A szabályok, stratégiák és technikai változások követhető története.
Filtering, strategy and technical changes are recorded here. Earlier history is approximate; new entries have a date.
Only non-draw selections with signal-time odds strictly below 3.50 are eligible. Exactly 3.50 is excluded. Home and away selections on tied matches keep their original settlement. The raw sources and prior audit are preserved. The shared-window comparison is retrospective. The complete Form Radar entry gate cannot be verified from the retained evidence.
Official club announcements support exact sport- and league-scoped mappings for Kobe/Nishinomiya Storks, Reilac/Mio Biwako Shiga, UBSC/Eagles Graz, Fjellhammer/IL and TV Hüttenberg/TV 05/07 Hüttenberg. Women, youth and reserve teams are excluded. Original bet identities and closed history are preserved; two fresh independent finals and correct market arithmetic remain required.
Exact HC Ajoie / HC Ajoie Porrentruy labels in National League and Rungsted Ishockey / Rungsted Seier Capital in Superisligaen identify the same senior club selection, supported by official club information. Mapping applies only in this sport, these leagues and for exact full labels, excluding youth, women and reserves. Original rows and closed history are preserved; two fresh independent finals and correct server market arithmetic remain required.
In the Portuguese cup, the exact senior men’s labels Atlético CP and Atlético Clube de Portugal identify the same settlement selection, supported by the club’s own information. Women, youth, B teams and other Atlético clubs are excluded. Two freshly checked independent final-result sources, exact event/date, market scope and server arithmetic remain required. Original bet identities, selections and closed history are not rewritten.
Full-match Over/Under lines with reference data receive the priority scan lane across sports on the server and dashboard; other markets retain rotating coverage. With a small monthly budget, The Odds API uses three of four requests for totals and one alternately for spreads or winners, preserving credit pacing; normal budgets request all three. Full-match Total Games labels are supported. Browser SX matching shares exact-line validation; saved totals evidence uses the same line for both sides. Budgets, EV and quality gates and history are preserved. Priority is a user research preference, not validated future performance.
Bookmaker comparison matches full-match totals by Over/Under direction and the exact line rather than team-name similarity, in reference consensus and Stake selection alike. Bare direction labels include the line in saved selections. Different lines, partial periods and team totals remain excluded; EV, freshness, quality gates, notification filters and existing history are preserved. This corrects matching and does not validate future returns.
Tracker Stake links use saved event identities up to one hour old. Expiry of the shorter refresh cache no longer starts full fixture pagination alongside Tracker; missing link metadata does not block the ledger or results. Future, non-live event checks remain. This is link metadata, not fresh odds: the EV view and server scanner retain their normal data path, quote timestamps, freshness gates, quotas and complete history.
Tracker data loads before EV statistics, CLV analysis, the FormRadar test tracker, mail retries, email recovery and supplementary reads. These features remain available without competing with the initial event-list request. The header reports Tracker loading/error status, its last successful read and one-minute read countdown; a loaded list no longer waits for hidden odds sources. Five-minute full source refreshes and focus listeners run only in the EV view, and opening Tracker starts no source-health probes. Manual refresh, half-hour server scans, history, formulas, filters and notification conditions are preserved.
The Tracker checks completed historical repair markers in one database read; missing repairs retain their original order and exact-row guards. The dashboard requests only quality fields used by its interface; full evidence remains in the database and full API response. Public result charts fetch their strategy data without a duplicate full tracker list. Opening Tracker no longer starts hidden EV-source downloads or a separate browser settlement run; the fixture catalogue for match links loads after tracker data. First opening the EV view loads its sources, while manual refresh and the existing half-hour background checks remain. History, quotas, filters, strategies and settlement calculations are preserved.
After delivery, the existing minute Telegram callback can continue remaining odds checks from a completed half-hour cycle. Completed cycles may continue for up to 24 passes and 25 minutes, releasing the lease before the next half-hour start; each callback has a forty-second budget. It reuses prepared references and the cached fixture catalogue, rechecks original quote freshness, and starts no new reference downloads, model research or email digests. Atomic pass claims prevent duplicate work. No new external scheduler is needed. Quality, sport, odds, EV, risk and notification conditions, quotas and history remain; more checks do not guarantee more eligible alerts.
Live reference and odds comparison recognizes exact identity after a leading AC/PFC/HNK/GNK/CSD/FCI club prefix when at least two name tokens remain, or an explicit parenthesized leading poll rank from 1 to 99. Examples include CSKA Sofia/PFC CSKA Sofia and ranked NCAA teams, without lowering thresholds or guessing partial names. Gender, youth, reserve, sport, kickoff, simulated-event and ambiguity guards remain. ASCII-hyphen and vs fixture separators are supported while typographic separators take precedence to preserve hyphens inside club names. Historical/result normalization, settlements, quality/EV/odds rules and history are unchanged.
The four-part reference preparation no longer needs a completed fixture catalogue: matching happens later in the odds step. Its first provider group therefore starts alongside fixture pagination. The catalogue network budget increases from 20 to 40 seconds within the reference step’s existing 45-second envelope, so sequential downloads no longer consume the expanded football pagination budget. Genuinely incomplete catalogues remain marked partial; no complete-coverage or signal-volume guarantee is made. Quotas, quality and notification rules and history remain intact.
The shared upstream cache preserves original Date, Age and stale-response headers. Evidence for an unchanged open Unibet/Kambi price therefore survives both the source snapshot and shared network cache without renewing its timestamp. Original price changes, prices, quality limits, quotas, history and notification rules remain intact.
Unibet/Kambi last price-change time is separated from a fresh observation of its complete open-offer listing. HTTP Date/Age evidence within five minutes, no stale-response warning, a complete OPEN market and the original changedDate are required; missing or old transport evidence retains the previous price-time policy. Reading cache does not renew observation time. Original price changes, response observation evidence, regulation/overtime scope and bookmaker identity remain in the audit. A scan may resume across at most eight separate bounded odds requests within the same ten-minute lease, without new provider refreshes and with quote freshness rechecked on every pass. Scheduler source consumes the remaining queue; the external scheduler needs separate publication. Already attempted failures are counted separately rather than becoming an endless deferred queue. The background status separately displays upcoming fixtures, usable references, comparable fixtures, actual odds checks and the remaining queue. Football pagination grows from 1000 to 1600 raw fixtures; a catalogue whose end is not reached remains partial. A/B quality, EV, odds, strategy and notification rules and complete history remain intact; more signals or future returns are not promised.
The 17 reference adapters are distributed across the existing data, hockey, esports and settlement steps, retaining half-hour scheduling and provider quota protection. Responses are consumed on arrival and fallback snapshots read individually; all four groups combine only within the same cycle before the shared odds scan. An expired ten-minute execution lease displays an interruption instead of perpetual running. Owner diagnostics separately expose private Telegram and channel delivery queues and their five latest actual delivery records. This does not send a test message or invent a betting signal; quality, odds, strategy and notification rules and complete history remain intact. No increase in admitted signals or future returns is claimed.
The V2 market → participant → line → bookmaker-price structure is now decoded. The previous reader expected lines directly beneath each market and produced no references even when fixtures were returned. Winner, spread and total markets preserve original bookmaker identity, signed lines, overtime scope and per-market price-change timestamps. An old handicap neither refreshes another quote nor hides a genuinely fresh market. Existing 15/60-minute limits, suspended/incomplete-market exclusion and independent-bookmaker rules remain. No paid key or subscription is introduced; the provider’s own free allowance remains a limit. Tracker history, settlements and saved notification settings stay unchanged.
Quotes outside the existing 15/60-minute windows are removed before expensive name matching. Fresh single-book inputs remain so independent providers can still form consensus; the unchanged A/B gate runs after merging. Daily statistical model refresh moves to the existing model-scan stage instead of sharing the full reference-preparation request. Model calculations, daily refresh limits, provider budgets, strategy rules and notification filters stay unchanged; tracker history and settlements are preserved. The split reduces resource contention without promising more signals or returns.
A new keyless Unibet/Kambi match-winner reference joins background scanning and the dashboard, with seven sport routes and a shared five-minute cache. Only future, non-live, non-simulated fixtures with complete open main markets qualify. Original price-change timestamps are required; existing 15/60-minute freshness and A/B-quality limits remain. Regulation and overtime are explicit, and different settlement scopes cannot create consensus or a false market match. Unibet quotes delivered by other providers still count as the same bookmaker. TheRundown sport and league rotation now follows the half-hour cadence: the former five-minute rotation could repeatedly select the same lane when sampled every half hour. At most two sport/date requests per cycle remain. Explicit billing resets are rechecked at most every six hours; minute 429 throttles retain short waits. No paid plan, subscription or quota increase is introduced. Tracker history, settlements, strategy rules and saved notification preferences remain intact. More sources do not guarantee more positive signals or returns.
Best sports & markets and Profit / drawdown now select named sports and one common broad odds range. There is no all-sports market-family fallback or tiny market/odds cell set. Five fixed broad ranges are evaluated without searching EV thresholds. Each sport needs at least 12 settlements and half the combined minimum, positive overall and latest-30% results; combined minima are max(24, 10% of training data) for the third profile and max(24, 15%) for the fourth. Risk-managed baseline remains default; its stakes and Highest staked volume are unchanged. PuntersEdge replaces a twelve-hour odds cache with at most one shared sport request per hour, rotating sports and leagues, requesting three markets with a five-minute upstream cache. Over a full 31-day month this is at most 2232 odds credits plus the daily sport catalogue, without increasing provider quotas. Monthly credits are measured separately from minute limits. Original quote timestamps and admission quality limits remain; more signals are not promised. History, settlements and saved notification settings are preserved; revised profile rules apply to new signals. Optimized performance is a retrospective backtest, not validated future returns.
Risk-managed baseline becomes the first and default profile across strategy views, homepage performance and notification choices. Its stakes and the Highest staked volume calculation remain unchanged. Best sports & markets replaces sport–market–odds cells with whole sports or one whole market family, without per-sport market/odds intersections or EV-threshold search. It requires at least 24 settled picks and 10% of training data, positive overall and latest-30% results, and retains the common odds/EV range and baseline 1 / 0.5 / 0.3-unit stakes. Profit / drawdown replaces Balanced Optimization and maximizes net profit / max(1 unit, maximum drawdown), with non-increasing stakes capped at 1 unit and at least 24 settled picks and 15% of the training sample. Positive results and genuinely different entry sets remain required; insufficient evidence is disclosed. The two revised cards show compact summaries with expandable rules. Existing profile IDs and saved notification selections remain; revised meanings affect new signals, without rewriting tracker history or settlements. Updated curves retrospectively backtest the new rules and are not forward-validated returns.
Fixtures deferred by the first odds request’s time budget receive one additional processing pass of up to 45 seconds within the half-hour cycle. Continuation reuses only the same cycle’s prepared references younger than five minutes; it does not refresh reference providers or retry fixtures already attempted in the first pass. Original timestamps, exact market identity, positive EV, A/B evidence and risk limits still apply. Actual counters from both passes are combined and remaining deferred fixtures remain visible. Existing tracker history and settlements are preserved. This improves coverage without promising more tips or better returns.
Every member receives a referral link in their account. Referrers earn 10% of actual Stripe and verified crypto fees from new email-verified signups, including renewals. Refunds correct rewards; disputes freeze them. Tests, free access and self-referrals earn no commission. Bank payouts can be requested after 30 days using encrypted IBAN details; admins record completed transfers. The interface does not initiate money movement. Existing registrations are not retroactively attributed; tracker history, odds, EV and the fixed one-unit reference remain intact.
Checked research destinations for AFC, UEFA, Maia Challenger, Shanghai and DreamLeague S30 China qualifiers are reused for those competitions. A source list is not final-result proof: exact date, team variant, completed full match and two independent sources remain necessary. Live and date-mismatched events stay open; old scores are never copied to new fixtures.
Verified result settlement recognizes Esteghlal FC/Esteghlal FC Tehran in AFC Champions League and Sabah FC/Sabah Masazir in UEFA Champions League as the same first-team selection. These aliases apply only to exact base names within those competitions; women, youth, reserves and other clubs remain distinct. Two independent final-result sources and date, market, handicap and score checks remain required. Original records and historical settlements are preserved.
Telegram direction protection and previous-pick warnings use the sport and stable fixture ID when a valid Stake event link is available. The Hurkacz–Davidovich Fokina kickoff change from 13:10 to 14:50 therefore does not open a new direction. Derived direction indexes are refreshed from original delivery snapshots, with private and common-channel sends sharing protection. Related-pick warnings use the recipient’s own accepted or uncertain delivery history even if the tracker row was archived or its kickoff changed. Without a fixture link, exact kickoff and participants are still required; rematches with different Stake IDs remain separate. Historical message content, send times and tracker results are preserved.
The exactly identified pending October 12 2 de Mayo–Sportivo Ameliano bet is relabelled from Apertura to Clausura using two completed match reports. Its original label and checked sources remain in evidence metadata. Only the still-pending row matching every identity field and odds can change; odds, EV, settlement and all other historical data are preserved.
Full-match winner/1X2 and handicap signals with the same sport, participants and exact kickoff share a team direction across private bot recipients and the common channel. After a sent, sending or uncertain notification, new picks on another team or the draw cannot be sent. Over/Under and BTTS yes/no have separate direction groups; opposite directions are conservatively blocked even across different lines. Different markets/lines on the same side remain eligible with the existing related-pick warning. Atomic database claims protect both same-message and competing sends. Known failed or merely queued deliveries do not lock a direction; unresolved full-match selections are held. Protection derives from the original payloads and times of existing sends; where both sides were already sent, further opposing notifications are held. Half, quarter, set, map and player markets do not inherit the full-match direction. Intake order takes priority over newly arriving picks. Only derived protection metadata and reasons for newly blocked notifications change; prior messages, odds, EV, bets, settlement and retrospective strategy results are preserved.
Private and common Telegram messages containing multiple signals separate each complete match card from the next with a bold horizontal line. Each card keeps its pick, match link, related signals and stakes together; separator length counts toward the 3900-character sending limit. Previously sent messages, filters, duplicate protection and settlement are preserved.
Private and common Telegram signals warn about different picks with the same sport, participants and exact kickoff. Related picks already sent to that recipient and in the current message are shown; uncertain earlier sends are labeled separately. The match-specific Stake link is prominent directly below the pick. Filters, duplicate protection, strategy stakes, settlement and history are preserved; the warning does not quantify correlation or confirm a placed bet.
Financial, observation, pending and settled tracker counters render directly in all ten UI languages, so refreshing counts cannot leave a Hungarian observation label in English. Mobile and table statuses and the observation filter use the current locale. Counts, filtering, settlement and history are preserved.
Odds strategy analysis renders verdicts and composed sentences in the selected language; all six verdicts cover all ten UI languages. Missing English copy on the home page, automation, FAQ, EV bands and payment interface is added. Stake pagination retains only event fields and shared provider fetching reads each response once without buffering an unread clone. Coverage and freshness rules remain intact. Result search links include the event date; local sites from unrelated leagues no longer crowd out relevant league and general result sources. Telegram cards show sport, market and a separate strategy stake block. The October 10 open San Lorenzo–Trinidense pick’s incorrect Apertura league label is corrected to Clausura using APF and VS Sports evidence, preserving the original label and an audit entry; settlement, odds and EV are not changed by this metadata repair. Filters, dispatch timing, duplicate protection, email opt-outs, settlement evidence requirements and tracker history are preserved. Strategies remain retrospective backtests updated from settled results, not verified future profit.
Results, Automation and the strategy audit now show excluded odds ranges alongside selected bands, scoped to the relevant sport and market. Lower bounds are excluded and upper bounds included. When there are no gaps within the shared 1.10 < odds ≤ 5.00 range, that is shown explicitly; odds outside it always remain excluded. EV remains a separate filter. This presentation clarification changes no picks, stakes, results or notifications; optimized profiles remain retrospective backtests.
The active baseline replaces flat 1 unit with 1 U for 1.10 < odds ≤ 2.50, 0.5 U for 2.50 < odds ≤ 3.50, and 0.3 U for 3.50 < odds ≤ 5.00. Summaries, charts, analyses and new notifications use these stakes. Volume caps stakes at 2 U; balanced at 1 U with stakes tapering as odds rise. Available profiles differ pairwise by at least 10% of their combined selected settled entries and at least 5 picks; simply raising stakes does not create another strategy. Without a qualifying distinct positive sample, a profile produces no entries. Filtered profiles remain retrospective optimizations, not validated future profit. Original odds, EV and normalized settlement profit are preserved; existing strategy subscriptions follow the new baseline without resending old emails.
The previously confirmed Bodrum, Beveren and Desenzano aliases also apply when selecting the 1x2 winner. Similarity thresholds and youth/women/reserve distinctions remain intact. This change alone does not alter closed bet statuses.
Source links from older manually settled bets can also be prioritized using the stored league, normalizing accents and punctuation. Settlement recognizes the former professional Desenzano Calvina name as Calcio Desenzano while preserving youth, women and reserve guards. Old scores are never applied to new events: each new settlement still requires two fresh independent final sources and correct market arithmetic. Existing settlements, odds and EV data are preserved.
The former professional Bodrum Belediyesi Bodrumspor/Bodrumspor name is recognized for Bodrum FK handicap settlement while youth, women and reserve labels remain protected. Successful manual writes remember every supplied result source by league and prioritize those sites in subsequent result searches, including machine tracker reads. Existing sport sources remain available; an old score never proves a new event. Two fresh independent final sources and market arithmetic remain required for each new settlement. Closed history and odds are unchanged.
Detailed and 10% EV bands show mean and median odds beside each current strategy’s settled sample count. Odds mean and median count each bet once, without stake weighting. Expected net EV units sum strategy stake × recorded EV/100; the difference is actual net result minus expectation on the same binary, standard-payout sample. Push/void, half wins/losses, cashouts, nonstandard payouts and invalid implied probabilities are excluded from that comparison; partial coverage shows its own count and actual result. Bands load independently in parallel with the full tracker, skipping repeated historical repairs, observation audits and unrelated aggregations. Profiles and bands share one response sample; failures offer retry and mark retained results. Stored odds, EV, settlements, strategy rules and Telegram/email settings are preserved. This is a descriptive check of recorded EV and retrospective current-profile analysis, not proof of a positive edge.
Basic mode shows simple strategy and notification selection. Telegram markets by sport, conditions by strategy, odds/EV filters and quiet hours are editable only in Advanced. Switching modes and saving in Basic preserves existing detailed conditions. Server-side email opt-outs can be extended without replacing the original list; opted-out addresses receive no new signals and unsent queue entries can be closed. Telegram links, strategies, entitlements, Tracker history and already accepted mail remain unchanged.
PinnWire replaces the previous sport rotation with at most two shared requests per half-hour across visitors and the scheduler: one sport with upcoming Stake fixtures and one oldest sport. Cold startup no longer bursts through every sport; 96 scheduled requests per day fit the verified 100-request free allowance. Original quote timestamps remain. Browser and background scans prune market cohorts that cannot meet existing freshness and independent-source quality gates before fetching Stake prices. A stale attached bookmaker no longer downgrades independently fresh evidence for a near-start fixture; the existing two-source, one-hour allowance for later fixtures remains. Team, market, signed line, odds, EV, risk and notification checks still apply. State diagnostics separate cached inventory from eligible references, raw positive EV from quality-approved EV, and admission candidates from actual tracker inserts, with reason counts that may overlap. This does not guarantee daily alerts, create missing upstream entitlement or rewrite history or settlements. Existing layout, filters, strategies, delivery cadence, credentials and access are preserved.
The existing data stage prepares bookmaker and exchange references, so refetching all providers no longer consumes the separate odds collection window. Preparation is usable only in the same refresh cycle for five minutes; original quote timestamps and quality gates remain, with a short separate window for refreshing old exchange prices. Stake fixture pagination visits every sport’s first page before deeper pages and prioritizes previously failed sports. Partial reference lists retain one-day scan progress; failed fixture requests no longer keep the same event at the front forever. Internal odds requests avoid rereading the full fixture catalogue per event. The state response reports matched, attempted, successfully scanned and quoted counts for every sport; saved prices and tracker admission candidates have separate counters. HTTP-200 quota errors and API-Sports minute limits start the appropriate recoverable cooldown, while daily quotas do not suspend every API-Sports sport. API-Sports free or unknown daily plans request today/tomorrow odds; a confirmed larger daily allowance retains the four-day rotation. ParlayAPI odds requests run sequentially at least one second apart, stop on access errors and report their cause; missing quote timestamps never become fresh data. This does not create provider entitlement or renew exhausted allowances. Keys, access, interface, dispatch cadence, filters, strategies and tracker history are preserved; EV and settlement requirements are not relaxed.
Advanced automation gives selected email strategies their own odds, EV and markets-by-sport conditions. Enabling copies common choices; custom filters then override that strategy’s common filters while retaining its own rules. One selected matching strategy is sufficient and each pick is sent once. Background intake, browser intake and pre-dispatch filtering follow the selection; delivery reconciliation uses the same conditions. Existing settings, already rendered messages, Telegram settings, strategy formulas and tracker history are preserved. This customizes alerts and makes no new performance promise.
The landing page and dashboard default to English when no language preference is saved. English is the first language option; Hungarian remains selectable. Existing cookie and browser language preferences are preserved. Tracker data, signal filters and notification languages are unchanged.
Football handicap settlement recognizes the club-confirmed Waasland-Beveren → SK Beveren rename. Youth, women and reserve variants remain distinct; KSK Beveren is not mapped. Two fresh final-result sources, signed handicap, market scope and server recalculation remain required. Previously closed bets are not changed.
MCP manual settlement no longer appends text after FINAL_EVIDENCE JSON, leaving evidence readable for later verification. Refunds retain their source link. English strategy and performance-curve copy is completed; localized accessible skip links move to the shared language module with unchanged text. Independent final sources, complete basketball periods, tennis game totals, market recalculation and closed-row protection remain required. Existing tracker history, external provider allowances and access are unchanged.
The desktop main menu returns to the same row as the logo, language selector and Dashboard button, removing the separate lower menu row. From 1024 pixels the central menu scrolls horizontally when needed, keeping every link reachable without expanding the header or page width. Active-page indicators, focus, translations and the mobile dropdown remain. Per-sport notification filters, access, strategies and tracker data are unchanged.
Advanced automation supports multiple markets per sport for email recipients and browser alerts. Private Telegram supports the same selection in common and per-strategy conditions. Enabled mode allows only checked sport–market pairs; empty sports and an empty overall selection send no alerts. It replaces that scope’s common sport/market filters while retaining odds, EV and strategy conditions. Persistence, background intake and pre-dispatch checks share filtering; totals exclude game/map/set handicaps and winner markets. Existing settings without the new field retain their behavior. Already rendered email payloads, the shared Telegram channel, access, delivery schedules, strategies and tracker history remain unchanged.
The four current profiles are ordered: Balanced optimization, Best sport–market–odds segments, Fixed 1 unit, Highest staked volume. The main Results/dashboard strategy chart, its bonus calculator and the analysis selectors initially choose the optimized profile. Strategy choices and notification settings use the same order. All four profiles retain their rules, measurements, calculators, tracker history and saved alert selections; the fixed one-unit control remains separately measurable. Only ordering and initial presentation change; retrospective optimization is not validated future performance.
The landing page replaces the fixed one-unit reference chart with the net-unit curve of the current Balanced optimization profile shared with Results. The chart, net result, settled sample and ROI use variable stakes and filters from the same response, refreshed every minute. A missing profile shows a loading error rather than substituting another strategy. The curve retrospectively backtests current rules; it is not executed betting history or validated future performance. Tracker history, the fixed reference, strategy calculations and notifications are preserved.
From 1024 pixels the desktop menu shows every link in a separate header row that wraps when necessary. The logo, language selector and dashboard link remain in the row above so longer translations do not push controls out of view. The expandable mobile menu appears only below 1024 pixels. Active-page indicators, languages, keyboard focus, links, previous account-security fixes and the other workflow’s changes remain. Only shared navigation layout and its general accessible label change.
All four strategies are independently selectable, with shared or per-strategy sport, market, odds and EV conditions. Matching any selected strategy is sufficient; each pick is notified once per person. Browser/background Tracker saves and emailed Tracker recovery durably queue notifications for all matching active linked recipients before delivery, without a five-member limit and independently of a running dispatch lease. Private dispatch then starts immediately, prioritizing new picks before common-channel delivery. Private bot messages no longer require joining the common channel. Saved personal choices, language, quiet hours, access and deduplication protections remain; existing background checks retry provider limits. The channel’s common fixed reference, half-hour email digest, every existing schedule and Tracker accounting remain unchanged. Immediate dispatch initiation does not guarantee delivery or reading time.
General sign-in, registration and code-login labels are available in ten languages. A validated, one-year SameSite preference cookie passes the selected locale to the server, so the login page, header and footer links appear in the saved language in the server response. Legacy local-storage-only preferences migrate on first load and may still show a one-time language switch. This release does not claim complete translation of every existing page or server error. Code-mode initial state comes directly from the URL; mode switches expose accessible pressed states. Password fields accept up to 256 raw code units while the server’s 15–128 normalized-character rule remains. Post-login return paths also reject backslashes and control characters that browsers may normalize to external destinations. Accounts, sessions and history are preserved.
Provider half, quarter and period statuses are rejected even when accompanied by a completed flag. API-Sports and BALLDONTLIE must supply exact terminal match statuses. Basketball needs two separate agreeing providers plus at least one complete period list summing to the final score, including NCAA men’s two halves and applicable overtime. Open bets expose 6/12/24-hour milestones; after 24 hours pending bets enter audited manual review independently of provider success, without a settlement time or profit. Automatic searches continue there. Existing hourly scheduling and manual-queue retry timing remain; there is no duplicate schedule, record deletion or rewriting of previously closed results.
Login checks password length after the same NFC normalization used during password creation, keeping accepted long decomposed Unicode passwords usable. Legacy PBKDF2 hashes with unexpected trailing fields are rejected. Valid existing hashes, accounts and sessions remain intact; no password reset or historical data rewrite occurs. Odds, EV, strategy, billing and signal-delivery behavior are unchanged.
The full horizontal navigation appears from 2200 pixels with a wider header container. Smaller screens retain the scrollable expanded menu so longer French and Dutch labels do not push controls outside the viewport. Links, content, access and historical data are preserved.
Menu open/close, mobile navigation, language selector, appearance, contact, development log, general privacy/terms links and theme states gain dedicated labels in ten languages. Shared navigation renders the selected language through React; the language selector has a visible keyboard focus indicator. This is a general interface fix, not a claim that all page content is fully translated. Historical data and odds, EV, strategy, billing and notification behavior are preserved.
We checked 87 available daily LiveScore archives across four sports. Another 110 historical bets have completed scores agreeing with a previously inspected separate provider, preserving their correct status and profit while fixing 13 missing scores or metadata. The September 19 Beer Sheva–Kiryat Shmona 1–0 FT result is corroborated separately: Kiryat Shmona Over0.5 changes from WIN to LOSS, a −1.56-unit correction. 34 additionally reviewed provider-name aliases and twelve specifically reviewed club, league and news proofs further document history. Full prior-state guards and audits preserve original dates, odds and EV. Same-name men/women or senior/U21 fixtures, lone sources and unfinished results cannot justify unverified overwrites. Gaps in complete independent historical authentication remain explicit.
The shared public page footer uses a brighter base text color so its 14-pixel navigation and contact links are easier to read against a dark background. Link destinations, wording and wrapping are preserved. This is a general presentation fix; data handling, calculations, billing, notifications and history are unchanged.
Two separate result pages corroborate the September 18 Nantong–Guangxi 0–1 FT result: Guangxi Over0.5 changes from LOSS to WIN at +0.42 net units, a +1.42 correction. Yanbian–Guangdong 1–2, Rapid–Tirol 3–0 and Laval–Sochaux 1–1 receive two-source documentation with unchanged correct profit. Full prior-row guards preserve original timestamps, odds, EV and audit. Generic Polymarket types cannot substitute half, 45-minute, quarter or period questions for full-match bets. The precise origin of older unsourced errors is not asserted without evidence; full historical authentication remains unfinished.
The shared public page shell provides a keyboard-accessible content shortcut in ten languages. Escape closes the expanded menu and restores focus to its button. Internal links preserve browser behavior for Ctrl/Cmd/Shift clicks, separate windows, downloads and external URLs. Existing mobile layouts and page content are preserved. This is a general navigation fix; odds, EV, strategy, billing, notification behavior and historical data are unchanged.
The completed September 18 Bandırmaspor–Ümraniyespor 0–1 result is corroborated by the visiting club and a separate Ajansspor report. Ümraniyespor Over0.5 changes from an unsourced, score-free LOSS to WIN at +0.46 units, a +1.46 correction. Arema–Persik 1–1 receives official league and club documentation; Persik Over0.5 stays WIN with unchanged profit. Both rows require their complete prior state, preserve original timestamps, odds and EV, and audit once. Independent authentication of other incomplete historical results remains a separate task.
The two-source historical evidence bundle gains 24 rows, for 259 in total. The extension introduces no further financial correction; full prior-state guards preserve previously repaired and later edited rows. Asian Total now calculates the full-match total, retaining team totals and quarter-line settlement. Simulated or virtual basketball identities in provider event IDs cannot merge into real-event quotes or match a real Stake fixture. This remains partial historical authentication; missing old evidence is not represented as acquired.
Two retrievable sources document the completed result of 235 identified historical bets. Ružomberok, Orange County and Birmingham team goals Over0.5 become LOSS; Miami match winner in Toronto’s 105–129 NBA game becomes WIN. The four financial corrections total −1.89 units; other score/source backfills preserve profit. Full prior-row guards protect later manual edits and all changes are audited. This is partial historical authentication, not a claim that every older row has been proven correct. Result matching also requires agreement on youth, women, reserve and simulated variants carried in league identity.
Cairns +7.5 in Sydney’s 111–90 result is LOSS; Denver −6.5 in the October 4 97–109 Utah game is LOSS. These corrections total −4.48 units; the related Under 237.5 remains WIN with correctly oriented scores. Only the three rows matching their entire old state can change, with two final-result sources and audit records. Automatic basketball score settlement waits for agreement from two separate result providers; repeated copies of one feed or unverified sibling bets cannot count as the second source. A verified two-source manual result with period sums and an exactly resolved Polymarket market remain usable. A lone feed leaves the match pending for the existing later verification.
After a calculation audit of the entire 960-row tracker, four closed game handicaps for Harris–Hurkacz, Maristany–Grabher, Dietrich–Searle and Wang–Prozorova receive verified game totals instead of set counts: 10–13, 8–13, 13–8 and 15–16. Win/loss status, profit, odds and original settlement time are preserved. Only the four rows matching their entire previous state can change, with an audit record; later owner edits and repeat runs are protected. This is a calculation audit, not fresh independent verification of every historical final result.
Active linked accounts can request a test message to their own private bot chat from notification settings. The recipient comes from the saved server-side account link, never a supplied address. The manually requested test uses the saved language regardless of quiet hours and saves neither settings nor bets nor tracker results. At most three requests are allowed per ten minutes. The interface confirms Telegram provider acceptance, not that the message was read. The members channel, filtered alerts, access rules and history are preserved.
The homepage primary and result buttons open public results directly; the paid dashboard no longer interrupts the initial preview. The sitemap includes existing public signals, results, how-it-works, contact, responsible-use and development pages. The FAQ now describes the existing calculator’s baseline rakeback estimate and illustrative weekly/monthly ranges separately from measured tracker profit. No new promotional sending, visitor tracker or future-profit promise is introduced; prices, access, signal rules and historical settlements are preserved.
Four exactly identified prior settlements are repaired: Sacramento +11.5 WIN (Lakers–Kings 114–110), Lakers -6.5 LOSS, Cleveland–Boston Over 222.5 WIN (113–124), Fortaleza +0.5 WIN (1–1). Two older tennis game-handicap scores are corrected from set counts to verified game totals; partial scores are cleared for one receipt-confirmed refund. Detroit–Phoenix score orientation is also corrected to 109–107. Their statuses and profits are preserved. Original partial scores and incorrect completion claims remain in the audit; repairs require the exact original erroneous row state. MCP manual writes require two freshly checked final-result sources, explicit completion evidence and server-side market arithmetic. Basketball period totals and tennis set-game totals are checked; the VOID refund path remains. Old manual scores lacking separate completion evidence cannot automatically settle sibling bets. Tennis and esports sibling results must also match the market. Source diagnostics distinguish plan restrictions from quota and preserve rejection reasons; external allowance or authorization is not bypassed. Signal filters, notification dispatch and strategies remain unchanged.
The payment return page shows server-verified access, with a retryable check while confirmation is pending. Cancellation preserves the chosen plan. Active Stripe access shows subscription management instead of another purchase, and the server rejects a new Checkout request. Price validation also checks the recurring interval. Same-second cancellation revokes access, and a late invoice cannot reactivate a canceled subscription. Sandbox requires separate test credentials and a strictly local return URL; live configuration is unchanged. Provider copy no longer claims registered sole-proprietor status. Tracker history, odds, strategies and past payments are preserved.
The homepage signal preview shows its last successful load in Budapest time; this is a view check, not evidence of each quote’s freshness. Failed refreshes get a warning while retained signals are explicitly labeled; initial loading failures are not presented as an empty market result. Lifetime pricing now describes access for the lifetime of the service, matching existing terms without promising perpetual operation. Prices, server-side billing and tracker history remain unchanged.
An incorrect English dictionary entry labeled USD as EUR in the currency selector. The entry is corrected; shared translation preserves USD, EUR, HUF and supported crypto asset codes, including reused text nodes. Amounts, actual USD card charges, crypto networks and existing payments remain unchanged.
The homepage now includes Telegram and Pro email digests. Pricing removes obsolete pre-integration copy denying active payments and distinguishes recurring card subscriptions from prepaid crypto access. The Telegram activation warning appears only after a verified disabled configuration, never while loading or after a failed check. Both pages show a three-step account–Telegram–filter guide. Risk copy is more legible, the performance curve has English labels, and keyboard focus and mobile bookmark icons are provided. Prices, payment entitlements, tracker history, EV calculations and alert filters are unchanged.
Automation now shows per-account counts of pending, forwarded and review-needed private Telegram signals. Status refreshes every minute without overwriting edited filters. The page distinguishes dispatch after saving from the half-hour email digest.
New accepted tracker signals start Telegram dispatch as separate background work after durable saving, without delaying the browser response. Private alerts are grouped per recipient into formatted messages up to 3900 characters; the dispatcher checks 50 pending records with a separate twelve-second private-delivery budget. A database lease and atomic record claims protect overlapping runs. Filters, active access, quiet hours and pending event status are rechecked before sending; uncertain acceptance is never automatically resent. The independent scheduler separates minute Telegram dispatch from half-hour provider refreshes and email digests.
Private bot and members-channel alerts use sport emoji, bold underlined picks, separate odds and green EV lines, readable Budapest kickoff time, source and event link. EV is explicitly an estimate, not guaranteed profit. External fields are safely escaped and bounded without cutting HTML entities. Deduplication and filters are unchanged; existing messages are not resent.
The contact page EV-risk heading now appears in English in the English view.
Resumed crypto invoices display their own token and plan in the selectors. Verified deposits direct customers to My account for current access, so expired or revoked access is not described as active.
Shared-channel delivery and durable queue preparation can no longer exhaust the private bot’s delivery budget. Preparation and private delivery each have a bounded eight-second budget, with at most ten pending messages per cycle. Filters, plan permissions, kickoff and data-quality rules, equivalent-outcome deduplication and uncertain-acceptance protection remain intact. Previously sent alerts and tracker history are unchanged.
Deposit-address reads retrieve the asset’s existing addresses, then the server matches only verified names for the selected network. Deposit availability and token contracts remain checked against MEXC configuration codes. Empty or mismatched responses cannot produce payment instructions; existing invoices, credited-payment evidence and tracker data are preserved.
USDC/BEP20, USDT/TRC20, native BTC, ETH/ERC20 and SOL receive account-bound exact payment requests valid for 30 minutes using fresh USD rates and an invoice-identifying discount of at most USD .05. Access requires a credited MEXC deposit matching asset, network, receiver, memo, exact amount and time; transactions and invoices are single-use. Prepaid 30/365-day access has no automatic charges and does not cancel Stripe subscriptions. Mismatches and delays require support review. The existing refresh cycle checks pending orders even with the browser closed; bot and email follow the same plan permissions. A contact page, Telegram/Facebook/TikTok links and more precise EV, variance, withdrawal and privacy notices are added. Complete legal operator details still require finalization. Existing tracker, Stripe and notification history is preserved.
Production billing also accepts a restricted Stripe key with the necessary permissions. Test and publishable keys still cannot activate live payments; merchant approval, live prices and authenticated live events remain required. Existing subscription and tracker history is unchanged.
The dashboard allows up to 90 seconds for the initial paginated 72-hour Stake catalogue across sports. Concurrent refreshes share the same in-flight request. Partial or failed coverage remains visibly marked; odds, EV, signal filters and historical tracker data are unchanged.
Stripe sandbox keys and signed test events use separate configuration and an endpoint that never grants paid access or rewrites production billing history. Live payments still require live credentials and merchant activation. A private members channel receives shared verified market signals through its own durable queue; personal filters apply to private bot messages. Only signals observed after activation and still eligible before kickoff can be broadcast. Equivalent winner/minus-half outcomes produce one alert, and uncertain acknowledgment is not automatically retried. Admission requires active access and a short-lived identity-bound linking code. Existing tracker, email and billing history is preserved.
The owner tracker connection now accepts an explicit VOID value for pending or review bets. Confirmed refund or bookmaker-rule evidence and a source link are required; both scores remain empty rather than inventing a final result. Saving uses the existing refund status, zero units and a VOID audit note. Closed or changed bets remain protected; historical calculations, filters and notifications are unchanged.
After a recognized transient network or internal D1 failure, a SELECT read can retry once after 150 ms. Persistent failures remain visible; writes, counter increments, transactions and sends are not retried by this guard because an operation may already have committed despite uncertain acknowledgment. Access rules, betting history and notification history remain unchanged.
Existing mail outbox and delivery audit tables are adopted into the official schema. The backward-compatible migration creates missing tables only; prior mail and evidence are not rewritten. Repeated table-creation requests are removed from delivery and retry paths. A transient queue failure returns an explicit 503 and defers retry; already accepted mail still cannot be resent.
The visible dashboard rereads tracker, history and source diagnostics every minute and on return. Historical CLV and the open availability tester’s timing and price data follow the same updates. Failed background reads retain prior data with a warning; the countdown renders separately and larger analyses retain calculations for unchanged inputs. Existing five-minute odds-provider and fifteen-minute model cycles are retained. USD 20/month Early Access includes Telegram; USD 50/month Pro includes Telegram and email. Manual access, admin and a valid introductory Pro bonus retain email eligibility. Telegram linking, subscription-gated group admission, personal filters and a durable delivery queue are prepared; without configuration the interface shows activation pending. Equivalent winner/minus-half handicap outcomes produce one Telegram signal; uncertain Telegram acceptance is not automatically retried. Tracker history, settlements and strategy calculations are preserved; historical optimization remains a backtest, not validated future performance.
The outer CLV quality dashboard panel in the tracker now also starts expanded and can still be collapsed and reopened by its header. The inner historical CLV summary already started expanded; both levels are now immediately visible. Calculations, price data, historical bets, ordering and notifications are unchanged.
The full performance tracker shows settlement status, status controls and the initially expanded, collapsible event list after its Market EV, Model EV and Combined summaries. Strategy, sport–market, sport, market-type, odds, EV, CLV and signal-source statistics follow below, preserving their relative order. Filtering, calculations, historical bets and notifications are unchanged.
The four-strategy EV table now defaults to detailed bands: 0–0.2, 0.2–0.4, 0.4–0.6, 0.6–0.8, 0.8–1, 1–3, 3–5, 5–10%, followed by ten-percentage-point bands through 100%. The previous 10% breakdown remains selectable. Lower bounds are included, upper bounds belong to the next band, and 100% to the final band; no double counting. Both views use the same baseline and current strategy rules, preserving total net/staked units and settled counts. Green/red indication, saved bets and notification rules remain intact. Dynamic profiles remain retrospective backtests.
The radar also shows only one row for archived equivalent full-time football winner and same-team Asian −0.5 signals: better odds, with 1X2 on ties. A fresh accepted signal takes priority over its archived equivalent. Original records, EV values, emails and settlement remain intact.
The historical CLV and unit-profit block starts expanded and can be collapsed. Its header retains positive-group net units with signed green/red indication and a label matching the selected CLV or price-movement view. Full sample, detailed breakdown, filters, calculations and data retrieval are preserved.
Full-time football winner and Asian −0.5 for the same team now share one raw win estimate for the same event and exact start time. Each actual bookmaker counts once: complete 1X2 within two minutes takes priority over that bookmaker’s AH price, while a genuinely newer AH can replace old 1X2. Complete signed quotes up to one hour old use proportional margin removal followed by averaging independent bookmakers. Existing freshness A/B/C gates and 75%/50%/15% caution adjustment remain; these are not validated true probabilities or future profit. Identical Stake prices, shared evidence and quality yield identical adjusted probability and EV. Offer selection uses better executable odds first; equal odds prefer 1X2, never a larger separately estimated EV. Missing quote evidence yields only conservative grade-C diagnostics, not a new signal. Different lines, periods, sports, teams and same-day rematches stay separate. Browser and notification recovery retain raw probability, quotes and the calculation audit; existing watchlist evidence is readable. Expandable calculations below fair odds show break-even, raw/adjusted probability, EV formula and shared reference. Historical sends, betting prices, EV and settled profits are not rewritten.
Browser notification requests pass already-recorded provider names and source labels into the durable digest. Existing attribution is no longer lost between browser and email. Missing historical provenance is not invented; sent emails remain unchanged and are not resent for this correction.
Dashboard matching shares gender/youth/reserve classifications for at most 5000 names. Previously calculated similarity for the same name pair is reused without reclassification. Cache size is bounded; match scores, exclusions, time windows and betting filters remain unchanged.
Live verification identified repeated work in reference merging as well. It now uses request-local prepared team names and sport/time groups, preserving first-match order, original quote times and independent bookmakers. The 80% merge threshold, six-hour window and gender/youth/reserve exclusions remain intact. Search budgets and deferred-fixture rotation are unchanged.
Each background scan prepares team names once and groups fixtures by sport. The 85% similarity and 15 percentage-point ambiguity thresholds, gender/youth/reserve exclusions and six-hour time window remain intact. Quote collection shares 35 seconds, reserving time inside the existing 50-second work window to persist every collected quote and final state. Deferred fixtures remain in the existing rotation. History, quality gates and notification entitlements are preserved.
Bronze–Obsidian levels receive automatic weekly/monthly estimates from the frozen Gold USD 3130 / approximately USD 15 weekly reference. The linear five-percentage-point rank adjustment is our scenario assumption, not a published Stake multiplier or measured tier average; its illustrative range is ±50%. Personal weekly references use ±33% and override the model baseline. Monthly falls back to a weekly-based ±50% approximation without sports-only data. USDT assumes nominal USD parity. Rakeback follows the official 3% × 3.5%; no fixed, reload, level-up reward or loss compensation is added. Bonuses are excluded from tracker profit. Availability net units, ROI and differences gain green/red/neutral indications; lower drawdown is favorable. Historical CLV moves above signal-source ranking and shows broader last-price movement coverage and missing-measurement reasons; the three-hour near-start boundary is retained. Three exactly identified October 7 open bets can settle from fresh official results: Piacenza winner and AH 0 WIN (1–3), Nybro winner LOSS (7–1). Teams, timestamp, sport, league, market and selection must match; only open statuses change and source links are preserved. Later manual corrections remain intact. Other data, strategy rules and notifications are unchanged. CLV subsets and optimized profiles remain retrospective analyses.
Search order updates from the shared settlements known before the day begins. A minimum of 20 bets and 12 events is required for a preference or penalty; preference also needs a positive final 30% sample. ROI weights are reduced for smaller samples. Three priority searches alternate with one exploration, half of exploratory places reserved for handball/water polo when matched references exist. EV, A/B quality, odds, stake and settlement gates are unchanged. This is retrospective research allocation, not validated future profitability. PinnWire combined volleyball/handball category 10 requires a unique Stake handball match, at most 15 minutes difference and at least 90% team similarity. PulseScore Bet365 handball and NetBet water polo feeds rotate with no more than one shared request every two hours. Existing sports stay in rotation. Original reference times are preserved and expired data cannot trigger new alerts; availability depends on the existing provider plan. More searching does not guarantee more signals. A full-time team win and Asian −0.5 share one notification identity for an identical selected club alias, even when the event caption uses a different club label. Different periods, lines, teams and sports remain separate. Sent emails and tracker history remain intact. Translation examines only changed HTML subtrees; chart updates and components that manage their own language do not trigger whole-page walks. Performance charts skip parent-driven rerenders when inputs are unchanged.
The dashboard starts with the tracker view open. Market EV, Model EV and Combined summaries come first, followed by low/medium/high odds strategy analysis, the custom sport–market explorer, sport and market breakdowns, detailed odds bands, EV bands and signal sources. The event list remains open. After measured profiles, the custom 2–6-step staking strategy precedes the availability tester; the FormRadar test tracker sits immediately above the CLV quality dashboard. Only initial view and layout change. Historical data, calculations, filters, subscription gates and notifications are preserved. Optimized profiles remain historical backtests.
Custom sport–market statistics use compact comparisons. Positive net units and ROI are green, negative values red and zero neutral across strategies and CLV. Signal cards, lists, details and new emails show recorded providers in badges; restored signals inherit existing provenance from an exact matching tracker event/time/market/selection; missing historical provenance is not invented. Strategy ROI plots are replaced by estimated accumulated rakeback, weekly and monthly bonuses and their sum for the full settled sample. Details group Monday-start weeks and calendar months in Budapest time without extrapolating partial periods. Push/void, cashout and missing/future timestamps are excluded. Bonuses use earlier personal turnover/bonus references; monthly falls back to a weekly-based approximation without a sports-only reference, and is not a verified Stake payment. Bonuses never enter measured profit. Previously queued or accepted email content remains unchanged. Settlements, strategy rules, notification conditions and history are preserved. Optimized profiles remain retrospective backtests.
The user clarified the weekly reference as 313 recorded baseline-eligible OddsRadar signals in the October 1–8, 2026 seven-day tracker sample × USD 10, or USD 3130 turnover, with approximately USD 15 weekly bonus. The denominator is frozen; later signals do not change it retrospectively. This is the automatic Gold/USD baseline; stake starts at USD 10, old blank turnover is migrated and custom values are retained. Without a separate sports-only monthly reference, the monthly card applies the same ratio to the strategy’s 30-day bonus turnover and labels it as a weekly-based approximation. The USD 50 monthly including other games is excluded; no VIP multiplier is invented for other levels. The +/-33% range is illustrative, not an official payout formula. Tracker profit, strategy rules and historical settlements are unchanged.
The four dashboard strategies, availability tester and custom stake ladder use the same complete server strategyBets sample as Results. Source statistics show only signal sources and sort either direction by net units, ROI, settled count, win rate and staked units; unknown rates remain last. CLV profit joins the complete odds archive to the shared settled strategy sample using exact event, sport, kickoff, market and selection, requiring an independent post-entry, pre-start quote. Near-start CLV requires a quote within three hours of kickoff; older last-price movement has its own view. Positive, negative, unchanged and unknown buckets preserve original settled profit and ROI. Positive-rate denominators include measurable bets only, with coverage shown separately. Full analysis replaces misleading CLV cards previously calculated from the truncated positive-odds list. Missing prices never manufacture historical CLV and extreme jumps stay excluded. The subset uses closing information known retrospectively, not a strategy executable at entry or a promise of complete coverage. Historical settlements remain intact.
Collected digests and pending outbox mail recheck the active recipient before sending; deleted or suppressed addresses cannot be retried. Server-side suppression applies to the recipient’s account and owner lists while historical accepted mail is retained. The calculator’s main view contains only unit stake, currency and VIP level. Staked units produce a baseline rakeback estimate and illustrative weekly/monthly ranges; weekly/monthly ranges are calibrated from the user’s previous turnover and bonus for that level. Turnover underlying the reported USD 15 weekly is unknown, so no rate is invented from it alone. Calculation details are collapsible and bonuses never increase measured tracker profit. Development log moves immediately before FormRadar. Sportmonks uses the existing key for targeted football result fallback with a six-hour cache: only FT/AET/FT_PEN states and unambiguous regulation scores are accepted, without mixing extra time or penalties into 90-minute scores. The secret MoneyLine key is activated in production while its existing six-hour quota-protected data path is retained. Dashboard copy also describes scheduled background notifications accurately. Shared strategy rows retain cashout/void notes so those bets cannot inflate bonus turnover. Historical settlements and strategy rules are unchanged.
OpenLigaDB joins football and hockey result checks with a 30-minute cache and separate regulation/overtime handling. OpenDota can use only decided BO3/BO5 winners for explicitly identified Dota 2 picks; one map or an unknown game cannot settle a bet. MoneyLine odds/results await a key: six-hour requests, first 50 odds and 100 event rows, original quote timestamps and observed credit metering; coverage is partial. Tracker filters and selectable strategies drive source/combination and result-source statistics with win rate, settled/open counts, staked/net units and ROI. Shared signals belong to one group without duplicated profit. Old rows retain only provable main-source attribution; new evidence also retains provider lists. Historical settlements remain unchanged; dynamic profiles are backtests.
Local API-Sports pacing has its own state instead of appearing as a provider quota failure. Successful result and status requests refresh access evidence without claiming complete odds coverage; older success cannot overwrite a newer provider failure. Expired membership or a disabled recipient also cancels interrupted sending states, and a concurrent failed attempt cannot restart them. Actual provider acceptance is preserved without resending. Email status distinguishes new acceptance from verified reconciliation of a previous 409 conflict. Tracker history, settlements, strategies and admission rules are preserved.
A new table shows net or staked units and settled counts in 10-percentage-point EV bands from 0 to 100%, using all four current profiles’ own rules and stakes. Lower bounds are included; upper bounds enter the next band, except exactly 100% in the last band. The full active settled sample is independent of list filters; pending and review picks contribute no financial result. The analysis updates with profiles and data and remains a historical backtest, not a future profit promise. Existing history and settlements are preserved.
The tracker keeps its earlier odds bands and adds 0.20 bands above 1.10 through 5.00, ending at 4.90–5.00. Bands do not overlap and share tracker/EV filters. Below the existing sport and market breakdowns, multiple sports and markets, inclusion/exclusion and odds bounds can be combined. Detailed sport–market analysis remains Pro. All four strategies get a collapsible USD/USDT calculator for unit value, VIP level, turnover, rakeback and custom weekly/monthly rates. Sportsbook defaults estimate 3.5% of a 3% house edge; 7/30-day event-start windows are samples, not actual Stake credits. Push/void and cashout bets are conservatively excluded from bonus turnover. Weekly/monthly values are user estimates, not published fixed VIP formulas or guaranteed income. No live USD/USDT conversion is provided. Tracker settlements and alert rules are unchanged.
Every Pro feature is included for 30 days from the first successful Early Access subscription payment, for monthly and annual plans, while paid access is active. The bonus is once per account; renewals and repeated confirmations do not restart it. The original Early Access plan remains afterwards, without automatic Pro upgrades or price increases. Revocation removes payment-based access. Pricing and Early Access checkout show the offer, and the account page shows its active deadline. Card payments are still being prepared. Tracker history, invitations, Pro and Lifetime plans are preserved.
Guest or inactive dashboard visits redirect to pricing. Entire locked Pro panels link to pricing for Early Access; direct FormRadar and model-statistics pages behave consistently. Pro model APIs return 403 with a pricing destination for Early Access. Pro, Lifetime, active invitations and owner access are retained. Background alerts recheck the owner membership and close unsent alerts after expiry or revocation; accepted mail, settings and tracker history are preserved. Plan-interest monthly prices use the shared USD 20/50 pricing; EUR/HUF illustrative conversions are labeled separately and card payments use USD. Strategy calculation reuses verified selections with unchanged search boundaries and results; identical eligible samples share a cache. Notification audits recognize identical markets and lines across source IDs. Dynamic profiles remain retrospective optimization.
New email digests show the other pick on the same event with exact market, selection and line, recorded odds and saved suggested stakes. Earlier notifications, queued mail and picks in the same digest are labeled separately. Complete stake records produce separate suggested-unit totals per strategy; actual bets are unknown. Browser notifications show related tracker and current-notification picks, with full details on click. Different handicap and total lines appearing only in parentheses remain distinct bet identities. The 30-minute schedule, recipient isolation, existing emails and tracker history are retained.
A uniform stake is shown once at the start of the strategy. Separate intervals sharing a market and stake appear in one row while excluded gaps remain visible; markets merge only when their complete intervals match. The duplicated staked-volume display is removed from cards. Dynamic sport/market/odds data, strategies, notification conditions and historical results are unchanged.
Live strategy sports, markets, odds intervals, stakes and EV limits replace generic sample-threshold descriptions. Sport–market–odds segments are directly visible; adjacent bands with equal stakes are merged. Tracker, Results and both Basic and Advanced automation share the same dynamic display. Unrestricted sport/market scope is labeled and an empty scope produces no notifications. Calculation, delivery filters and history are unchanged; curves remain retrospective optimization.
The fixed one-unit baseline is retained. The three stable dynamic profile IDs now represent maximum staked volume, a balance of volume/profit/consistency/drawdown, and selection of positive sport–market–odds segments. They learn from settled results; settlements, corrections and sample changes invalidate derived caches. The fourth profile has no whole-sport exclusion: requires at least 10 picks per segment, 24 overall, and positive total and latest-30% results. The first two dynamic profiles require 24 picks or 15% of the sample. Stake caps are 2/2/1 units. No qualifying positive sample means no dynamic entries. Pages refresh every minute and new notifications use the same rules. Curves show retrospective optimization, not proven future performance. Tracker outcomes, past emails and recipient strategy IDs are retained.
60 exact open records have result evidence and market-specific settlement: 22 awaiting review and 38 pending. An unfinished full-match tennis handicap is VOID/PUSH; game handicaps sum per-set games. Suspended events and events without a verified final result stay open with research notes. The owner ChatGPT tracker connection reads and saves without a separate web-login cookie, requiring exact identity, source evidence and matching current status. The connection requires installation. Closed or concurrently changed records are protected; evidence and audit history are preserved.
Result sources and market-specific arithmetic cover 40 exact historical records: 17 awaiting review and 23 pending. Tennis game handicaps use per-set game totals; U21 and senior national teams are identified separately. Two unfinished tennis bets are VOID and Córdoba −1 is a handicap PUSH, all at zero units. The repair applies only to listed records that remain open, preserves closed history and is recorded in the settlement audit.
Basic and Advanced automation support Hungarian or English emails independently of the site language. Test emails, digest subjects, guidance and warnings follow this choice. Existing recipients default to Hungarian. Event, market, selection and event links are preserved; immutable queued emails, deduplication and 30-minute timing remain unchanged.
Choose a 5–60 minute window after returning and an absolute odds-decrease tolerance of 0–0.30. Use the first qualifying recorded positive-EV price before kickoff; higher odds are accepted within the global range. Profit and stake steps use the new price. Missing observations are disclosed, never fabricated as prices or executions. Main tracker, scanning and notifications are unchanged.
Tracker filters and event rows can be collapsed independently on mobile and desktop. The header shows the filtered item count; summaries and subsequent testers remain accessible. Filters and data are unchanged.
588 verified Resend sending records and the market contents of 403 owner EV emails were recovered. The availability tester also uses provider record timestamps instead of only previously audited messages. Legacy undated messages require one unambiguous exact event/market/selection match within a 72-hour kickoff window; ambiguous matches are excluded. Sport–market analysis and availability testing are listed and gated as Pro extras. Settlement history and notification settings are unchanged.
A collapsible Pro tester supports multiple daily unavailable windows and time zones, including overnight intervals. Recorded provider timing evidence for the signed-in user’s accepted emails is the default; missing timing can optionally use explicitly labeled queue-time approximation. These timestamps do not prove inbox delivery or reading. The sport–market panel includes or excludes a market family and compares fixed-one-unit profit, ROI and chart. Notification settings, settlements and tracker history are unchanged.
Best opportunities use the OddsRadar market catalog without extra quality and matching thresholds that could empty the panel. Global odds/EV and start-time conditions remain. The Real Stake matches table is removed; underlying data needed for scanning remains.
Advanced loss limits and warnings are removed; preparation no longer stops on legacy personal-loss limits. Recipient settings add maximum EV, minimum/maximum odds and multiple sports/markets; the same check runs in the browser, background intake and before digest sending. Existing settings retain behavior without new fields. Sending is correctly labeled as a 30-minute digest. Tracker renders results before separately fetching optimized profiles so range search does not delay the table. Scanning, settlement, email links and history remain.
The turnover profile replaces flat two-unit scaling with a genuinely smaller odds/EV-filtered entry sample and three variable stake patterns capped at 2 units. The fourth profile ranks odds/EV windows with sport–market exclusions, capped at 1 unit. At least 24 settled bets, positive overall and latest-30% profit and at least half of periods profitable are required. Missing qualifying distinct profiles are explicitly marked. Volume/bonus explanation is also displayed next to the profile; bonuses are not added to profit charts. Lifetime gives full Pro access for life. Retrospective optimization; raw historical records and active fixed baseline remain.
Early-access monthly price is USD 20 and Pro USD 50; the annual 20% discount and USD 500 Lifetime remain. Early Access includes the main dashboard, market signals, tracker and alerts; Pro adds FormRadar tests and the custom staking builder. Owner and valid invitation full access remain. New Stripe checkout requires an active Price matching the displayed USD amount; payments are not activated by this change. Future list prices are not set. The FAQ explains the new profiles and related-event notification warnings without claiming a placed bet.
Highest ROI is replaced by highest staked volume: at least 24 bets or 15% of the sample, positive overall and latest-30% net profit, at least half of periods profitable, stakes capped at 2 units. No qualifying profile shows an explicitly marked fixed baseline. The fourth profile can exclude non-positive sports and sport–market combinations with at least 10 settled picks; stakes capped at 1 unit. Both are retrospective optimization, not guaranteed rising curves. Bonuses are excluded from measured profit; explanation is in the FAQ. Temporal backtest and model calibration blocks are removed from Results. Historical records, odds boundaries, fixed baseline, notifications and signal admission are preserved.
Email identity uses event, market and selection rather than changing provider signal IDs. Historical outbox picks are compared again before sending. Different signed handicaps remain eligible; another previously notified pick on the event adds an exposure reminder without claiming an actual placed bet. Tracker history and admission filters are preserved.
Cadiz Under 2.5 LOSS; Jeanjean +1.5 games WIN (10–16 games); North Carolina +19.5 and +18.5 WIN (26–37); Boston College +22 WIN (25–16); Linkoping winner LOSS (2–3). Only exact matching rows still in review may change; source links are retained. Result-provider requests share a 45-second deadline, then settlement proceeds with available evidence.
Background scans pass all collected quotes to the existing endpoint in write batches of at most 75; a single request no longer silently drops inputs beyond 75. Existing price, EV, source and tracker-admission validation remains. A monitoring type-check error is corrected.
The 16-fixture selection cap and four-adapter reference rotation are removed. All available reference adapters run within the shared deadline and their provider quota controls. Successful fixture-check times are persisted; unchecked and older fixtures advance on the next run. Three priority checks alternate with one exploration check, then remaining work continues. Priority considers all available reference markets. Four concurrent Stake requests and the safe deadline remain; fixture count is uncapped.
The half-hour odds scan selects up to 16 distinct fixtures instead of four reference pairs. All available references for a fixture share a single Stake market fetch. At most four concurrent fetches and the existing shared deadline remain; unfinished fixtures are reported as deferred. Priority and exploration allocation remain. EV and odds admission, strategies, notification timing and historical results are unchanged; additional signals are not guaranteed.
Each of the 23 exactly identified October 2–3 bets receives result evidence: 7 WIN, 15 LOSS and 1 VOID (zero units) for the Rocha–Kym walkover. Tennis handicaps use summed set games; women's football and Jong Almere reserves are identified separately. Only review rows matching ID, key, event, sport, date, market and selection are repaired. Source links remain in settlement notes and changes in the audit; later manual corrections cannot be overwritten. Strategies, notifications and all other historical bets remain unchanged.
The fixture catalogue is preloaded during the existing data stage. Odds-stage requests share a 50-second window; serial multi-page or multi-sport providers cannot restart their full timeout for each call. Expired sources do not count as fresh references; other usable data is retained. A partial Stake catalogue is shared for five minutes across cycle stages and then retried; the full catalogue uses 30 minutes and odds four minutes. Successful API-Sports status checks replace old access failures with verified daily allowance without claiming odds coverage. The external half-hour schedule, notification timing, strategies and historical records remain unchanged.
The Stake fixture catalogue is shared for at most 30 minutes; legacy two-hour cache entries also refresh at that boundary. Actual market odds retain their separate cache of at most four minutes. Monitoring records catalogue and actually fetched market timestamps separately, so old access probes cannot override fresh odds evidence. An old exhausted daily observation from the previous UTC day cannot block the new day; explicit provider resets remain respected. Unspecified HTTP 429 rate rejections use a short retry window while proven daily/billing quotas and Retry-After remain enforced. No new credentials or paid allowance are created; strategies, odds limits, Stake links and historical tracker data remain unchanged.
The test tab gains filterable sport, league, market and match breakdowns using 1-unit paper stakes, win rate, ROI, net units and pending counts. Excluded rows are omitted by default; pushes count toward staked units. Background checks pending football/NFL result sources independently of training: at most one league per cycle, six hours between checks per league. Settlement requires exact teams, date, league and market; scores and sources are retained. NHL/Liiga use existing result refreshes. Main OddsRadar and actual-bet statistics remain unchanged; these are recorded forward observations, not validated profit claims.
Watchlist and beyond-48-hour fixtures are intentional deferrals, not missing emails. Shared notifications for equivalent full-time winner/Asian −0.5 outcomes are recognized by the audit. Recovery only fills not-yet-started active picks within 48 hours and normalizes recipient addresses. The 30-minute batching, working Stake links, strategies and historical records are unchanged.
Tabur +1.5 games: 6–2, 3–6, 4–6, total games 13–14, WIN. Zagłębie–Unia 5–2 in regulation: Unia winner LOSS, Over 5.5 WIN. Wydad Temara–Union Touarga 0–1: Draw LOSS. Only the four exactly identified rows still in review can be repaired; later manual corrections and all other data are preserved. Source links are retained in settlement notes.
Dashboard and background use the same bookmaker comparison. Each cycle checks SX plus four rotating reference adapters and one quota-paced The Odds API sport; original quote timestamps determine freshness. Provider Retry-After and observed daily/billing exhaustion reduce repeated requests. Each successful cycle reconciles eligible tracker notification queues. Owner-only development audit measures chronological performance by sport, market, odds, EV, probability and reference bookmaker without changing predictions or strategies. Accepted mail is distinct from delivered mail. Historical records, odds boundaries and stakes are retained; Stripe activation still requires missing account configuration.
A recipient-level delivery lock prevents concurrent duplicate sends when the same address is configured under multiple accounts. Invalid login field types return validation errors. Invitation redemption and capacity updates are transactional. Account deletion also removes the account notification queue; disabling a recipient stops not-yet-started deliveries. TheRundown uses two rotating sport/date requests instead of twelve to keep refreshes bounded. Recipient settings always load from the current account; the owner dashboard cache is separate. Two missing English translations were added. Odds limits, strategies and historical betting records are unchanged.
The tracker preserves the exact Stake event URL and notification recovery restores it. Missing legacy links use an unambiguous match by sport, teams and start time; the sportsbook home page is no longer presented as an event link. Already sent email content cannot be changed retrospectively. Odds policies, strategies and historical results are unchanged.
Recipient-filtered market tracker picks whose events have not started no longer wait indefinitely for new odds. Emails show the original observation time and disclose when odds and EV are more than 35 minutes old. Missing notification recovery runs independently of source failures; quiet hours defer sending rather than collection. Sending clocks are isolated by account and address. Tracker intake is authoritative for notifications; rejected candidates are not separately emailed. Historical results, odds boundaries and strategies are preserved.
Eligible tracker saves also collect recipient-filtered email notifications. Digests use half-hour boundaries and a 35-minute freshness window rather than expiring before the 30-minute digest. Old signals remain awaiting new odds. Missed market picks from the previous day are recovered with their original timestamp, and matching available markets receive recheck priority. Old prices are never presented as fresh; delivered items are not repeated.
Original delivered provider messages were verified for three earlier idempotency errors. Recipient, event, selection and event link matches are recorded in the delivery audit; queue state is corrected without sending another email. Health reporting includes the latest complete radar refresh, but one successful run does not verify a recurring scheduler.
Automatic and custom stake profiles are capped at 2 units. Active odds statistics cover odds above 1.1 and up to 5; historical records are retained. The experimental FormRadar tracker follows the main strategies. Refunds and payment disputes also revoke access using the subscription reference. Retrospective optimization remains distinct from validated future performance.
Sources, provider-confirmed quotas, tracker waiting times and mail queue summaries are stored as small rows for owner monitoring. Original observation time and staleness are explicit. The snapshot uses existing evidence without additional provider calls. Missing quotas remain unknown.
Translation dictionaries are reused by language instead of rebuilt for every label. Strategy charts first render when tracker view is opened; simulator details render when expanded. Eligible samples are memoized. Recommendation rules and data are unchanged.
OddsRadar opportunities can be sorted by EV or odds in either direction, or grouped by sport. Sorting applies to the complete filtered list before pagination; recommendation and notification rules are unchanged.
Custom simulator stakes are 0.1–0.9 in 0.1 increments, then 1, 1.5 and 2 units. Unsupported loaded stakes round down for the backtest. Historical bets and automatic profiles are unchanged.
The misleading Restored badge is replaced by Backtest, with simpler odds-band and stake instructions. Missing English strategy and bankroll text and mobile control/save layouts are corrected. Strategy rules and history are unchanged.
The simulator accepts custom odds boundaries and stakes, used by both backtest and chart. Bands are increasing and contiguous with a fixed 5.00 upper bound; invalid settings hide results.
The main market recommendations and performance tracker start expanded and remain collapsible. The FormRadar experimental tracker appears below the main tracker and strategy sections, collapsed and marked experimental. Models and historical data are unchanged.
The panel, calculation and CSV export blending market and internal model signals were removed. Market EV comparisons across reference bookmakers, separate experimental models and historical tracker records remain unchanged.
Guide appears immediately after Results with a highlighted menu entry. FormRadar moves to the end with a small Test only badge. Operating and filtering rules are unchanged.
For football with matching full-time settlement, a team win and its Asian −0.5 handicap count as one recommendation; the stronger ranked version remains. Half-time, corners, cards, qualification, early-payout and overtime markets are not merged; existing actual bets remain. Dashboard view controls use separate buttons that contain their content, and FAQ bubbles are smaller and aligned inline.
Small question-mark links beside EV, ROI, fair odds, CLV, quality and market relations lead to their FAQ explanation. Automation and FAQ suggest units of 0.5–2% of betting bankroll and illustrate aggregate concurrent exposure. The 1% default, strategies, odds boundaries and historical records are unchanged; this is not a performance guarantee.
The displayed FormRadar name follows every selectable language, including compound headings and explanations. A collapsible menu replaces crowded top navigation on narrower screens; mobile menu items have separate rows. Translation changes affect display only; strategy boundaries, model IDs, routes and historical data remain unchanged.
The guide now focuses on practical steps; definitions and market relations are in FAQ, with changelog and audit on the development page. Help links target the relevant FAQ section. Automation has Basic and Advanced views sharing the same save workflow. Review queue and management diagnostics require owner access; other users do not receive review items.
The fourth profile selects on earlier settled results to balance ROI, net units, consistency and drawdown, with stakes capped at 1 unit. This is retrospective optimization, not validated future edge. The shared boundary remains 1.10 < odds ≤ 5.00; stake-band displays follow it. Historical records are preserved. FormRadar has a prominent experimental notice and methodology links use button styling.
All sports: 1.10 < odds ≤ 5.00. Picks at or below 1.10 and above 5.00 leave the active tracker, recommendations and new notifications. Exactly 5.00 remains eligible. History is retained and active statistics are recalculated. This decision uses retrospective analysis and does not promise profit.
Consensus requires the same signed handicap line. Available original reference quotes are saved with new market signals so EV can be audited later. Background email and tracker use the same adjusted probability. Missing historical quotes are not fabricated.
Strategy cards selected using historical data carry an explicit label. Latest-30% performance remains visible.
Dispatched requests, cache hits and local pacing are measured separately. Provider allowance is shown only when reported by the provider. Historically stronger sport/market groups receive priority while other markets retain capacity.
Event and market matching, repeated-bookmaker deduplication, conflicting versus related markets, per-event limits, multiple odds-based stake profiles and a fixed-one-unit control. Tennis game handicaps require detailed scores and evidence, with a review queue. This is an approximate history summary, not a complete release archive.
Multiple recipients, 30-minute email digests, durable delivery queue and duplicate-send protection. App login, access permissions, mobile improvements, translation fixes, FormRadar/OddsRadar naming and a guide.