Product updates

Changelog

What changed, when it shipped, and what it means for your next visit.

Looking for a correction to a specific pick or article? Read the Corrections log.

  1. Latest update

    A clearer look inside the current prediction engine

    Site explanations now distinguish the two-advocate debate engine from historical seven-lens Kimi reports.

    • Prediction methodology

      Two opposing cases, one final adjudication

      New official forecasts use one frozen, market-blind evidence packet, two Luna advocates and one Sol adjudicator. The site now explains the bilateral winner discriminator, premise and phase checks, always-forecast v6 rule, conditional method and timing estimates, and disclosed confidence caps. Historical seven-lens reports retain their original labels.

      See how the engine works

    • Homepage

      Every fight has two sides. We make both prove it.

      The homepage now leads with the engine’s opposing-case architecture, explains the adjudicator’s role in plain language, and connects free official picks to the complete fight blueprint available with Sharp.

      See the new homepage
    Link to this update

  2. More reliable automatic upkeep

    Improved background maintenance and independent monitoring of the site and prediction pipeline.

    • Reliability

      Catch interruptions and keep maintenance focused

      Background record checks focus on upcoming active bookings, preserve progress when interrupted, and keep unresolved identities visible. Pipeline monitoring now accepts its own recovery links, and an independent check watches for missed scheduling and service outages.

      Read the methodology
    Link to this update

  3. More reliable saves and a leaner application

    Saved fights follow your current account, and interrupted status checks can recover.

    • My Blueprint

      Keep saves in sync with your account

      Save controls wait for the current saved state, refresh when you switch accounts, and offer a retry when a status check stalls. Delayed responses from a previous fight or account cannot change the current control.

      Open My Blueprint

    • Reliability

      Simpler upkeep, with existing records preserved

      Removed unused demo-authoring tools and retired pipeline branches. Article recovery reports unavailable data explicitly, and the citation pack now describes the current OpenAI engine. Existing videos and prediction history remain available.

      Read the methodology
    Link to this update

  4. Current quotes when you return to the Desk

    Visible Best Bets refresh as quotes expire, and odds collection prioritizes the nearest event.

    • Whale Betting Desk

      Keep quote comparisons current

      The open Desk checks stored prices when a quote expires or you return to the tab. Hidden tabs stop refreshing, overlapping requests are combined, and expired prices remain clearly labeled until a fresh quote is available. Scheduled odds collection checks the nearest card first, including events with special names such as Noche UFC.

      Open the Betting Desk
    Link to this update

  5. Clearer lineup refreshes and reliable page styling

    Preparation waits for a current verified lineup, and returning visits receive updated layouts.

    • Evidence preparation

      Wait for fresh roster verification

      When an accepted lineup needs fresh verification, preparation waits without repeatedly consuming retry attempts. A new verified observation can resume the queued refresh even when the matchups remain the same. Existing forecasts and source-disagreement safeguards stay intact.

      View upcoming events

    • Design

      Keep returning visits up to date

      Updated page styles receive a new cache identity, so returning visitors can see layout improvements without manually clearing their browser cache.

      Open the Betting Desk
    Link to this update

  6. A calmer Betting Desk and clearer page layouts

    Simpler navigation, more readable matchup summaries and consistent spacing across the site.

    • Design

      Find the useful information sooner

      Page headers use quieter typography and consistent margins. The Betting Desk brings its five tools together in a compact toolbar, keeps each Best Bet beside its matchup, and improves light-theme contrast and keyboard focus. Account, pricing, methodology and editorial pages share the refined presentation.

      Explore the Betting Desk

    • Best Bets History

      Keep long-term records responsive

      History loads successive pages without repeatedly sorting the full revision archive. The first priced selection still determines each matchup’s scorecard result, and every later revision remains recorded.

      Review Best Bets History
    Link to this update

  7. A Best Bet for every fight, with a record to follow

    Clearer matchup reads, estimated-value rankings and an enduring Best Bets scorecard.

    • Whale Betting Desk

      Compare value at the actual price

      Each published fight has a Best Bet selection. Supported current quotes are ranked by model-estimated return, with the book, quote time and price needed for positive expected value. An unfavorable or unavailable price stays clearly labeled; sportsbook prices never alter the official AI forecast.

      Open the Betting Desk

    • Best Bets History

      Track results without counting a fight twice

      Selections and quote revisions remain recorded. The performance scorecard uses one first priced selection per matchup, showing win rate, flat-stake units and ROI from verified results. Unpriced, pending and void selections are reported separately, and later refreshes cannot multiply a result.

      Review Best Bets History

    • Matchup analysis

      Read the tactical explanation first

      Weekly matchup summaries remove technical fighter identifiers and introductory identity labels before shortening the original explanation. The saved prediction and its probabilities remain unchanged.

      View the weekly picks
    Link to this update

  8. Safer card updates and more reliable preparation

    Stronger checks for changing matchups, interrupted collection and published forecast history.

    • Official predictions

      Resolve source disagreements before changing a card

      Updates to an accepted lineup now pass roster checks before they can change an established matchup. Conflicting opponent or weight details remain visible and hold new prediction work until resolved.

      View upcoming events

    • Evidence

      Keep interrupted updates consistent

      Source collection and worker cleanup have bounded deadlines. A fighter-history repair saves its related changes together or leaves the existing record intact. Missing statistics remain unknown, and documented submission finishes retain their method classification.

      Read the methodology

    • Publication history

      Recover a verified forecast without changing the pick

      A verified publication repair can reuse the original saved forecast while preserving the invalidated publication and its article. Recovery does not generate another prediction or alter its probabilities.

      View the prediction record
    Link to this update

  9. More dependable updates and clearer privacy controls

    Stronger source collection, automatic discovery of new picks and easier cookie choices.

    • Official predictions

      Keep evidence fresh and results precise

      Source collection recovers more reliably from interruptions and reuses unchanged evidence while recording each new retrieval. Event completion follows verified results, and monitoring distinguishes current failures from work already replaced or recovered.

      View the prediction record

    • Discovery

      New picks reach the sitemap automatically

      Published predictions enter the live sitemap without waiting for a site rebuild. Search notifications retry independently if a service is unavailable.

      Browse official analysis

    • Privacy

      Revisit your cookie preferences

      Cookie settings are available in the footer. Withdrawing consent stops analytics collection, and private account pages share stricter analytics exclusions.

      Open Cookie settings
    Link to this update

  10. Sharper evidence rules for both prediction engines

    Clearer opposing cases, stricter comparisons and stronger probability safeguards.

    • Prediction engines

      Ground each case in the complete matchup

      Each advocate combines the relevant physical, statistical, stylistic and historical evidence for its fighter. The adjudicator checks both cases against the same evidence, including how positions are reached, maintained and denied. Relative comparisons require verified values for both fighters; unknown control stays distinct from an even matchup.

      Read the methodology

    • Probabilities

      Keep documented winning methods possible

      Unclear tactical access can reduce a method’s estimated likelihood, but cannot erase verified historical winning methods. New forecasts also reject absolute win certainty. Published predictions retain their original prompts and results; these safeguards do not establish a measured accuracy gain.

      View the prediction record
    Link to this update

  11. A clearer, more consistent Blueprint experience

    Refined reading, navigation, and account flows across desktop and mobile.

    • Design

      Readable reports and consistent page layouts

      Official and Custom reports now give each opposing case and tactical phase a clearer place. Editorial pages, event cards, account forms and support states share more readable typography, spacing and controls, with improved light-theme contrast and keyboard focus.

      Explore the methodology

    • Navigation

      Clearer ways to continue

      The menu, footer, empty states and recovery pages make next steps easier to find. Keeping a membership during cancellation now returns to Settings. Prediction records, model outputs and paid access rules are unchanged.

      Open Settings
    Link to this update

  12. Published blueprints recover safely from temporary read failures

    A temporary loading problem no longer makes a published blueprint appear missing.

    • Published analysis

      Keep valid public pages and retry safely

      When a read temporarily fails, a valid saved public page remains available. If no verified public page is available, visitors see a retry message. Missing or superseded articles still close, and full analysis continues to require verified paid access.

      View published picks
    Link to this update

  13. More reliable prediction recovery and event progression

    Interrupted reports can resume from saved work, and card monitoring follows the current OpenAI publications.

    • Engine explanation

      How opposing cases become a forecast

      The methodology and Custom Report guide now explain the two parallel OpenAI advocates, the adjudicator’s treatment of conflicting evidence, raw win probabilities and conditional method estimates. They also explain how verified career history can support a cautious forecast when tactical evidence is sparse, without claiming a measured accuracy improvement.

      Read the methodology

    • Official predictions

      Less repeated preparation and clearer holds

      After a fight publishes, the next eligible matchup can begin immediately. A changed review stops unnecessary preparation. Unstarted work with an outdated review can advance through a verified revision, while published picks and paid attempts retain their history. Card monitoring counts current predictions and their linked articles together.

      View upcoming events

    • Custom reports

      Recover saved work and request fresh reruns

      Eligible interrupted reports automatically resume from saved results within their original spending limits. A requested fresh rerun creates a new report instead of returning a cached one. Uncertain charges and explicit holds still require review.

      Create a Custom Report
    Link to this update

  14. OpenAI across both engines. A lasting record of Best Bets.

    Custom reports join the OpenAI prediction workflow, and the Whale Betting Desk gains a history of its recommendations.

    • Prediction engines

      The same debate workflow for official and custom predictions

      Two OpenAI advocates examine the opposing fighters before an adjudicator weighs their arguments. New custom reports use this workflow, while previously saved reports keep their original model and results. Missing tactical evidence stays unknown; qualifying regional history can support a prediction.

      How the engines work

    • Whale Betting Desk

      Best Bets history

      Newly recorded recommendations retain their original selection, quoted price, timestamp and source prediction. Later results remain connected to that record, including losses and voids. The history begins with recorded recommendations; past prices and bets are never recreated after the fact.

      Open the Betting Desk

    • Access

      Full blueprints require verified access

      Free visitors retain the official winner, win probability and publication receipt. Full tactical analysis requires an active membership or matching Event Pass. Cached pages no longer retain private analysis or another viewer’s access, and signing out closes an open blueprint.

      View access options

    • Site

      A public changelog

      Product changes now have dated, shareable release notes. Find this page in the site menu or footer. Corrections to specific predictions and articles continue to appear in the Corrections log.

      View corrections
    Link to this update

  15. Official picks move to OpenAI

    The Silva–Delgado card has 13 OpenAI predictions, each linked to its published analysis.

    • Official predictions

      Opposing arguments before the final prediction

      Two lightweight advocates and a reasoning adjudicator produce winner probabilities, method probabilities and round forecasts. Historical winning methods retain a small nonzero possibility. The probabilities are raw model estimates; improved accuracy has not yet been established by prospective results.

      View the card

    • Reliability

      Reviewed results and safer event progression

      The preceding Paris card is fully settled. Upcoming predictions check the accepted lineup, evidence, review, budget and lock time before running. Updated official result receipts can refresh the review while preserving earlier forecasts and paid attempts.

      View the Track Record

    • Corrections

      King–Rosas reach comparison annotated

      The evidence contains no verified King reach measurement, so it cannot support the claim that Rosas has the longer reach. A visible correction preserves the original prediction and explains that the effect of the unsupported comparison on its probability is unknown.

      Read the correction
    Link to this update