zest.win 1.6.0: CS2 Leaves Beta, Clearer Track Records
CS2 predictions leave beta in zest.win 1.6.0, alongside team comparisons, clearer forecast evaluation, and faster browsing.
zest.win 1.6.0 is live. This release moves Counter-Strike 2 predictions out of beta, adds more context to CS2 and Dota 2 matchups, and makes our prediction record easier to understand. It also brings faster browsing and a cleaner homepage.
Here’s what has changed since 1.5.0.
CS2 predictions leave beta
Counter-Strike 2 now joins League of Legends without an experimental label. We’ve removed the Beta badges and Calibrating model message for CS2 across the platform, including the homepage, game menus, match cards, recent results, and blog navigation. The methodology and evaluation labels reflect the same change.
CS2 coverage includes series and map win probabilities, pistol-round predictions, and round-total forecasts where available. Match pages bring those forecasts together with bookmaker odds and model edge, subject to your subscription access.
Leaving beta is a change in product status. Predictions remain estimates, and their results should be judged by game, target, time period, and sample size. Dota 2 continues to carry its experimental label.
More context before a match
CS2 and Dota 2 match pages now include team comparison statistics, giving you more information alongside a forecast.
For CS2, the comparison includes map and round win rates, CT and T performance, pistol-round win rates, and average kills. For Dota 2, it includes win rates, Radiant and Dire performance, towers, barracks, and average kills. Season and recent-form comparisons help you see how the teams’ broader records compare with their latest results, where data is available.
These statistics provide context for a matchup. They don’t make an outcome certain, and a limited history still deserves a closer look.
A clearer prediction record
This release makes the distinction between historical coverage and verified recording more explicit.
Historical coverage describes stored predictions and their graded outcomes. A historical accuracy figure can include forecasts recorded after play began, so it should not be read as verified pre-match performance. Retrospective backfills are excluded from that sample.
Verified performance uses the last eligible, immutable forecast revision recorded before the source-reported actual start, counted once per event and prediction target across model versions. Both generation and archive timestamps must precede that start. This verifies when a forecast was recorded; it does not establish when it first became publicly visible.
Sample sizes matter in both cases. Where there are no eligible settled forecasts, there is no accuracy percentage to report. We’ve also refreshed the recent-results feed to use settled recorded predictions and updated evaluation requests to avoid showing cached results from an earlier policy.
The methodology page now has clearer navigation, probability examples, settlement explanations, and guidance for reading the track record. Its historical evaluation snapshot is labeled as belonging to a previous policy.
Faster browsing and a cleaner homepage
We’ve improved homepage and match loading, simplified the homepage prediction cards, and removed the Startupbar promotion widget. The result is a more focused view of the matches, model statistics, and recent results.
Game prediction hubs are easier to reach from the navigation. League of Legends and CS2 team profiles also now deliver their initial profile content in the page HTML, helping people and search engines access the available information before the app finishes loading.
Available now
Version 1.6.0 is deployed at zest.win, and the site footer displays the new version. Your existing account and subscription continue to work.
Open a CS2 matchup to explore the team comparisons, or visit the methodology and track-record pages to see how forecast results are reported. If you spot incorrect match data or a settlement issue, send the match URL and supporting details to hello@zest.win.
Predictions and statistics are for analytical and informational purposes only. Outcomes are uncertain, and past results do not guarantee future performance.
Related Analysis & Predictions
View all posts →
Procyon vs BORRACHEIROS: model prices a 64% series edge
Our model makes Procyon 64% for this ESL Challenger League BO3, above the bookmaker line, with BORRACHEIROS at 36% in a first meeting.
Read →
Galorys vs MEIA NOITE: model favorite meets better form
Our model makes Galorys a 70% series favorite over MEIA NOITE in ESL Challenger League South America, though recent results favor the underdog.
Read →
MOUZ vs Spirit: model edges MOUZ 52% in ESL Pro League BO3
MOUZ vs Spirit at ESL Pro League Season 24: our model makes MOUZ 52% in the BO3, against a 47% bookmaker implied probability.
Read →