Straight answers about how the predictions work and what they can and can't tell you.
Every match is simulated thousands of times, start to finish. Each simulation plays out the map veto (bans and picks), then plays every map round by round, first to 13 with overtime at 12-12. Round chances come from team ranking, recent form, player ratings, head-to-head record and each team's map history. The final probability is simply the share of simulations each team wins.
No model knows a match's exact probability, so we don't pretend to. Simulations run across many plausible "strength scenarios" for both teams, and the range shows the middle 90% of outcomes. A tight range means the data is solid; a wide range means the data is thin (a new roster, few recent maps) and you should treat the headline number loosely.
Here's the whole idea in one example. Suppose our simulations say a team wins 67 times out of 100, but a bookmaker's price only pays as if they win 61 in 100. If our number is closer to the truth, betting that price makes money over many bets, because you're being paid for a 61% chance while holding a 67% one. That's what the Genie's verdict looks for, and it recommends the bookmaker paying the most for it. It's a disagreement, not a promise: sometimes the market knows something the model doesn't, like a last-minute stand-in.
Decimal odds implied directly by our probability, with no bookmaker margin baked in. If our model says 50%, fair odds are 2.00. Bookmakers quote worse than fair odds on both sides; the difference is their margin.
We measure instead of claiming. Every prediction is logged before the match starts and graded against the final result using the Brier score, a standard accuracy measure where lower is better and 0.25 equals coin flipping. Once enough matches settle, the live track record appears on the homepage, and the model's weights are periodically refit against those real results.
The slate re-evaluates itself automatically every few minutes, and inputs move: form updates, new matches appear, market odds shift. A prediction is a snapshot of the best available information at that moment.
Yes. The What-if lab on each match lets you drag a team's form or player ratings up or down and watch the whole simulation re-run live. It's marked "hypothetical" so you always know when you're looking at your scenario instead of real data.
Treat everything here as analysis, not advice. Predictions are estimates with stated uncertainty; upsets are normal and expected. Never bet money you can't afford to lose, know the rules where you live, and if gambling stops being fun, seek help. In many countries you must be 18 or older to bet.
Match schedules, teams and live bookmaker odds come from licensed sports-data APIs. Some stats aren't available for every team or event; when data is thin the app widens its confidence range rather than guessing confidently.