Compare 50 AI models per game: free access and Casino Pro
AI Models Lab brings 50 selectable model variants per game into one catalogue. Compare recorded bin accuracy, exact accuracy, model family and sample counts, then use an available model in Overview. Free users can select the five lowest-ranked ready models. Casino Pro unlocks the remaining ready choices, additional challenger comparisons with their own forecasts, and private AI Strategy Lab sessions.
One catalogue, several approaches
Explore statistical references, payout-focused machine learning, neural networks and participation-aware approaches in one place. Their different history windows and inputs give you several ways to examine the selected game. Names, model families and measured results stay visible together, making comparison easier than relying on a single unexplained predictor or an impressive-looking label.
Readiness tells you whether a current forecast is available; ranking summarizes recorded performance. A high past percentage is not a substitute for enough recent information. Check the model's training count, scored sample and freshness alongside its position. When a model is waiting for usable data, you can still inspect its recorded performance and compare another ready choice without confusing an older estimate with a new one.
What the ranking measures
The primary ranking uses recorded raw bin accuracy over a shared rolling 24-hour period anchored to the last completed hour. It refreshes hourly. Exact accuracy and log loss help distinguish otherwise similar records. The displayed count is important because two models can have different numbers of eligible scored forecasts even within the same clock window.
Raw bin accuracy asks whether the category with the largest model probability matched the result. An elevated signal asks a different question about change from baseline, so its hit rate is recorded separately. Do not substitute one percentage for the other when judging how a selected model behaves in your live workspace.
Choose access that matches your workflow
Free access keeps live results and five ready models available for exploration. Other model cards show their scoring information with a Pro lock, so you can compare the recorded differences before subscribing. Active Casino Pro unlocks those additional ready forecasts and the challenger comparison. If your subscription ends, the Overview returns to an eligible free model and identifies the change beside your selection.
Casino Pro costs USD 40 for 30 days and is separate from an AviatorTrends subscription, although existing account credentials can be shared. Pro also opens private paper-strategy comparisons with virtual balances. Select a model for its recorded evidence and behavior, then evaluate its later results rather than treating an hourly ranking as a permanent guarantee.
Common questions
Does the free model selection stay the same forever?
Not necessarily. It is the five lowest-ranked models among those currently ready under the scoring and freshness rules. Rankings refresh hourly, and unavailable models do not become live choices. The interface identifies a fallback when access changes.