Latest Model Results
Probabilist trains and validates a set of models — covering calls, puts, and long/short strangles across a broad universe of tickers — to estimate the probability that a given option will be profitable if held to expiry. This page describes how those models are evaluated; the full methodology and detailed results are available in the whitepaper below.
How the model is evaluated
Each model is validated using walk-forward testing: the model is repeatedly retrained on an expanding window of historical data and evaluated on a subsequent, entirely unseen out-of-sample period, so that no test period ever overlaps with the data used to train it. This is repeated across many sequential windows spanning multiple years and market regimes, rather than relying on a single train/test split.
Beyond raw classification accuracy, we also evaluate how well the model ranks opportunities relative to one another — whether the option candidates it scores highest actually outperform the ones it scores lowest — since ranking quality is what determines whether the model's output is useful for selecting trades in practice.
Read the full whitepaper
Methodology, validation results, and known limitations, in detail.
Important caveats
- All results are based on historical data and do not account for slippage or transaction costs
- Options trading involves significant risk, including the potential loss of the entire premium
- Past performance does not guarantee future results