Total Strategy Performance

Every model behind EdgeFinder is built and tested on real historical match data spanning 10 seasons across 24 leagues worldwide - never synthetic or back-fitted data. Performance figures you see (win rate, Return on Risk, drawdown) come directly from that same historical record, computed the same way whether the period is 2016 or last week.

Models are validated using walk-forward testing - evaluated only on data they had never seen during development. This is the standard defense against a model that looks great on the data it was built with but fails on new data. Live performance is tracked continuously against this same standard.

Every recommendation is sized as a fraction of your bankroll and automatically scaled down when confidence drops or live results drift from historical expectations. Risk management is applied identically whether a result is about to be a win or a loss - it manages exposure, not outcomes. See How Risk Works for the full breakdown.

Every publicly shown signal is timestamped at the moment it's published - always before its match starts - and locked from that point on. Nothing is ever added, edited or removed after the fact. See Methodology for how that proof works.

Every result - win or loss - is published and stays visible permanently at edgefinder.tech/results. No losing signal is ever hidden or deleted. What we don't publish is the selection logic itself - the specific rules, filters and signals stay proprietary, so the numbers you see can never be gamed by publishing the "answer key" alongside them.