A strategy that only works when it gets filled has a hidden dependency on fill rate. Low fill rates on passive orders are frequently a warning sign: you are being filled selectively, which means the fills you do get are the adverse ones.
Measure it by quantity, not by order count, and measure it separately in rising and falling markets. Asymmetric fill rates are the fingerprint of adverse-selection.
Example: a mean-reversion bot rests 200 bids a day for 500 shares each, 100,000 shares offered. It fills 23,000, a 23% fill rate — and 78% of those fills occur in the 20% of intervals when price is falling. The backtest assumed 100% fills in all conditions, which is why live results are 40% worse.
Related: partial-fill, adverse-selection, queue-position, execution-quality