Incubation is the gap between finishing a backtest and trusting it. The strategy is frozen - no parameter changes, no rule tweaks - and tracked on data that did not exist when it was built, which is the only genuinely out-of-sample evidence available.
Set the period by trade count rather than calendar time: 50 to 100 trades is a reasonable minimum for a frequently trading system, and the comparison to make is the live r-distribution against the backtested one, not just the profit. Any change to the rules restarts the clock, which is precisely the discipline that makes incubation work.
Small live size beats paper-trading here. Paper fills are optimistic and paper losses are painless, so both the slippage-budget and the behavioural component stay untested. Trading one-tenth size costs little and produces honest data.
Related: live-vs-backtest-gap, paper-trading, backtesting, sample-size-for-edge