To me, handling a drawdown is all about knowing when something is going normally, vs when something is going wrong.
Whether you are a discretionary trader, mechanical, or algo bot, in order to know if something is out of the norm you need to have enough data to compare your results to a prior set of results.
In the very beginning of collecting this data, you only have historical backtesting data to compare your live/sim/forward test trades against.
But as you forward test in a live market (cash or sim), you can start to gain much more valuable data. The longer this period the better.
Then you can compare this new data to your old data, and look for variances. Some items to compare would be:
- MAE and MFE
- Expectancy
- Win/loss ratio
- Win/loss percentage
- Avg time in trade
- Profit factor of trades
- Number of trades per day
- Avg time between trades
And the list goes on. But by comparing this type of data, you can determine "hey, this is a perfectly normal drawdown period, all of my results are within the norm". Or you can say "holy crap, something is broken, all the latest results are way outside the norm".
What you use to determine the normal variation is up to you. I'll leave it to the math guys to answer. But you might look at simple things like standard deviation. Just keep in mind, your original strategy may no longer be profitable even with the smallest deviation from the highly curve fitted "norm" you created in your backtest.