I should also note that this is not a case where the equity can drop inter-trade by $7k, as my stop limits are fairly narrow (like $200).
It's impossible for that kind of dip to occur on any one trade. --- In [email protected], "ozzyapeman" <zoopf...@...> wrote: > > Hoping someone can explain how Max. Sys Drawdown is calculated in the AA > Walk-Forward report. I used to think it was pretty straightforward. But > recently I've compared WF and manual backtest reports and notice some > very weird things. > > For example, if I do a Walk-Forward test on a trading system using a > 1-month In-Sample, and 1-month Out-Of-Sample period, I might see > something like the following in the report: > > Mode Begin End No. Net Profit Max. Sys Drawdown OOS > 01/01/2008 01/31/2008 1 10200 -7353 > > If I manually run a backtest on the above month (1/2008) with the same > values for all variables that were used in the WF test, I do in fact see > the correct Net Profit for that month as $10,200. However, the System > Drawdown never dips below zero. It happens to be a great (theoretical) > trading month, and starting from the first trade I am in positive > territory. The few losing trades are small, and happen well into the > month when I am already far into positive equity territory (e.g. drops > from $3k cumulative profit, down to $2.5k, then climbs again). > > Why would the Walk-Forward backtest should a Drawdown of -$7,353 when a > normal backtest with the exact same values shows only positive equity > during the entire month's worth of trades? > > I should note that most of the Walk-Forward backtests appear to match > with manual backtests. But there are alwasy 2 or 3 in a dozen that > exhibit the above type of discrepancy. Since the discrepancies are huge, > as opposed to minor, I am of course perturbed. > > Any input appreciated. >
