Why does DD increase and returns differ for the "common years" when I reduce the backtest period?

Research
1

I have developed a ranking system and have been running simulated strategies using that system. I was testing the performance of this strategy by changing the backtest period, running MAX, 10y, 5y, etc. with the latest date as the end date. The strategy sells all positions at each rebalance using the Sell rule "Rank < 101". I'm just testing something out so have not yet defined an in-sample and out of sample timeframe.

To my surprise, the statistics for the most recent periods are different when I look back MAX vs 10 yr vs 5 yr. In particular, the DD is actually worse when I look at 10yr or 5 year vs MAX. How is this possible when I am running these time periods to "today"? What would cause this and cause the yearly returns to also differ? The only variable I am changing is the lookback period. I would have expected the stats for the years in common to be the same (e.g., 10 yr vs 5yr, I would expect the last 5 years of returns to be the same since I am selling all positions at each rebalance).

Thanks for any thoughts on this!

3

looks like if I want to have the "common years" show the same performance data, the key is to choose the start and end dates so that the rebalancing happens on the same dates.

this can be by choosing the same starting dates and varying the end dates or choosing the same end dates but then manually adjusting the start dates to make sure the rebalancing dates are the same (by looking at the transactions)

it is a good exercise in showing how sensitive the strategy is to different time periods, even if they change by just a week or so

any other thoughts welcome - thanks!

1 Like