I'm writing to tell you about a bug in our software that affected the results shown by our backtester. I apologize on behalf of the team for any difficulties this might cause.
Impact: For any backtests that you ran between late 10/31 and earlier today, 11/15, the benchmark returns were offset by several months. Specifically, when you looked at the the graph of your results, both the red line indicating the benchmark performance was incorrect, and the top-line, "Results Overview" risk metrics were incorrect. If you shared your backtest result, the incorrect benchmark data is visible in that share.
Most of the backtest results are unaffected. Your algorithm's returns were calculated correctly. If you drilled into the detailed risk metrics, where the results are displayed in tables, those were calculated correctly.
How we're fixing it: This afternoon we corrected the problem in the backtester. If you go back and re-run your algorithm, the new backtest will be correct. Shortly, we will put a message on the display of all affected backtests.
How we'll prevent this in the future: We're increasing our automated testing against known-outcome algorithms to verify future code changes.
We know that this kind of bug is one of the worst. You need to trust our backtester, and we need you to trust us. When we do make mistakes, we will be as transparent as possible, providing timely and accurate information about the error. We're very sorry this happened.
Please tell us any questions or concerns you might have. You can ask us below, or, if you prefer, email us privately at [email protected].