In the contest, bigger your draw down, the higher you are ranked.
In the contest, bigger your draw down, the higher you are ranked.
One of my algos has a max drawdown of -9.117% for its backtest, and 2.263% for paper trading. See https://www.quantopian.com/leaderboard/22/5782ed7239bd8b7892001179 . Seems that drawdown should be reported as either negative or positive but not both (it should go from 0 to negative infinity or 0 to positive infinity, but not both). Otherwise, we are measuring both drawdown and drawup.
And at the top of the leaderboard for the current contest, a drawdown of 0.01126% is ranked at 406. See https://www.quantopian.com/leaderboard/22/57c5c8b9ae7033ce8500014b .
On leader board Contest 21:
Backtest Max Drawdown are from 0% to -99% ranked correctly.
rankMaxDD_bt maxDD_bt
14.5 0
504 -0.990415633
Paper Trading Max Drawdown are from 0% to 60% ranked not correctly in opposite order.
rankMaxDD_ pt maxDD_pt
1 0.605514656
466 0
The same on other leader boards.
Q are you going to fix the bug?
This time I won't even bother asking if the rules have changed.
What I'm wondering is if these kinds of extremely silly mistakes are made in such a simple calculations then what does this mean for more complicated matters like fund risk calculations etc. This may be a bit harsh thing to say but quality control doesn't seem to be Q's strongest points..