Hi, the returns revealed after we ran port.backtest() for 2007-2016, and what is shown on the public leaderboard differs a lot. My impression is that the public leaderboard results should match the given backtest , and only the private leaderboard results should differ, is this understanding correct?
The reason is that the backtest does not account for the transaction costs. Each time you trade / rebalance there is a transaction cost associated with it.
You can find more details on this forum post.
I see, thanks for the clarification!
But please do note, if you wish to change your rebalancing frequency, you may submit another revision before Monday and we will accept it.
Ah that’s good to know, thank you!
The algorithms are not evaluated on out-of-sample data?
Yes they are. That’s what the private leaderboard shows