Back to Community
Will Quantopian get better short modeling?

This topic was already discussed last year when we were told that it is not a top priority.

I really can not understand how can any long-short algorithm be accepted for the fund when the quantitative inaccuracy of its short positions is going to be substantial. Just yesterday I observed one of the stocks from QTradebleStocksUS() universe - the one used for the contest - have a borrow rate of 23%. The stock in question is SRNE and 23% borrow fee is a huge headwind the algorithm will face in the real world - the other ones being short liquidity, short sale restriction and other rules. These are hard to model, so at least some way to assign 'a short sale fee' would be a good start.

2 responses


My thoughts on this is that the things you've listed might be more applicable to retail traders trading via retail brokers. I believe Q trades via a Prime Broker, which depending on the terms they've negotiated (in general, the bigger the fund and the more you trade, the better the terms), might have a lot lower (possibly negligible, net rebates) borrow costs and a lot more long inventory available, including ability to borrow 'hard to borrow' stocks inter-broker if needed (at a cost of course). The default Q slippage (and commission) model I believe is quite aggressive as well, at least for liquid stocks, which on average might even-out higher borrowing costs not accurately modeled.

Also, while this may not be a high priority for Q, one great thing about Zipline being open-source is that anyone can contribute and add a better borrow cost model if someone has it as a higher priority, e.g. for retail trading.

Well said. I guess the unknowns regarding dealing with a prime broker is really the issue here. Anyone here had any exposure to them to comment about how it changes the playing field?