-Fama French factors like small vs large cap stocks, value stocks, momentum stocks, etc
-Short vol (it does work in the long run despite what happened with the vix blowup last week)
-Long credit premium (High yield vs investment grade etfs)
Fully automated futures trading
- Thread starter globalarbtrader
- Start date
-Short vol (it does work in the long run despite what happened with the vix blowup last week)
-Long credit premium (High yield vs investment grade etfs)
Sir,
I find it difficult that avg holding periods of 2 weeks has a sharpe ratio of 3+. This is because so many things can happen in 2 weeks to affect the pnl variance of a trade. Also 2 weeks holding period mean that the number of trade samples you have can easily lead to overfitting.
I have no issue doing this kind of work; I really enjoy it. The issue is, building the code around it all is a lot of work, and 95% of it is building 'stuff' and not actual strategy. Guess I should just get on with it and build it!
You can use a vectorised approach to build strategies which is quite fast to build. I've seen some python vectorised stuff on the web, can't recall right now. I don't recommend building so much infrastructure if you have no idea of what type of strategy is gg to pay off. GAT knew exactly what he had to build so he cut the event driven stuff and also simplified the execution portion which made his code really sweet and nice.
I did that too; I built it all vectorised. I'm in the process of tidying it up to open source. When I think about how much work it would be to move to an intraday strategy, I start to get the shakes. It's such a pain integrating data from different sources/timezones etc. I'm also not sure what kind of strategy makes sense on an intraday level.
That's true if you are only trading one instrument. But of you are trading equities like this case, with a universe like the Russell 1000, even a biweekly turnover with a long short portfolio would be 2000 bets a month.
I guess that such a trading strategy would cost quite some money in terms of trading costs: commissions and slippage.
This guy claims that his sharpe is around 3.2, well it all goes back to the question "what kinds of returns are actually possible" in systematic (or any kind?) of trading, which was one of the first questions GAT discussed in his books\blog\forum.. I personally have very little experience with it as I only do this as a hobby., But very simply, if I look at the Leaderboard on fundseeder in the "Systematic" section, and order by sharpe desc (https://fundseeder.com/leaderboard/tm_systematic/sharpe_ver/1), the top-best performer at the moment has sharpe 2.59 and his track-record only goes from April 2017. If I search for the best systematic sharpe with at least 3 years of history(which still isn't that much I think) it's only 1.62 (this guy at the moment https://fundseeder.com/trading_account/517).. This kind of tells you something I think about expectations.. Browsing around some more on fundseeder, it seems that Discretionary guys are doing slightly better - there are at least 2 sharpe ratios above 3 with 3+ years of history..