i was reading Advanced Futures Trading Strategies. There in footnote 69 there's a formula for the standard deviation of sharpe ratio estimate. (1+0.5*SR*SR)/N So if the strategy have sharpe ratio of 0.5 and we have 250 days of data. This will be equal to (1+0.5*0.5*0.5)/250 = 0.0045. However the standard deviation bands in figure 20 are much bigger compared to 0.0045. Perhaps I am making mistake in my interpretation of this formula. Anybody has ever looked into this?
First, that's the variance of the estimate not the std. deviation. You need to square root it. The footnote says that. Second, N and SR need to be in same time units. You are mixing daily and annual units up. SR of 0.5 annualised, equals 0.03125 per day (divide by sqrt[256] which is approximate trading days per year). Std dev will be sqrt[1+.5*0.03125^2)/250] = 0.063. That's daily. To annualised back up the std dev, multiply by 16. That gives a std dev of 1.01. So for one year of data, your 95% confidence interval would be roughly 4 SR units. The biggest in the graph is around 3 SR units, because we usually have a bit more than one year. Rob
I would recommend holding off on adding ATW. I did a test single contract trade on it yesterday using IBs adaptive algo and got a terrible fill. I am also thinking of creating a seperate list of 'decepticon' contracts like this where volume appears high but is poor in reality. For now I am thinking of using a combination of Robs liquidity and cost report to create some sort of top filter.
Also I hope you guys had a better Feb than me - I ended up with a nasty -9%. Lots of reversals - oj was particularly savage. By the end of the month, the system ended up closing both oj and cocoa longs and reducing risk from ~23% to ~17%. I am curious as to what levels of turnover on average do you guys run? I am around 4 times a year. My formula (using daily wts) is: 0.5 * days per yrs * avg absolute trade/avg gross leverage
February was tough for me too. Down 9.68%, and now down 6.51% on the year (before today). It's not just us. TransTrend's main fund lost 9.94% in February, is down 9.12% YTD, and it's now in its largest ever drawdown.
I use IB's volume rather than CSI's to estimate my trading costs (a market impact model). It's not hard to download IB's volume and compare it to CSI.
Yep that's what I was thinking. I now modified the liquidity check so that some markets have a higher minimum volume requirement. IBAlgo - I also noticed terrible fills previously - even on Patient it just crossed the bid/ask. It really looks like IBAlgo just flat out fails to work on some markets and just reverts to being a pure market order, but there's no documentation I've found that says what these are. Possibly IBAlgo can't work on IPE contracts properly? For markets like Malaysian Palm oil @Bursamy where the Algo is not supported it's very clear from the TWS interface that it will be pure market and I think there's an API warning too. Another flag I've setup is to use Limit orders for these type of markets, but it's been trial and error to find out what they are. Was down -8% MTD on Friday, but it's up to -5% now (does include some stock strategies though). Today had good moves in NG, Coffee, Gold, Euro Indices & Short GOIL/COIL, RTY, German bonds to outperform some other shitty positions. YTD looking better than MTD so I'd rather watch that, since monthly moves are more likely to revert (seems like it's starting to). I don't have IBs volume in my data for input into the model. My execution engine only uses API for tracking positions, entries, and exits. So I just have to make an assumption on a multiplier that I need. However I'm still concerned about the IBAlgo not being usable on contracts like these either. Possibly a limit order and a notification that I have to manually check on it, otherwise I might just kill the contract from the universe.
At least we are not alone... At a factor level, both carry and trend sucked so no place for me to hide. I went live 1 October 2024 and this move has taken me back to almost zero gains, but I guess its important to see the big picture that moves like this are actually quite frequent in the simulations - so things are within expected specification.
Looks like it's all a bit random with the IB Algo orders. I noticed Malay Palm Oil because it actually warns you. I dont have any data subscriptions on IB, so not sure if I can download historical prices and volumes. It doesn't even show me the price for ATW on a watch list but it will show me charts on TWS because I now have a position on.
I'm about 4.5% down YTD, also haven't seen a new HWM since Sept 28 2022, although, was close on Jul 11 2024..