I programmed something similar; where VAR was inversely linked to drawdown (that is, as the system drew down, it backed out of the market). Cut a long story short, you make more money by taking risk and doing the uncomfortable thing.
It's dangerous to think you can forecast Sharpe Ratios, and there isn't any statistically significant difference in performance. But based on it's diversification properties TF should probably have a higher weight than I give it; a third is probably reasonable. But I don't have a job; and my book sales plus occasional consulting gigs don't bring in enough money to pay the bills. I need a regular income and you can't trust any trading strategy to provide that, least of all a positive skew strategy like TF which only makes money every now and then (maybe something with a heavy negative skew would provide a regular 'income', until it went pop). Finally I couldn't cope with seeing a 5% daily swing in my entire net worth - the actual £££ value involved would just scare me too much. It's plausible that a bad day in the stock market would cause that kind of swing, but I don't notice it happening. I can only be so sanguine about occasional large losses in my trading strategy (see other ongoing discussion on this thread) because the amount of money is relatively small (at least relative to my entire asset base). GAT
Fair point; I've seen it mentioned in a number of other places that TF works best when blended with a standard 60/40 portfolio (not sure any of these are statistically significant, but I can believe it). It's been on my mind to write some other strategies that should be complementary (market making & relative value). Also, for anyone interested, Ernie Chan will be down the pub tomorrow. I'll be popping in. https://www.meetup.com/London-Systematic-Traders/events/87587112/
Hi GAT, What determines when you sample the market? I see in the above snapshot you've sampled the market and recorded prices at irregular frequencies. Also, is it possible to get pysystemtrade to work with intraday data as above? I assume that sampling the market intraday will lead to overtrading and potentially a lower Sharpe ratio. Is there a reason why you allow your system to run and trade intraday? Thanks, AG
@AvantGarde some of this is discussed here https://qoppac.blogspot.co.uk/2015/04/system-building-data-capture.html
In theory sampling the market intraday won't speed up your trading, that will be determined by the size of your buffer. In practice you'd probably see a slight increase in costs due to intraday autocorrelation effects. What I do isn't 'true' intraday trading since I use the last sample of the day, and then add that to my daily samples from before. pysystemtrade is not tested for intraday sampling; I plan to add something to make sure it can cope with it but I wouldn't trust it for that right now. Actually it's my plan in my next system iteration to run only daily trading for the core system; but I might add another system which does faster intra day trading. GAT
Hi GAT, The recent Eurodollar sell off has got me thinking whether you or anyone else on here have tried to successfully trade a hybrid approach: 1. long term target position determined by TF system 2. trade around long term position using discretion getting back into TF position when possible / at better levels It seems to me that Eurodollar does have a particular pattern that could be successfully short term traded. Backtest needed of course and could be talking from 20/20 hindsight. But interested to know other peoples experiences with this.