Do your momentum rules on the equity/bond long only holdings include a trend filter? (Eg if MA50 < MA200, go to cash) Or is there a reason you’ve decided against it?
Fully automated futures trading
- Thread starter globalarbtrader
- Start date
Yes, as discussed in smart portfolios
Yes because there is overwhelming evidence that TF performance is mean reverting
Rob
(That was sarcasm by the way)
@wopr I just saw your post from a while back relating to “Strategy twenty-seven: Safer fast mean reversion” from the AFTS book
I agree, there is most definitely a lookahead error with the equilibrium calculation whereby we are using today’s close in order to calculate order levels for earlier that same day.
Quite extraordinary the difference between using today’s and yesterday’s close makes, where we go from a strategy with a Sharpe of 2.1 to a strategy where every single market consistently loses money over all time frames.
So no I don’t think you’re missing anything as my results also produced one of the most consistently unprofitable backtests I’ve ever seen!
It’s a pity because the underlying strategy logic seems very sound and I’m sure there was definitely no intention by Mr Carver to mislead. In fact he modestly makes several disclaimers …
“ I do not trust a backtested Sharpe ratio over two. There are certainly plenty of reasons not to trust this one.”
“There may well be assumptions or errors in my code that make the results look better than they really would have been.”
“I would advise very careful testing”
Which I think all just goes to reinforce the fact that regardless of an author’s pedigree, one should always do one’s own testing.
I've come to the same conclusion myself and at some point was going publish a mea culpa blog post, but it's hard to find time to do something which won't add value.
My own tentative fiddling about suggests that a version where you wait until the MR divergence is higher is profitable (discovered by accident when noticing a big difference between rounded and unrounded results for small capital), which is the more standard way of running such a strategy.
Apologies
Rob
Thanks for the update, Rob. So, just to be clear, you no longer have faith in the wonderful sharpe ratio of your chapter 27 Safe Mean Reversion system, as published in AFTS?
Let's assume risk tolerant person would like to allocate 75% to futures trading and the rest to systematic ETF investments. Let's also assume small capital. Obviously allocating 25% from futures into diversified investments translates to lower risk and thus lower expected return. However, opportunity cost here is also the fact, based on my understanding from AFTS, that DO will be less efficient with smaller capital, and therefore the lost expected return is even higher.
Based on AFTS, I would assume the point where risk tolerant would start diversifying to investing is somewhere between 100k (enough to start) and 500k (where DO is practically as good as Jumbo portfolio). So my question is, what would be a reasonable amount of capital, with which DO will work well enough and the relative extra penalty of allocating capital elsewhere at that point is not that significant?
Personally I think the absolute minimum account size to trade futures is probably around the 50k mark; I can't remember if I use that or 100k in AFTS. Irrespective of performance expectations.
Rob
So you would diversify into investing right after minimun capital is met, whatever that is? I was wondering how much capital DO would need to function well enough.
I'm debugging something with my DO implementation (gist of the issue is, in my backtesting, performance of the DO drops significanly with lowering the capital, eg. from a 1.14 SR for a Jumbo portfolio to 0.91 SR for DO with 800K, for the same forecasts), I think I narrowed it down to correlation and covariance calculation.
Rob, I wanted to ask, what's the difference between `instrument_correlation_estimate` key in the config (https://github.com/robcarver17/pysy...d514549e2cf/sysdata/config/defaults.yaml#L226) and `instrument_returns_correlation` 50 lines below, besides obviously the parameters? I see that the latter one is used in reports, so the correlations on the risk report are not the same as the ones used in the DO run (besides obviously not being shrunk).
If I read the code correctly, I think the former is used in the DO run, with `floor_at_zero` set to false. Also, since the lookbacks are so short (25/75 weeks), the roll years parameter is irrelevant, is that right?
Thank you in advance!
Rob, I wanted to ask, what's the difference between `instrument_correlation_estimate` key in the config (https://github.com/robcarver17/pysy...d514549e2cf/sysdata/config/defaults.yaml#L226) and `instrument_returns_correlation` 50 lines below, besides obviously the parameters? I see that the latter one is used in reports, so the correlations on the risk report are not the same as the ones used in the DO run (besides obviously not being shrunk).
If I read the code correctly, I think the former is used in the DO run, with `floor_at_zero` set to false. Also, since the lookbacks are so short (25/75 weeks), the roll years parameter is irrelevant, is that right?
Thank you in advance!