I'll take a look at adding another strategy. By selling vol, I'm guessing you mean options (implied vol vs historical vol?)? (Sorry if that's a dumb question, no actual experience in this space, this is sort of a pathological hobby gone wild). After adding 45 futures, it doesn't look like I can diversify much more. I'm trading the majority of liquid contracts- there seem to be hundreds more, but either for similar things (London Sugar, American Sugar), or completely illiquid (e.g. Uranium). I'm going to add the cryptocurrency futures now as it's the only extra source of diversification I can find, and they offer bonkers leverage, so no counter-party risk. I've been experimenting with PyMC/Bambi (a Bayesian network framework for python); I'm going to try and use it to model my trading costs/slippage first. It seems like it could be a very useful tool both for that and for trading models too.
Hi GAT, I think you mentioned that you trade relative value. Could you explain how that works (if not proprietary)? Seems like a great addition to a trend-following system (correlation-wise), but I'm not sure about the Sharpe ratio and consistency. Thanks!
Relative value carry - simply take the carry system and deduct the average for the asset class. Relative value mean reversion - again compare the recent price trajectory to the average of the asset class, and bet on it mean reverting. I appreciate that is scant detail, and when I get time I'll do proper posts on this subject with code. GAT
Very much looking forward to that, thanks! Your system seems intentionally long-term. Do you not see value in intraday trading? Intraday trends may help, perhaps especially when volatility ticks up.
From doing a quick backtest of both of these ideas, I can see that each of the performances used to be solid and stable since the 1980s, but have essentially remained flat/down since the early 2000s. Do you see a similar deterioration in the performance of your relative value carry/mean reversion indicators?
It is almost impossible to make money from intraday trend following as the costs are horrendous. You have to cross the spread and pay up on slippage every time. Once that is factored in you need double digit Sharpe Ratios to make it pay (without giving up more than a third of your expected returns in costs). I would consider running faster systems, but they'd have to be relative value (so I can put in limit orders and get the spread). GAT
is this the definition of market making? I'm guessing you need to keep cancelling orders as the price moves?
There is a picture in this post https://www.elitetrader.com/et/threads/fully-automated-futures-trading.289589/page-6#post-4094709 I would say that I probably agree with you on the relative mean reversion strategy, but not so much on relative carry which if anything has made more money in the last 15 years than before that. But I would also say so what? these are negatively correlated strategies with trend following which means an optimiser would still give them a positive weight unless they were demonstrably awful There is now way enough statistical evidence to suggest that these things are 'broken' If I were to now remove them from my backtest I'd be guilty of implicit in sample fitting GAT