"NY Harbor ULSD" is probably the least helpful description of an instrument I've ever seen... apparently it's heating oil. I wish people would just give the name of the instrument rather than these cryptic codes. My codes for these markets are HEATOIL and GASOILINE. Anyway yes they both make money Rob
Dear Rob Thank you so much for your reply as well as your time to generate those result. I'm glad to see that the strategy no.9 works very nicely with heating oil as well as gasoline future. Now I have to re-write all the code. Below is a Euro FX future from CME. I tried to show the N of strategy no.9 with $1,000,000 with 20% risk target. I see contract size goes above 20 easily even 40. Do you think I'm on the right track...?
Again would be very helpful if your plot had an x-axis on it... https://photos.app.goo.gl/jZz39sqeNXJTA64X9
My bad Here is a N that I've got for Euro FX and Gold future I excluded ewmac2 and ewmac4 which I thought would beneficial for me(for now). Maybe I should write code containing both, but still can't figure out why dynamic approach makes worse profit than static one. Must've written code wrongly.
Based on your reaction, it seems like what I've created isn't terribly strange or incorrect. That's enough for me. I will keep fixing my code.
After finishing the latest book, I still can't get the Fast mean reversion (or the safer 'buy the dip') out of my head (thoughts of enormous profits keep me up at night) So how about this idea: 1. we pick ~20+ "smallest" (by margin\notional\instrument risk) instruments out of our large pool of futures and start trading the fast MR strategy on them (limit orders and all). We want the cheapest instruments to be able to trade as many of them as we can to get the max diversification in MR with a small capital. 2. but we don't fully exclude them from the main "slow" system: we allow it to "see" but not change the positions of these instruments, and put them as constants (current positions) into the DO of the “slow” system. I.e. the MR strategy will be allowed to trade the ~20 instruments, and the slow system will include their positions as they are currently set by the MR into all the test portfolios of the DO’s greedy algo without trying different variations for them (but DO will still be allowed to try different weights for all the other instruments). It seems it can work..? We will be able to sort of share the same instruments across both strategies to some degree, in general the “buy the dip” MR strategy should put positions only in the same direction as the slow system (because it doesn’t trade counter-trend)., the slow system will probably try to compensate ‘excessive’ positions with other instruments, but maybe it’s not so bad, because it will be trying to “preserve its general view\desired overall exposure”.. The questions will be how to allocate capital\risk to these 2 sub-systems., and the number of trades will probably increase in the slow system somewhat, but I think it can be addressed… Another question is will we be able to find enough small instruments for the limited capital for the MR strategy to still be profitable..
Hi this is quite a nice idea and along the lines I had been considering - but the genuinely novel idea you have come up with is allowing the DO to see the fast positions - I had just thought of constraining them at zero, so only the faster MR system could actually trade them, although they would still have forecasts plugged into the DO. By the way in theory you might not want *cheap* instruments; if you think you will mostly be executing limit orders, then you want a low commission but maybe high bid/ask spread. But for safety for sure you'd want to have cheap when you started trading in case there were more market orders than expected - this is one of the big unknowns about this strategy. I think the capital/ risk question is moot to begin with; I'd want to run this on a single instrument to begin with which would mean minimal capital allocation to test both infrastructure and a proof of concept. Then you could gradually increase your allocation. The relative allocation between a (in theory) higher SR strategy with say one instrument per asset class, and a pseduo diversified DO with forecasts from (in my case) nearly 200 instruments, but realistic exposure to perhaps 15 or 20 at once in my experience with my account size; well that's an interesting theoretical debate. Especially considering it would be hard to get one instrument per asset class in MR if you just focused on minimum capital/cheapness; considering the fact that the 'knowing of positions' would hopefully reduce in more of a marginal improvement in diversification. Really the challenges around the MR are all around the implementation; especially into a system that is driven purely around doing things once a day and not knowing about equilibrium prices. Realistically it's not something I am going to have to time to do in 2024, for reasons I will explain in my end of year post to follow shortly. But I will watch any progress here keenly. Rob