Wow, now corn and soybeans are at it again, and everything is great Dead-cat bounce? oh, I would definitely took my money and ran at this point if I traded this manually
Obviously we use a different set of rules to determine forecasts. In my system have currently European equity (ESTX50, DJ600) and V2TX their forecast capped at the maximum value. As has HE (lean hogs). Grains are positive but not exceptional.
yeah, actually ESTX50 is also killing it (I only trade it in paper), V2TX and HE are making some profits but not much for me at the moment..
Tried to analyze the effectiveness of my execution algo, looks like it has some benefits (though it's a very crude analysis and I might've messed something up). This is cumulative sum of execution profits and losses from the "worst" side of the spread to the actual fill (current ask-fill if I'm buying, fill-bid if selling). (it can't achieve mid-point execution, but it seem to do better than giving away full spread every time): So looks like I made about 600$ in the last year or so compared to executing on the worst side of the spread. Although the large positive jumps look strange, this thing should have negative skew, so small frequent gains and large rare losses, the large rare gains are weird. Probably some edge-cases like a lost connection or something, maybe I'll investigate them more closely..
Would be interesting to see the breakdown by future (provided you have enough data). My bet is there would be some with a graph like a straight line up, and some that are like a straight line down. In which case you should turn off the ones with the straight line down.
Why do you think an execution algo can have different effectiveness on different markets? I mean they are all "equally-liquid" I think, what other factor might influence execution., like maybe we expect to have better execution in grains vs equities, but why? I can pretty-easily get the total sums for every market, generating individual plots is more cumbersome., But I would want to formulate a hypothesis first i.e. what am I trying to test., because for sure some markets will have better execution that others but how would I know that it's significant and jot just randomness (especially because there's not much data, some markets had only 4-5 orders in the last year) Or you mean do it just to find the bug easier (the instruments which had these weird high execution gains)?
Different microstructure across markets. Tick size most importantly. Logically, futures with big tick sizes would gain the most from passive orders, whereas small ticks could be easier to take aggressively.
Whenever I've done this (Admittedly not for a few years) I haven't seen this - just a very random series with a slight edge but a hell of a lot of noise in every market. GAT
My breakdown is like this (this is only last 9 months as I fixed some bugs in the exec algo at that time): here's the individual plots (sorry for excel, I really need to get more fluid with python ) So as I said, the sample size is very small especially for some instruments. ZT seems to be really good, and MES, Nikkei, BTP(only 10 points) and ZS even beat mid-point execution. Don't know if there's any pattern there, only maybe that smaller-size contracts have better execution, which might make sense because I'm a very small fish and my orders are tiny so easier to "scalp" especially on smaller contracts?..