A hot new hedge fund is based on smart computers picking off dumb ones

I somewhat fail to see the value proposition here. This confluence of HFT and stat arb sounds a bit dubious to me because they have been related for a long time. It's not as if HFTs either black and white just deterministically arb or quote. Stat arb is not new here. HFTs know how to do PCA on an ETF and every single dispersion trading desk has a statistical idea of where implied correlation likes to hang out. Yes, some HFT strategies are fast enough to sacrifice analysis of edge cases in favor of speed for deterministic arbs or first on new price level formation quoting but to say that a combination of HFT and stat arb is new is untrue, and I think Manoj would agree with me there. That being said he does mention some cool stuff.

Does anyone have any examples of the non linear (spread) modeling techniques he's talking about? We've covered some linear techniques in school but I think the only nonlinear idea we've covered that applies is copulas.
 
Last edited:
Buddy I do not know non linear spread modelling or similar fancy stuff, but if this guy is right, a time has come when trade positions will not be merely taken on fundamental or technical variables but strange and exotic facts like how many folks in starbucks order a type of coffee on Mondays and/or what percent of people sleep in more than 2 hours on saturdays etc. etc.

Now how such facts are going to determine the price of IBM stock is a matter of statistical modelling. These models will eventually tell them strange items like " when this happens then on thursday trading there is 90% chance of IBM stock gaining more than 1%".

In the last 10 years, artificial intelligence based algorithms took over more than 60% of trading jobs and in the next 10 years "statistical intelligence" based algos will have a major showings in the order flows.

HOWEVER !! these models are based on the data from the past ONLY. Thus any new "outlier" type variable hits on above IBM example on thursdays trading, then all "exotic variables" can go out of the window and IBM stocks dips instead and stops are taken out at the light speed of the algos.
 
Last edited:
It's called marketing. Manoj is playing up his HFT street cred to differentiate his quant fund from the rest. What's more likely is that his prior firm was beaten by competitors and lost its alpha as markets and participants evolved. Same thing appears to be happening with Teza. Survival of the fittest. The outlook for marginal players in HFT isn't pretty, and established players have to keep spending more to maintain ever-decaying alphas.
 
You must log in or register to reply here.
Back
Top