What is the most troublesome thing about quantitative trading?

kuebiko · Post by **kuebiko** » Sun Aug 19, 2018 10:38 pm

@anonq I honestly don’t know but I think the RKS stuff/paper is all a joke...

BTW what frequencies do you trade?

anonq · Post by **anonq** » Mon Aug 20, 2018 12:37 am

That would make sense... something more than mildly interesting coming from one of the banks quant research teams would be rather surprising

Basically a standard stat arb book these days with maybe higher turn than the norm, everything from very short to month hold alphas, colocated but primarily to reduce market impact, couldn’t survive on the purely intraday signals

RubyYao · Post by **RubyYao** » Mon Aug 20, 2018 4:37 am

I share a strategy on another thread named "Blockchain Quantitative Investment Series Course (3) - Intertemporal Arbitrage", which includes complete code. glad to talk about it with you guys Big Smile

kuebiko · Post by **kuebiko** » Mon Aug 20, 2018 7:12 pm

@anonq cool. Are you saying your intraday stuff just can’t scale/generate enough vol without creating significant market impact? Presumably the lower frequency stuff is also lower Sharpe but higher capacity. Curious how you think about combining “alphas” on these different timeframes (say a collection of hourly signals and a collection of monthly signals)....

anonq · Post by **anonq** » Mon Aug 20, 2018 8:48 pm

@kuebiko yeah can’t scale due to market impact without significant work on the execution side and/or additional short term alphas, but always tempting to play that game but kinda intimidating with the likes of virtu who probably has more invested in tech infrastructure than I have gmv

Yeah lower frequency is incredibly high capacity but much lower sharpe and am seeing some alpha decay there, my guess is that it’s so easy to execute that more funds have that kind of stuff. Used to know people over a decade ago running multi billion gmv rank based portfolios on a handful of signals off IBES but that doesn’t work at all anymore, whereas the short term stuff from back then still holding its own

Assuming one has accurate expected returns and good market impact modeling then the best way by far is a multi period optimization, decent number of papers on the topic now but easier said than done. Currently have what I’d call a ghetto version of that, still working on getting better expected returns

Do you do anything in this area?

kuebiko · Post by **kuebiko** » Mon Aug 20, 2018 9:39 pm

@anonq yeah I run lower frequency futures strats (~week up to months turnover). I’m interested in trading faster but my current firm doesn’t have the interest (they want scale) or really the expertise/infrastructure to do it competitively at the moment. The dimension is so much richer at shorter timescales, but I’m limited by the lack of a good market impact model due to execution people being siloed (or vice versa).

I’m definitely aware of and interested in the multi-period optimization stuff, have read the papers and played around a bit. As you say it’s far from trivial. Just curious if there were any other good ways to think about it.

Thx!

Jurassic · Post by **Jurassic** » Mon Aug 20, 2018 10:33 pm

Why are the dimensions richer at shorter timescales?

kuebiko · Post by **kuebiko** » Mon Aug 20, 2018 10:39 pm

I guess I’m being a bit loose with “dimension,” but at the longer extremes it seems everything is explained by a few principal components/risk factors. You haven’t got much data, and it’s very hard to get away from things that look like momentum or carry. Of course, it’s not impossible and I think there’s still some interesting stuff that can be done, but there are many more uncorrelated, robust, and predictable effects, both directional and cross-sectional, when you go intraday.

Jurassic · Post by **Jurassic** » Mon Aug 20, 2018 10:41 pm

What do you classify as not much data? you could get daily equity prices for 15 years +. Its only things like interest rates where its once a month

Zoho · Post by **Zoho** » Mon Aug 20, 2018 10:47 pm

@anonq

>Used to know people over a decade ago running multi billion gmv rank based portfolios on a handful of signals off IBES but that doesn’t work at all anymore, whereas the short term stuff from back then still holding its own

It is also said that lower volatility levels during the last several years preclude long-short books from earning. Low frequency ranks based on fundamentals partly confirm this but it looks like short-term based ranks still hold their edge. I would explain this happens due to the fact that it is not that obvious to produce short term stuff that works and this helps for their alpha to keep on working despite low volatility environment... What do you think?

What would you think about this topic

[url=/Show%20Post.aspx?PostIDKey=187134]Pca-based portfolio properties[/url]