I did consulting with WorldQuant and I am not a big believer in how these funds are built. Algo developers all chasing the same edge. Adding in noise so it doesn’t seem correlated but actually is. Using ML to super-fit nonsense algos. Mean reversion over windows of a few days which works until markets trend down (or up) strongly. Alternative data that works well in-sample but disappoints in real time or works for months - year. Algo developers have some knowledge of what the funds strategy will be and can front run on private books or leak to others.
Getting thousands of consultants to independently crank out algos and then combine them has issues.
Where did you hear about how respected they are? They cut all their freelance consultants and kept all their IP without paying ongoing royalties for it. Shafted thousands of poor quants around the world. Then started operations back up under a new name WQ Brain and disallowed any previous person who previously worked with them to get back on. Honestly, that was one of the biggest scams vs humanity I have ever seen in this business. Quants in 3rd world countries slowly built up to $500/month in residual revenue from their algos (took them years to do so). Then in 1 night that was taken away and they were given nothing in return.
WQ had every manner of alt data you can think of. But the stuff that really worked was various spins on short term mean reversion. They neutralized everything and had holding periods of a few days.
Although their ‘alphas’ were secret, I could tell because all the ‘alphas’ I submitted related to mean reversion were already in production and too highly correlated. Add some noise in there and they’d accept it though.
I don’t doubt your take on their ethics. Apologies for the confusion an any personal feelings towards them. I was just referring to alpha and their performance to build and be trusted by or with: 1,000 employees, 9 billion dollars in AUM, 24 offices, 16 year existence etc:
Sounds like there may be an business opportunity if that IP is valuable and those less fortunate are looking for work. Business can be harsh. Apologies again for any confusion or misunderstanding.
Re neutralization, do you have any ideas for how that would work? I suspect it’s not dissimilar to the alpha / risk model discussion.
No apologies needed. They are a big fund but I am not a fan of how they use aspiring quants dangling a carrot in front of them. They work for years to make chicken scratch and then basically give all of them the boot and keep their algos. Anyway…moving on…
Everything is run long / short with liquid stocks and they neutralize for sectors and industry groups and so forth. I am not 100% sure how it was done because I would design an alpha and then the backtest would already incorporate all of this in it. They didn’t let you drill down and see individual holdings because then for sure you could easily front run it. You knew which analytics you were using and what you were designing. You could pick your stock universe somewhat but mostly just the top 3000 stocks based on liquidity.
So to answer your question…sorry…I am not 100% sure how they neutralized it behind the scenes. Maybe a risk model. Maybe some kind of sector exposure balance both long and short…not sure.