Comment by bofadeez
3 months ago
It's a lot more complicated than that. Eugene Fama might be a good starting point for you.
(Spoiler: most PhD level machine learning quant trading funds do not make money. Marcos Lopez de Prado can explain why that is the case: they accidentally invest in false discoveries. It's so pervasive that I often argue it's a numerical analytic proof for EMH. Most quant funds you've heard of are engaged in high-frequency market making, they provide liquidity by continuously quoting prices and earning the bid-ask spread with limited directional risk. It's not behavior forecasting. Something like order flow toxicity has predictive ability for like a few milliseconds.)
Something like this does not work: https://www.tradingview.com/v/9Ut0yL2p/
Even if it has a good p-hacked backtest with some parameters, it's not going to work in practice. It's just a novelty by/for academics.
The question is not "do all quants make money modelling human behavior?" It's "do any quants make money modelling human behavior?"
And are there any quant funds that consistently make money? No. It's like professional gambling. Susquehanna actually bets on sports too. Constant losses (hopefully slightly) offset by gains. There's no such as a trader with high accuracy/precision. If you can even edge out 1% (gain-loss) you'll be a billionaire from $0 very quickly. A positive edge is so exceptionally rare, it's like a startup -> unicorn. It would be the equivalent of owning a casino with trillions of dollars being played nonstop. It almost never happens.
If there was any fixed analytic solution to behavior then a simple neural network would approximate it quite easily. It's almost all noise with only spurious correlations and misleading patterns that don't repeat.