I’ve been asked by my esteemed colleague @goodalexander to write a thread on recent industry trends within Quant Trading. My background is mostly within the mid-frequency space, so I’m likely to miss hugely obvious points to those in parallel industries. Apologies: 🧵

The easiest way for a fund manager to improve their performance is not by finding outright alpha (the space is too competitive now), but rather, it’s the ability to use existing knowledge in new spaces.

New spaces can mean newer asset classes, or different timeframes, resulting in two independent but closely related paths with relatively less competition: these paths have driven trends in recent years.

Intraday trading is well known but over the past few years, the barrier to entry has really lowered, and with that, there’s been a big push towards these higher performance strategies. A lot of systematic infrastructure can lend itself to intraday models (due to being built on…

generally evenly spaced time intervals), but more importantly, banks have been pushing more intraday index products which IMO, aren’t the smartest/best/worth buying, but it’s a sign. To me, it signified the mainstreaming of fast strategies.

Likewise, and parallel to this, newer geographical regions make for good progress as traditional strategies like momentum/carry/value etc work kind-of-OK everywhere, so when certain regions open up their domestic markets to international traders…

the results are often (initially) robust. Usually, there’s much less competition, less arb’ing, and as such, more alpha to take. The perfect combo is where there’s a structural block that pertains long-standing alpha (say, a high domestic retail participation).

The majority of asset classes are still pushing ahead with quantification & automation. Concepts like using optimisers for optimal position sizing & factor neutralisation are becoming more important in less quant-y spaces (think Equity L/S, Macro / Commods etc) because…

alpha is richer if you remove certain factor exposures, or balance your risk better. Becoming a bit more technical offers higher rewards. Unfortunately though as well, some mutli-strats now penalise PM’s for factor exposures!

There’s also been developments in other asset classes. Systematic single name credit trading has really pushed forward with most (high quality) teams making an effort in this space. Trading these names is difficult though because the market isn’t the most liquid, prices…

aren’t that transparent and historical data isn’t very trustworthy - but a few teams have propped up making good coin, so again: robust & existing strategies in newer places.

Crypto is, of course, omnipresent. A lot of long only managers and funding arb players initially gained but eventually both suffered. Now, alpha is much harder to find but can be in various weird settings (think the guy who broke Mango).

I’ve not seen much stat-arb being prevalent in the space. Although, successful traders exist, I would direct any question to @worstcontrarian here.

I know nothing about Market Making - so please ask @AlexanderGerko or @jgreco for more information. These guys are the bee’s knee’s when it comes to this stuff. Moving onto retail trading…

Retail trading makes me jealous now. “Back in my day”, I was stuck using yahoo finance and executing after dinner on my retail account, but now, there are so many platforms like Quant Connect with cheap & easy ways to digest data to help retail traders run quant strategies.

Honestly - the models you make yourself are no different to those at a Hedge Fund (only except we work full-time at it) - so keep trying and keep inventing. I’m actually surprised with the amount of alt-data out there for retail as well. Lot’s of alpha.

This leads me onto my final few points about the industry. HF launches are nearing all time lows as every manager now has to question if they should launch their own fund (headaches including regulation/asset raising/hiring/operations/marketing - ….

and that’s all before making industry leading performance)… or…joining a pod-set up, or launching a new arm to an existing fund. Fee structures are also changing as established behemoths try to hold onto all their cash for as long as possible, making it even hard for new funds.

So if you’re a PM with a sharpe > 3 and can scale to over $1bn, I’d say you probably have a shot. Anything else…probably better off with a 15-20% pay out at a multistrat! Anomalies exist, of-course.

Anyways, TLDR: ML/AI doesn’t work. Crypto is complicated. Use age old strategies in new ways & new markets. I’ll do my best to answer any questions and apologies if I missed anything out. The above is already quite lengthy!

@ThePythonQuant As someone who is not a trader, do you know of any good resources for reading about this se "age old strategies"?

@ThePythonQuant Forgive me but where in the thread were you alluding to in your conclusion that AI/ML doesn't work? Do you have more context from your experience?

@ThePythonQuant I didn’t see you mention ML in the thread. Doesn’t it work at all? Or are there some uses for which it does work?

@ThePythonQuant Thanks, appreciate you sharing your obviously extensive knowledge & experience.

@ThePythonQuant You earned a sub

@ThePythonQuant @readwise save thread