Balyasny cuts human-machine hybrid team Synthesis
Haven't seen this posted yet: Balyasny just cut 10 people running a $2 billion book. The hedge fund axed the 1-year-old team because of poor performance, sources say.
Apparently Balyasny created a unit called Synthesis last year that was supposed to blend human and machine. I'm guessing they got the inspiration from the study that showed human-computer teams outperformed both human-only and computer-only teams in chess. Interestingly, the study showed the best human-computer team wasn't comprised of grandmaster + computer, but just some random dude + computer.
Anyways, they recruited 9 sector heads that were PMs at places like Point72 and Citadel to provide the sector-specific knowledge (I'm guessing to provide the domain expertise to train the machines). Guess it hasn't been doing too well and they shut it after barely a year. Too bad, it sounds really cool and I bet this could have worked if they maybe worked out the kinks. Anyone else have any other info?
Hi Acidophilus, check out these links:
Fingers crossed that one of those helps you.
I made a post about balyasny on here a while back, they dont know how to hire tech/ML people and they suffer for it
Chemical synthesis is a very interesting and important question that is clarified every day in our life. But only several people know the real significance of this tools influence on our life. First of all, such type of devices provide convenience in using your mobile phones, tablets etc. Protecting their screen from dust and other garbage. But it is not the only using of such type technologies.
Most non tech people don’t know how to hire technical people. It’s how a lot of software companies swindle management into paying for their service
I'm having a hard time figuring out how this could have potentially worked...
The algo provides some type of action with some confidence level and then what? How do you combine the human and digital input? How did this even happen in the chess situation? Computer and human each make a turn sounds retarded... But then how does the human know when to interfere? Is it some type of randomized process? Maybe the computer asks the human to take a decision based on intuition or something when multiple different moves have exactly the same expected outcome
Freestyle chess
Good question, I'm not sure. The link above sort of describes how it works in chess. The idea is that the computer takes care of the number crunching, while the human governs the overall strategy and direction i.e. tells the computer where to number-crunch.
maybe I am a luddite but i think machine learning is great for identifying images of cats, and not proven in predicting financial markets. at least in its current form.
These hybrid processes are very tricky to get right. You need a good pragmatic quant team, and open-minded, collaborative fundamental specialists. Generally, the humans (sector specialists) provide domain expertise, which is used across a large systematic book (1000+ positions).
One example: each sector specialist forecasts current quarter's EPS/revenue, which the algorithms use as part of its process (factors).
The gist being that a sector specialist has better insight into a company than the sell-side consensus.
It's hard to get right because fundamental specialists don't know how to hire quants, and vice-versa. You need a good leader who can speak both languages.
Your last point about hiring right is crucial. The "hiring manager" (for lack of a better term) needs to be good at evaluating/selecting both fundamental analysts and quants, and needs to be able to strike the right balance between the two to avoid building a monoculture.
what I usually see happen is the manager for fundamental analysis just hires some guy with a fancy pedigree, like they do for analysts, and then the guy sucks
usually the fundamental PM hires a quant so they can either market as using quant or hoping it will juice returns. Its not because they suck but the fundamental PM isn't trying to really build out a quant / fundamental team and usually tries to hire as cheaply as possible. This usually means a PhD with no experience or a quant who comes in and just does their own thing. Actually trying to combine the two is difficult.
The example you gave seems to be limited to one specific type of alpha. For a pure quant team it wouldn't make sense to hire a fundamental analyst covering 50 stocks. Not sure how much value they would add in the case you mention. The quant can always just not bet on earnings.
Honestly it was a interesting concept to try, as someone who isn't a quant I'm king of glad it didn't. If it did would further make the case humans (as the majority) wouldn't be needed.
Is it just me or the link not working?
That's right. It doesn't work for me either.
Alternative link
Hey sorry the original was a cached version of the BI story. I found another link that should work. Can't update the original post though.
Sunt dolorum iure exercitationem assumenda omnis. Officiis expedita iure sunt sed. Voluptatum ab dolore earum et tempora architecto. Voluptatibus laudantium nam aut voluptatibus delectus voluptas. Provident ullam dolor odit eum pariatur.
See All Comments - 100% Free
WSO depends on everyone being able to pitch in when they know something. Unlock with your email and get bonus: 6 financial modeling lessons free ($199 value)
or Unlock with your social account...