There's a lot of the talk on WSO about HF fundamentalists needing to learn how to code. However, most of the coding that is discussed is about web scraping and model making. Should fundamentalists also learn machine learning to provide insighs?
| +59 | Looking for a Fund Manager to partner with to start a new fund. | 28 | 1d | |
| +36 | How to destress at Pod HF? | 16 | 1d | |
| +25 | Any R course recommendations? | 14 | 3d | |
| +24 | Starting Personal Account | 12 | 5d | |
| +24 | Fastes PM progression ever seen? | 29 | 1h | |
| +20 | Track Record Expectation For PM's | 29 | 16h | |
| +16 | How extroverted is the junior MMHF job? | 5 | 2d | |
| +16 | HF perspective on Google (Alphabet) | 5 | 2d | |
| +14 | Recommend Pass for MM HF Case Study | 4 | 3d | |
| +13 | Why don't MMs do any activism? | 10 | 11h |
Career Resources
Get instant access to lessons taught by experienced private equity pros and bulge bracket investment bankers including financial statement modeling, DCF, M&A, LBO, Comps and Excel Modeling.
The core reason, in my understanding, is that the data points are too sparse. In the quant side, people are using features with frequencies at least on the order of minutes. Even with that it's still hard to make money by trading, say, 1000 stocks. On the fundamental side, each year there are only a handful earnings reports for a company. And each analyst only covers one or few sectors. How do you extract enough information to predict the price with machine learning, based on very sparse data, and expect to make money by trading less than 100 stocks?
Thanks! But instead of using ML with stock movements, what about using it with alternative data? Would be interesting to see trading executed by computers based on strategies designed by fundamentalists.
yes funds are already looking through credit card, web traffic, and satellite data to find signals via ML/stats
Yes, ML is being used at fundamental shops, usually as an overlay for confidence/weighting/veto over security selection, its also used by fundamental shops to gauge risk mgmt environment
all shops are heading towards ML shops, its not a binary thing its just a tool in the box
That's very interesting. I thought software was used to filter prospective investments/trades and then a human evaluates and makes the decision.
Is it worth it for a fundamentalist to learn ML? I'm already learning Python for web scraping and model building.
highly recommend understanding how it works, the statistical foundations.
even if you never touch an ML model, interacting with those who do is crucial
For ML to be useful in this field, you would need to gain enough skills and experience such that it makes you practically a data scientist already.
Unless you are truly extraordinary I am skeptical the amount of machine learning you can pick up by just casually dabbling in it would be useful in a professional setting.
dabbling will not lead to professional usage
dabbling will help build a vocabulary for interacting with the middle office who actually use it
I'm learning python right now, should I also learn SQL for ML?
In my view there are very few fundamental funds/strategies for which meaningful available alt data dense enough to require machine learning and also have a worthwhile return-on time. If you have huge AUM, a small coverage universe, sufficient infrastructure/team and shorter term investment approach and the sector is focus is one of the few that current have really deep alt data available than it can be worth the time.
Why the huge AUM?
I think you need a high potential P&L in absolute dollar terms to offset the data purchase price and time spent.
few clarifying comments about how ive seen ML used at $4b+ fundamental shop:
so while fundamental analysts will likely stay fundamental, the rest of the infrastructure is going to use best practices, including ML for whatever their function
I think the answer is you use exactly zero of these skills at a fundamental hedge fund on average.
I actually work as a well respected quant fund now after working at a pretty prestigious hf and we STILL don’t use machine learning. If you are pussyfooting around with overly complex models, you’re probably going to miss a lot of opportunities.
Do I believe hedge funds tell their LPs they use machine learning? Absolutely. LPs might on average be slightly dumber than people on this forum so it makes sense they’d say that. The “data scientists” at hedge funds are marketing spend. You buy them for the same reason you buy nice office furniture.
Or firms that have data scientists come in thinking they're going to be using statistics and ML to solve the market. In reality, they double-check that the S&P500 closing mark is right.
That's interesting! Would it be helpful, for a fundamental role, if I learned python?
no absolutely not. have fun instead. nobody wants your python skills.
i've heard for example that some pms will hire really impressive quants to do a basic linear regression in python rather than excel. the purpose of this is of course just marketing, not the actual use of your skills. learning python without the impressive credentials defeats the objective (marketing).
ML in this context is really about alt-data and not advanced statistics or programming. There are some funds using it successfully, but they were doing it before it became mainstream and don't market themselves as doing ML specifically. The basic ML methods are important but are a commodity that everyone uses, and are not an edge. Most of the funds that talk about ML and hire data scientists got into it after everyone started talking about it, and for them it's just a form of marketing to gather assets.
Consequatur enim nihil sint accusantium debitis et. Rerum aspernatur quia consequatur provident inventore et excepturi. Hic ipsa beatae inventore voluptatem ducimus qui. Accusamus qui tempora vitae tenetur et eos aliquid.
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...
Adipisci libero odit quod consequuntur. Quo alias voluptates sunt maiores. Excepturi dolorum ut sapiente. Quia nisi recusandae culpa ut adipisci expedita eveniet temporibus. Quia suscipit nemo deleniti nulla aliquid. Mollitia amet et voluptates dolor voluptatem nihil culpa hic.
Earum omnis sint ut sit recusandae. Ut ut error repudiandae incidunt distinctio impedit. Blanditiis ut quas nihil maiores laboriosam fugit consequatur. Repudiandae ut et alias eligendi et occaecati aliquam. Rem quidem voluptatum veniam.
Rem quibusdam tenetur culpa dolor ut in omnis. Voluptatem maiores suscipit ut commodi.