List of books on machine learning, data science, or just traditional quant stuff I actually read and enjoyed without pretending, have a nice weekend!
- Ernest P. Chan - "Quantitative trading: how to build your own algorithmic trading business"
- Keith Fitschen - "Building
@centrdevils
I think he's actually gone to Liverpool to be drafted to the first team on the spot, since their mid is half-empty, while at United he'd have to compete with Eriksen.
Probably one of the best machine learning repos out there: . A bunch of tips and solutions on applying ML to trading, from data research to even deep reinforcement learning.
@picotop
True, a former colleague told me a story of an intern who had thrown up on MD’s shoes and was the only one who received an offer. This was inv banking though.
Someone just said “screw it” and created ~1400 slides for a 2023-2024 course in portfolio allocation and asset management, all free, straight from Paris: . The last lecture is on machine learning in AM!
And here is yet another position on financial machine learning you may have overlooked by accident, really good one, though also very scientific which makes it a bit less comfortable to read
This repo is an absolute unit, based on a pretty rough but great book "Machine Learning in Finance: From Theory to Practice" by Matthew Dixon, Igor Halperin and Paul Bilokon, have fun:
@MichaelAArouet
Poland pretty much stopped being a joke about 15 years ago. The interesting trend nowadays is that Poles living abroad have been coming back and many Westerners or Nordics are moving to Poland as well (especially pensioners).
Here is an interesting repo on financial data science you can take a look at today:
Pretty good examples of applying various models or building strategies.
Junior quants will nail their interview by solving Riemann hypothesis in less than 10 minutes just to do this on their first day:
rets = np.log(px_close.diff())
Pretty cool repo with various tools & guides used in financial machine learning. The downside is that it’s outdated, so some of the repos listed may not really work anymore.
Alright, have something neat for you this weekend -- have stumbled upon this financial machine learning repo recently, you may find stuff that interests you, from theory to practice:
Probably the biggest transition shock is the biologist with a PhD from MIT joining RenTech after having applied complex deep neural nets in genomics or protein structures in their previous job, just to run a linear reg model with one feature now.
Reminder of an old modelling trick that never dies:
- Linear regs (l1/l2/elasticnet) as a starter.
- LightGBM for pretty much everything as it’s fast and accurate.
- CatBoost if you have many categorical features.
- Random Forest if your OOS variance is too high from LightGBM.
-
I've stumbled upon this book somewhere here on Twitter and have put it through the notebook LM; man, it's gold!
And finally a good position that talks about different asset classes, not just equities. Bond/credit/alternatives risk premia, fx carry, commodities; of course, vol,
Junior quant interviews: “try solving the Riemann hypothesis in less than 10 minutes”
Senior quant interviews: “your model has equity R2 of 0.7, is that odd? Also, how would you go about your day if you woke up as a worm?”
Have just started reading this one, as it was circling around on my twitter feed a while back; wonder how it differs from the Ernest Chan's book I've tweeted about recently.
After I had recalled the "100-page ML book", I got asked about other ML positions for starters. I think this one is also great, particularly if somebody wants to dive deeper into financial machine learning
Another interesting repo to check out, FinGPT: ; applying AGI to asset allocation and trading. The caveat is, it thrives on OpenAI API, would be better it the LLM was free but worry not, I want to do some research after my PhD is done.
Tech hiring nowadays: "oh, you have a PhD from MIT in rocket science and recently worked on developing software that sent Artemis to the moon? Cool stuff, here is the Hackerrank interview, we need to see if your Python is strong enough"
Using traditional quant and machine learning you can:
- detect a regime shift using a technique of choice (e.g., Markov, SKF) and create labels out of it;
- apply trees to predict the switch ex ante;
- create a strategy following the predicted regime and optimise it;
- profit!
You'll like this one: . A repo containing links to a bunch of packages on data sources, feature engineering, backtesting, trading, and risk analysis (among others). Interestingly, not only for Python but also, e.g., R, Julia, or Java. Enjoy!
Stealing from
@quant_arb
's manual, this is an interesting repo to go over:
A lot of helpful implementations, many you have to treat with a grain of salt (not "Kalman Filter and Pairs Trading"!) but at least they can direct you at the right path.
Thanks to someone who recommended me PythonAnywhere, it's far superior to Google Colab and way cheaper than SageMaker. It's literally as if you were working locally but can schedule jobs with ease and set up your own database.
My favourite quants of all time:
1. Jim Simons
2. Max Cohen from Pi (1997)
3. Alan Turing
4.
@CliffordAsness
5. Ken Griffin
…
69. Oscar Martinez & Ryan Gosling ex aequo
…
420. Alex Gerko
…
777. Myself
Did I mention that this repo exists: ? Pretty much all you need to start (and continue) your journey in classic and modern quant finance. Something to browse over the weekend!
I’ll keep reminding about this: fractional differencing is my favourite way of transforming time-series data, where your only worry is the order of integration between 0-1 (higher -> less memory, more stationary; lower -> more memory, less stationary).
OK, so I read about Nested Clustered Optimisation from
@lopezdeprado
couple of days back, and now I just found the way to simply implement it step by step with ease: . Riskfolio has tons of stuff like this, no need to join Goldman to become a top quant.
Great read from Two Sigma on regime modeling via machine learning:
Even though unsupervised, the GMM approach can be used as a label generator for a subsequent multiclass classification (ex ante).
IMO, you can do everything in financial machine learning with these two models: linear/logistic reg and lightgbm. Random Forest also works but then again you can just use a ‘rf’ booster in LGBM. And forget about neural nets.
"Machine learning for quantitative finance: fast derivative pricing, hedging and fitting"
TL:DR:
- Gaussian Process Regression (GPR) can be used to effectively model and predict various financial quantities, such as option Greeks like Gamma, implied
Pro tip for Sunday: Best way to move from finance and economics to quant / machine learning is to dive deep into econometrics. You'll learn statistics required to understand ML and become familiar with (almost) every problem related to time series data.
I very much like this repo with various code, videos, books, interviews, papers, etc on financial machine learning, even though it’s quite “old” (4y). Check it out:
Riskfolio tutorials are an absolute gold, like this one: "Multi Assets Algorithmic Trading Backtesting with Vectorbt",
I wonder, has anyone here tried playing around with it using expected (like preds) instead of historical returns; literally port =