I model professionally. HFT, MM, rust.
Joined January 2015
- Tweets4.3K
- Following506
- Followers3.5K
- Likes4.3K
Pinned Tweet
If anyone wants to read about one of the ways you can get flip flop bugs like this i wrote about it a few months ago. Its basically really easy to do and one of the fastest ways you can incinerate and account.
markrbest.github.io/position…
Mark Best retweeted
Inspired by recent discussion about whether it's possible to learn trading outside an elite firm.
If you're the kind of person that reads this project and thinks it sounds fun, imo you'll be able to learn employable trading skills just fine on your own.
At this rate dead Internet theory is going to be because if you open a website or use an ide your bank account gets drained. The last couple of weeks have been a mess.
1
7
679
Mark Best retweeted
Si jamais vous êtes un solo quant ou solo algo trader qui a déjà une bonne vision sur son environnement mais qui atteint un peu un plafond de verre niveau infra, on peut potentiellement essayer de collaborer (on est une société de trading quant pour compte propre).
Pour faire simple, vous avez déjà produit des strats gagnantes, vous savez chercher de l'alpha mais vous en avez assez de diriger votre stack technique et assez d'être seul (dans le sens où vous avez fait le tour et que vous savez que le partage avec d'autres sera EV+)
On a de l'infra, du capital, de la force de dev, de la data.
L’idée, c’est d’avancer étape par étape : un projet commun minimal, puis le partage d’une strat à faible alpha ou en fin de vie qu’on ferait tourner chez nous. On peut aussi simplement recevoir vos signaux en black box et n’être qu’exécuteurs.
Vous pouvez m'envoyer un MP.
Vous pouvez poser des questions en commentaires si besoin.
PS : l'idée est une collab long terme avec possibilité de s'associer à court/moyen terme (aligner les intérets le plus tot possible pour diminuer la std).
It is possible for both of these things to be true
1) LLMs are the future of programming
2) They currently write absolute throw away garbage
3
19
1,494
Pros of using codex for cooking: You get an easy Belgian waffle recipe. Downside: The mix made 5x more than I wanted.
Looks like I'm eating waffles all day.
1
7
734
Mark Best retweeted
What is the state of art for languages to use with AI ?
-> quality of code (to train the model)
-> quality/number of library (to have out of the box components).
-> constraint by compiler, fail-fast languages, strongly typed, language opinionated....
-> High-quality tooling for checks, tests...
A part of the AI's success is a strong early feedback loop (as a human).
Have you try A/B testing on this.
One backend in Rust vs Go, and Python (with ruff, mypy etc..). Vs. TypeScript?
Many tks !
The biggest mistake I made in life was not buying a flat in london when i was 12.
London Flat prices....
In terms of relative valuation (compared to overall England and Wales house prices), London Flats have round-tripped and are now close to the levels they were in 2006 (when the relative boom started).
10
1,170
Browsing baby toys. This is a great way to teach your kid how to play with plug sockets and defeat child safety locks.
7
681
There is nothing more human than AI vibe coding something one day and then saying it is utter garbage when reviewing it after some sleep.
10
750
Mark Best retweeted
When market making, cross asset correlations are your friend.
What's funny though is you can footgun yourself here if you assume markets are too efficient.
Suppose you're on a secondary exchange making BTC, and some ALTCOIN, using something like the binance feeds as your source of truth.
Your mm engine sees BTC jump 2bips and immediately reprices ALTCOIN, canceling your inner ask. But then nothing happens to ALTCOIN for the next couple of ticks, so your system forgets about the BTC jump and requotes ALTCOIN at the level implied by your feed + some basis adjustment.
A few milliseconds later ALTCOIN does jump and you get picked off. So you had the right idea but the wrong execution. The market for ALTCOIN was actually less efficient than you assumed, and you backed off the signal too quickly.
What's the solution? Modeling. You need to analyze tick data and model the correlations temporally.
The precise impulse response function is something to be discovered empirically, but it's well motivated by the fact that there's an easily observable statistical relation:
ALT_return ~ beta * BTC_return + epsilon
(Or more generally ALT ~ beta_i * f_i + epsilon where f_i are factors.)
For most coins. This leads to all kinds of modeling ideas.