The main risk with smaller exchanges when you are doing smol trader arbs is really just that so many have fake flow.
No point doing all that research only to realise it was all fake flow.
First thing to check
🔔🚨 Quantitative Developer Role In Our HFT Team:
Fantastic Developer with an Interest in Finance:
(Quant Developer or QD want-to-be)
Did you grow up tinkering with computers? Taught yourself to program at a young age, and now wish to approach analytical problem-solving with a high degree of rigor? Went to school knowing exactly what you wanted to do - get paid to solve engineering challenges in a competitive field?
We prefer outcomes over paths and a demonstrated history of problem-solving over credentials.
Extremely talented developers with at least an interest in quantitative finance are encouraged to apply.
Rigorous, engineering-heavy pipeline centered around building and maintaining a tight ship, with relatively simpler mathematical requirements. Speed is the name of the game.
We’re a small, tight-knit team that lives, breathes, and sleeps our work, and are looking for a like-minded team player to join us.
While our stack is in Rust, we’re interested in all extremely talented developers with an intense and deep understanding of networking, low-level hardware, and complex debugging. If you get excited about different flavors of garbage collection, the advantages and disadvantages of Rust over CPP, and computer networking, you’ve come to the right place.
Previous applicants for other roles at the firm, feel free to apply.
Help out with a retweet for a follow back :)
Application Details Below ⬇️
IPCA is a godsend over IV regression. Instrumental variables are just hell, I’d much rather work with IPCA’d data.
Also, whoever is doing daily delta hedging for their delta hedged returns should look at optimal delta hedging since volatility differs
https://t.co/5KtzfBxq9G
After lots of optimisation, Newton Raphsom appears to be faster if you dgaf about precision beyond 1dp for IV estimation.
Annoyingly it isn’t great at SIMD trying to get a whole lane to converge because it’s outliers / jitter that’s the issue.
Otherwise, Jackel is great.
Can obviously vectorise both but Jackel is much easier / also has way less jitter.
I think if I optimised Jackel a lot more / we were comparing vectorised then we’d see them have roughly the same speed or maybe faster for Jackel but wayyyy less jitter for Jackel.
…