From my many years working as a professional quant trader, here’s a piece of advice for everyone. Anyone here presenting you “alpha” is trying to sell you something. True alpha/edge takes significant time and energy to develop and must be protected at all costs.
Literally zero successful traders or fund managers do stuff like this (or spend lots of their time doing podcasts, writing a paid substack, selling a course etc)
Which makes for easy decisions — if someone does any of those things, by definition it’s not worth listening to them. They self-selected out of the relevance pool.
For actual trading resources:
1. Elements of Statistical Learning by Hastie, Tibshirani, and Friedman
2. Network Algorithmics by Varghese then build a deterministic PCAP replay simulator with a simulated clock
3. Red Hat’s low latency tuning guide, extra credit: ef_vi user guide
Good luck. No Discord.
@macrocephalopod Fund manager here, 17 years at top quant firms. Even if a successful trader wrote a guide, almost nobody would want it. Trading is too much actual work (e.g. kernel bypass) not get rich quick. If a “trader” is selling you a course instead of trading, that is the business.
@mert I’ll broaden to trading UI as it’s all the same. Unsatisfying answer, all consumer products are awful, professional ones aren’t far behind. The best I’ve seen are all internally built. Why? No web bullshit, fast, full featured, and customizable.
@Melt_Dem Not explicitly a business around trading of data, but similar in spirit. Alpha capture (started at Marshall Wace) was an ingenious idea for its time. Though it’s been replicated to hell now.
@Melt_Dem That’s quite broad, trad tick/rt data alt data? For the latter, Yodlee sells anonymized consumer banking data (think plaid). IIRC RS metrics sells satellite imagery, more relevant back in the day for parking lots to forecast earnings.
@lex_node Pour over or french press that should be fine, not great for espresso. Idk what reasonably high end means for you, but my mazzer has been running 20 years strong - recently upgraded to weber eg1, no complaints.
@mdudas@FalconXGlobal Haven’t alaways been their biggest fan, however giving them the benefit of the doubt - maybe they’re trying to build something like gsam? Would definitely require true separation like BDs do to avoid conflicts of interest.
@optimizoor If one ran kernel bypass on the networking side and had proper core isolation for processes, it would yield significant gains beyond evm change as well. Even more extreme measures would be offloading ecdsa recover/validation at ingress with a smartNIC.
@optimizoor The EVM itself is fine, it could be improved. However node design is a much larger issue. Running everything in generic async frameworks is incompatible with performance. In extreme case, if I had to guess rolling SPDK would yield heavy as @ben_an_adams alludes to.
@gakonst All this additional bloat complexity (e.g. dns, js, 3rd party vendors) inherited from legacy tech creates a lot of issues in defi imo. Make defi simple again.
@gakonst If we are talking about irt dapps, I threw together some scripts which generate simple html files, local dns like resolution for contracts from ABIs, triggers cast to be used in lynx. Would be better with foundry integration if all projects had deployment artifacts.
Retail and institutional traders aren’t even playing the same game. I surmise that most of retail would find what we do boring, as it often resembles a regular - albeit demanding - job.
This is kinda interesting because “you should iteratively improve your trading system” is excellent advice for a professional operating in a professional environment with the discipline and tooling that prevents them from overfitting (indeed it’s basically mandatory — if you don’t iteratively improve then you will be overtaken and die).
But it’s awful advice for a retail trader who doesn’t really know what they are doing, because they don’t know how to iteratively discover real improvements and they will just overfit instead.
@EvgenyGaevoy Agreed. The take is akin to labeling latency arb in traditional markets as “frontrunning”. The information being traded on is public and there is no fiduciary responsibility to other market participants.