Hacker News new | ask | show | jobs
by gpm 377 days ago
I think it's fair to question how proprietary your data is.

Like there's the algorithm by which a hedge fund is doing algorithmic trading, they'd be insane to take the risk. Then there's the code for a video game, it's proprietary, but competitors don't benefit substantially from an illicit copy. You ship the compiled artifacts to everyone, so the logic isn't that secret. Copies of the similar source code have linked before with no significant effects.
2 comments
short_sells_poo 376 days ago
Most (all?) hedge funds that use AI models explicitly run in-house. People do use commercial LLMs, but in cases where the LLMs are not run in-house, it's against the company policy to upload any proprietary information (and generally this is logged and policed).

A lot of the use is fairly mundane and basically replaces junior analysts. E.g. it's digesting and summarizing the insane amounts of research that is produced. I could ask an intern to summarize the analysis on platinum prices over the last week, and it'll take them a day. Alternatively, I can feed in all the analysis that banks produce to an LLM and have it done immediately. The data fed in is not a trade secret really, and neither is the output. What I do with the results is where the interesting things happen.

link
FuckButtons 377 days ago
AFAIK, the actual trading algorithms themselves aren’t usually that far from what you can find in a textbook, their efficacy is mostly dictated by market conditions and the performance characteristics of the implementation / system as a whole.
link
short_sells_poo 376 days ago
This very much "depends".

Many algo strategies are indeed programmatically simple (e.g. use some sort of moving average), but the parametrization and how it's used is the secret sauce and you don't want that information to leak. They might be tuned to exploit a certain market behavior, and you want to keep this secret since other people targeting this same behavior will make your edge go away. The edge can be something purely statistical or it can be a specific timing window that you found, etc.

It's a bit like saying that a Formula 1 engine is not that far from what you'd find in a textbook. While it's true that it shares a lot of properties with a generic ICE, the edge comes from a lot of proprietary research that teams treat as secret and definitely don't want competitors to find out.

link