Hacker News

New stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Deterministic Concurrency [video] (youtube.com)
    —discuss
  2. Buzz: Share spare compute and run open models together over P2P networks (iroh.computer)
    —discuss
  3. Show HN: A game about fake news and memes (unspin.app)
    —discuss
  4. Local-First Conf Recap (inkandswitch.com)
    —discuss
  5. Arm and SoftBank: Part 1 (thechipletter.substack.com)
    —discuss
  6. Brazil Bans Online Betting (reuters.com)
    —discuss
  7. U2 Celebrates 50th Anniversary (bbc.com)
    —discuss
  8. Show HN: KISS – A highly performant agent harness inspired off Pi built in Rust (github.com/racetozero)
    —discuss
  9. Zambia Approved an HIV Drug in 12 Days (asteriskmag.com)
    —discuss
  10. S3 Is the Future, S3 Is the Past (btrblocks.com)
    —discuss
  11. Nvidia 5090 DLSS 5 power hits 647W, power connector runs hotter than the GPU die (tomshardware.com)
    —discuss
  12. Primal Solver (github.com/c-vision)
    —discuss
  13. It Got to My Field (4gravitons.com)
    —discuss
  14. LeanAPI: API Servers for web applications written in Lean 4 (github.com/theoriclabs)
    —discuss
  15. Jev Plays Pokémon Red (LIVE): an AI decision model plays the whole game [video] (youtube.com)
    —discuss
  16. Bob Mackie dressed stars–if they were brave enough (economist.com)
    —discuss
  17. Why house prices may be in trouble (economist.com)
    2comments
  18. Luntrack – food, workouts and runs in one daily log (luntrack.com)
    —discuss
  19. Hasarak – don't miss what's missing (hasarak.com)
    —discuss
  20. Solon – Docker on Windows Without Docker Desktop or WSL (github.com/v94lere)
    —discuss
  21. How does one keep up with exponential growth? (ezzeriesa.notion.site)
    —discuss
  22. The Subtle Art of Advertising as Taught by Mad Men (teamlewis.com)
    —discuss
  23. Our Big Dumb AI Gods Are Wrong (massopen.ai)
    1comments
  24. Automattic has a new board after failed attempt to put CEO on leave (techcrunch.com)
    3comments
  25. Market making PnL theoretical limits
    —discuss
  26. NEC building 1-petabit capacity subsea cable for Meta (japantimes.co.jp)
    —discuss
  27. X Club (wikipedia.org)
    —discuss
  28. OpenAI’s Systems Went Rogue and Meddled With U.S. Government Websites (nytimes.com)
    2comments
  29. How to keep enjoying programming in a world of LLMs (haskell.org)
    —discuss
  30. Ending procurement and forced use of paper straws (2025) (whitehouse.gov)
    —discuss

Market making PnL theoretical limits

1 pointsby 49m ago
0 comments
I have more or less working more or less market making algorithm (eat the spread).

It works extremely well at 1 lot trades, but once you increase the volume profits reach maximum at some point once you do significant turnover compared to the rest. It's kinda obvious, you got adverse selected more and more.

What bothers me - maybe I simply quote inefficiently? I tried a lot of of approaches (imbalance, skewing, VWAP, hand crafted signals), and at some point they stopped improving my PnL.

Any way to estimate theoretical capability of the specific market? Is this market already exhausted and you just can't do anything else here.

Assuming that market indeed can be exhausted, it's not some US bonds/top stocks with infinite volumes.

And side question about backtesting. Is it even worth it? Now I filter past trades with new signals. Predictions of new trades suck for me, in real life competitors instantly react on my orders, so backtest overestimates PnL.

A quiet thread, for now.Start the conversation on HN ↗