Risk management for small account algorithmic traders
I think about small account algorithmic trading as a problem of limited room for mistakes. A trade can lose for an ordinary market reason, and software can also create exposure its operator did not intend. I run an automated Kalshi prediction market system from my Mac lab in Honolulu. Since the October 11, 2026 restart, as of November 11, its real settled record is 1,236 bets, with 1,115 wins, 121 losses, a 90.2% win rate, and net +$250.93. Beside that headline, 66 bets lost half their stake or more, including a gold bet at a loss of $61.44 and a SOL bet at a loss of $39.26. These are individual losses, not a measured account drawdown. This article treats risk as its own design problem: define what the account can actually lose, treat unknown order state as exposure, count shared causes rather than only market names, separate the decision to pause from the decision to exit, protect the boundary around live software, and keep risk reporting separate from reassurance.
Authors
- Jeremy Koski
Publication Details
- Journal
- Figshare
- Published
- 2026-11-12
- DOI
- https://doi.org/10.6084/m9.figshare.34993696
- Primary Topic
- Financial Markets and Investment Strategies
- Type
- preprint