Position sizing for a newly deployed trading bot: how to start safely
Conservative sizing for new bots: halving rules, volatility targeting, kill criteria, and escalating size only when live stats match research.
Position sizing new trading bot and risk management algo trading deployment searches are about the transition from research to live capital.
Principles
- Cap per-trade and per-day loss in dollars, not only percent of backtest profit
- Scale in after observing stable fills and slippage (Why live fails)
Pair with kill criteria
Define drawdown and drift stops before you need them (Strategy kill-switch, Freqtrade kill-switch).
Why small size is a feature, not a failure
The first weeks of live trading are an operational test: fills, latency, exchange quirks, and your own discipline. If you start at full size, you pay for operational surprises with maximum pain. A smaller notional keeps mistakes expensive enough to notice, but not existential.
Volatility-aware sizing (high level)
If your strategy risk scales with market volatility, write down how you will tighten size when realized volatility jumps beyond what your backtest assumed. If you cannot explain the rule in one paragraph, it is not ready for automation.
Escalation rules
Only increase size when live behavior matches research assumptions across multiple regime slices, not after one lucky week. Tie increases to pre-written criteria: stable slippage, stable trade count, and no unexplained drift versus walk-forward expectations (WFE).