Posted in My Daily Journal · 11 October 2026
A reader warned that fractional lots and tick rounding compound through capital. For a fixed-lot bot the drift lives somewhere else — in the fill convention — so we started measuring it on every trade.
A trader warned that backtests drift from live results because fractional lot sizes and tick rounding compound through capital. Does that apply to a fixed-lot option buyer?
Not that part. Our buyers trade a fixed one lot, never size by a percentage of capital and never compound, and the charge formula in the bot reproduces the backtest's to within a few rupees per trade when reconciled against the broker's contract notes. The same class of gap does exist in one place: the fill convention. The backtest buys at the next minute's close plus 0.1% and sells at the trigger close minus 0.1%; the live bot sends a market order a few seconds into the next minute and a market sell a few seconds after the trigger candle completes.
How big can that difference be on a real fill?
Several rupees per unit on fast bars. One live gap-and-go trade signalled at a 240.1 close and filled at 243.5; its trail triggered at a 233.7 close with the bid quoted at 233.65, and the real sell printed at 231.5 — ₹2.15 below the quoted bid. Each leg is a few rupees per unit, which on a 65-unit lot is a few hundred rupees per trade — the same order as the rule's whole expected edge on small-gap days.
What did you change?
The measurement, not the rule. From that day every trade — live or paper — records the backtest's reference prices (next-minute entry close, trigger-close exit) beside the actual fills, as rupees per unit and per trade, with a running cumulative in the end-of-day report. A review rule was set with it: if the cumulative live-versus-backtest gap per trade is worse than −₹100 after fifteen live trades, the first thing to cut is the small-gap trades, which only earn about ₹27 each on paper and cannot absorb that slippage.