Quant Memo
Foundational

Price Rounding and Precision Conventions

Every market rounds prices to a fixed tick size and rounds calculated values (interest, P&L, shares) using specific, sometimes non-obvious conventions, and getting these wrong quietly corrupts backtests and reconciliations.

Prices in every market are quoted in discrete increments called ticks — a US stock can't trade at $10.005, only at $10.00 or $10.01 under the sub-penny rule — and any calculation that produces a price in between has to be rounded to the nearest allowed tick before it can actually be traded. The same logic applies beyond quoted prices: interest accruals, converted share counts, and P&L figures all get rounded somewhere, and the convention used (round half up, round half to even, always round down) is a small detail with real consequences.

The rounding rule matters most in three places: converting a theoretical fair value into an executable limit price, computing per-share dividends or splits that don't divide evenly, and aggregating small daily rounding differences across a large position over a long backtest. A "round half to even" (banker's rounding) convention, common in accounting systems, will silently disagree with the naive "round half up" a spreadsheet does by default, and over thousands of rows those disagreements accumulate into real basis points of tracking error.

Rounding conventions are not cosmetic — mismatched rounding between a research system and the real trading venue is a common, boring cause of backtests that look profitable but don't reproduce live.

The safest habit is to round only once, as late as possible in a calculation, and to match whatever convention the venue or custodian actually uses rather than whatever a library defaults to.

Related concepts

Practice in interviews

Further reading

  • SEC, 'Regulation NMS: Rule 612 (Sub-Penny Rule)'
ShareTwitterLinkedIn