Skip to main content

Metrics (Win Rate, Avg R, Expectancy, Drawdown, Profit Factor)

Updated Jun 24, 2026 at 2:35pm

Research Draft High 1,225 words

These are the five core performance statistics a swing trader uses to judge whether a system actually has an edge and whether that edge is survivable. The central tension is that no single one is sufficient — a high win rate can hide tiny winners and one catastrophic loss; a great expectancy can come with a drawdown deep enough to make you quit before it pays off. They must be read together, and they only mean anything across a large enough sample. Together they answer two separate questions: does the system make money on average (win rate, avg R, expectancy, profit factor) and can I psychologically and financially withstand the path (drawdown).

How they're calculated

All of these are best expressed in R-multiples, where 1R is the dollar amount you risk per trade (distance from entry to initial stop × position size). Normalizing to R makes results comparable across tickers, share counts, and account sizes — a convention popularized by Van Tharp (Hightick).

  • Win rate = winning trades ÷ total trades. (Loss rate = 1 − win rate.)
  • Average R (per win/per loss) = mean R-multiple of winners and of losers. By construction losers average roughly −1R if stops are honored.
  • Expectancy = (Win rate × Avg win in R) − (Loss rate × Avg loss in R). With average loss ≈ 1R this simplifies to (Win rate × Avg winning R) − Loss rate. It is the average R you expect to net per trade; above 0 = positive edge (Hightick, P&L Ledger).
  • Profit factor = gross profit (sum of all winning $) ÷ gross loss (absolute sum of all losing $). PF > 1 = profitable; PF = 1 = breakeven (QuantifiedStrategies, Titan FX).
  • Maximum drawdown (MDD) = the largest peak-to-trough decline in equity before a new peak, usually quoted as a percentage (PrimeXBT).

A useful sixth derived metric is Van Tharp's System Quality Number: SQN = (mean of R ÷ standard deviation of R) × √(number of trades) — i.e. expectancy over the dispersion of R, scaled by sample size. It rewards a high average R and low variability. Tharp's published bands (his own heuristic, not validated constants) run roughly: <1.6 poor, 2.0–2.4 average, 2.5–2.9 good, 3.0+ excellent — and he notes the number is unreliable below ~30 trades (journalplus, Hightick).

How they're used in practice

A master swing trader reads these as a connected set, not a scoreboard:

  • Win rate alone is noise. The famous point is you can win <50% of the time and thrive if winners dwarf losers. A trend-following swing system riding a few large moves often runs 35–45% win rate with avg winner of 2.5–4R; a mean-reversion/pullback system runs 60–70% win rate with smaller ~1–1.5R winners. Neither is "better" — they're different distributions of the same expectancy.
  • Expectancy is the headline. It tells you per-trade edge. Multiply expectancy by trade frequency to get expectancy per unit time (Tharp's "expectunity") — the number that actually grows the account. A 0.4R system trading 200 times a year beats a 0.8R system trading 40 times. Tharp's commonly cited benchmark band for a sound system is roughly 0.2R–0.8R per trade (P&L Ledger).
  • Profit factor for quick triage. Commonly cited practitioner bands put a solid system at ~1.5–2.0+ (some sources suggest swing traders target ~1.5–2.5, scalpers ~1.2–1.5); a backtest PF much above ~3 — and especially above 4, or near ∞ — is usually a sign of overfitting, survivorship bias, or too few trades, not genius (QuantifiedStrategies, BacktestBase). Live results typically degrade from backtest (slippage, fills), so build in a margin.
  • Drawdown sets position size and survival. The actionable rule of thumb: every positive-expectancy system will produce losing streaks, and they're statistical certainties. For win rate W (independent trades), the probability that any given run of N trades is all losses is (1−W)^N; over many trades the expected longest losing streak grows roughly as log(T) ÷ log(1 ÷ loss-rate) — about log₂N for a coin-flip system, so ~6–7 losses in 100 coin-flip trades (Pomegra, BacktestBase). Concretely, a 40%-win system has 0.6⁵ ≈ 7.8% chance (~1 in 13) of any specific five-trade window being five straight losers, and over a realistic 50–100-trade stretch is nearly certain to hit at least one ~5-loss string — so a swing trader risking 1R per trade must plan for ~5–8R strings of losses and size so that doesn't end the account or the will to continue.

Concrete failure modes Augustus should flag: a system whose entire edge rests on one or two outlier winners (drop them and expectancy goes negative); avg loss creeping well past 1R (stops not honored, or gapping risk — a real swing-trading hazard on overnight holds and earnings); and a backtest with too few trades to trust any of these numbers.

Adoption, debate & evidence

R-multiples, expectancy, profit factor, and max drawdown are near-universal in retail and professional trading literature and in every reputable backtesting/journaling platform — this is settled vocabulary, not a fringe technique. The measured caveats are well-documented:

  • Sample size dominates everything. With <30–50 trades, win rate and PF are dominated by luck; a few dozen trades cannot distinguish skill from noise. SQN explicitly scales by √N to penalize small samples.
  • Profit factor and expectancy are outlier-sensitive. Both can be dragged up by a single huge winner that may never repeat. Robust evaluation removes the top trade and re-checks.
  • In-sample vs out-of-sample gap is real. The academic and quant consensus is that backtested metrics overstate live performance due to overfitting/data-mining; only walk-forward or out-of-sample results, and ultimately live trading, validate an edge. Treat any single in-sample number skeptically.
  • There is no universal "good" threshold. Benchmark bands (PF ~1.5+ good, SQN >2.5 good, expectancy 0.2–0.8R) are heuristics from practitioners (Tharp, quant blogs), not laws — they vary by trade frequency, market regime, and asset.

Strengths & limitations

Strengths: Cheap to compute, comparable across systems via R, and jointly they capture both edge and risk. Expectancy × frequency is genuinely the best simple predictor of account growth, and drawdown is the best simple predictor of whether you'll stay in the game.

Limitations: All are descriptive of the past — they assume the future distribution resembles the sample, which regime change violates. They say nothing about path beyond MDD (two systems with identical expectancy can have very different streak structures). Profit factor and win rate are silent on trade-size variance, which is exactly where ruin hides.

The #1 misuse: optimizing for win rate (it feels good to be right often), which quietly pushes traders toward wide stops and quick small profits — a high win rate married to oversized losers and undersized winners is the classic blown-account profile. Optimize for expectancy and drawdown, treat win rate as a byproduct.

Sources

Note on disputes: benchmark bands (PF, SQN classifications) are practitioner heuristics, not validated constants; all backtested values are widely held to overstate live performance.