Expectancy tells you a system makes money. It can’t tell you whether that money arrives smoothly or in terrifying lurches — or whether you’ve traded enough to believe the number at all. System Quality Number folds all three into one grade.
What SQN combines: mean R, std dev, and N
SQN, coined by Van Tharp, blends three ingredients that individually mislead:
- Mean R — your expectancy per trade, in R-multiples. More is better.
- Standard deviation of R — how consistent those results are. Less is better.
- N — how many trades you have. More is more confidence.
The insight is that a system’s quality isn’t just how much it makes; it’s how reliably it makes it, and how sure you are the edge is real. A metronomic small edge can be a better system to trade than a lumpy big one. If R-multiples are new to you, the R-multiple explainer and the broader expectancy write-up cover the groundwork.
The formula and Van Tharp’s grading scale
SQN = (mean R / standard deviation of R) × √N
with N conventionally capped at 100 so systems with different sample sizes stay comparable. If you recognize this as the t-statistic of your R-multiple distribution, you’ve spotted the whole idea: SQN measures how statistically distinguishable your edge is from zero.
Van Tharp’s published bands:
| SQN | Grade |
|---|---|
| Below 1.6 | Hard to trade |
| 1.6 – 1.9 | Below average, but tradable |
| 2.0 – 2.4 | Average |
| 2.5 – 2.9 | Good |
| 3.0 – 5.0 | Excellent |
| 5.1 – 6.9 | Superb |
| 7.0+ | “Holy Grail” |
Why SQN rewards consistency and sample size
The standard deviation in the denominator penalizes lumpiness. A strategy that earns its keep from a few rare monster winners has high variance and a lower SQN than a steady grinder with the same expectancy — because the grinder is easier to size, easier to survive, and easier to trust.
Meanwhile √N rewards evidence. A small edge proven over many trades can outscore a big edge shown only a handful of times. That’s what makes SQN a quality number rather than a mere profitability number: it blends how good the edge is, how consistent it is, and how confident you’re entitled to be — the three questions a serious trader should always ask together.
Comparing two systems on SQN
Numbers make it concrete. Take two systems, each with 100 trades:
- System A: mean 0.3R, standard deviation 1.0R → SQN = 0.3 / 1.0 × 10 = 3.0 (excellent).
- System B: mean 0.5R, standard deviation 2.5R → SQN = 0.5 / 2.5 × 10 = 2.0 (average).
System B has the higher expectancy — it makes more per trade on average. Yet A is the better system to actually trade: smoother, more predictable, easier to size on a prop account and far less likely to hand you a stomach-churning streak. That’s precisely the distinction raw expectancy hides and SQN surfaces. Pull your own mean and spread from real trades with the expectancy calculator before grading anything.
The sample-size caveat and score inflation
Because √N sits in the numerator, SQN keeps climbing as you add trades even if the underlying edge never changes — which is exactly why Van Tharp caps N at 100. Two consequences follow:
- Never compare SQNs computed on different N without the cap. The bigger sample wins mechanically, not because the system is better.
- A high SQN on a tiny sample is noise. With ten trades the √N term is small and your mean and standard deviation are themselves badly estimated. The grade is unreliable in both directions.
Treat any SQN below a decent sample as provisional. It’s a t-statistic, so the same statistical humility applies: a flattering score from twenty trades is usually luck wearing the costume of quality.
Using SQN to allocate between strategies
If you run several strategies, SQN gives a defensible way to rank them for capital. Tilt toward the higher-quality edges — the smoother, better-evidenced ones — rather than whichever strategy had the flashiest recent month. A system scoring 3.0 deserves more size than one scoring 2.0, other things equal, because you can trust it and survive it.
Two disciplines keep this honest. Recompute periodically — SQN drifts as regimes shift and edges decay, so last quarter’s ranking can be this quarter’s trap. And watch the confidence around it: SQN’s √N is a crude confidence proxy, and pairing it with a proper interval on your win rate tells you whether a change in score is real or just variance.
That’s the loop Shibiki is built to run for you: auto-journaling captures every fill, live edge health recomputes mean R, variability and sample per strategy continuously, and a Wilson confidence interval on your win rate is the statistical companion to SQN’s √N — it shows whether you actually have the evidence the grade implies. The hard risk limits enforced at the broker keep the sizing sane while you tilt toward your best-graded systems, and copying across prop accounts lets you run the ranked strategies in parallel without re-keying a single order.
Related: Expectancy calculator · Trading expectancy · What is an R-multiple