Consistency in racing: what does the percentage actually mean?

In a one-lap qualifying session, peak pace wins. In a 30-lap race, consistency wins. A driver who alternates between 1:33.0 and 1:36.0 averages 1:34.5 — and their rival who runs 1:33.8 every lap wins, even though the inconsistent driver's best lap is faster.

PaceBoss surfaces a consistency percentage on every driver-detail page. Understanding the formula turns a number that can feel arbitrary into a diagnostic tool.

The formula

PaceBoss prefers the consistency value SimResults publishes when it's present (the source of truth for official championship pages). When it's missing — usually for shorter sessions — PaceBoss falls back to the coefficient of variation:

consistency % = 100 − (stddev / mean) × 100

Where stddev is the population standard deviation of your valid lap times and mean is the average of those same laps.

A driver whose every lap is identical has stddev = 0, so consistency = 100%. A driver whose laps vary widely has a higher stddev relative to mean, and consistency drops accordingly.

What the percentage feels like in practice

For a 90-second lap:

• 99.5%+ — sub-450ms standard deviation. Elite. Your fastest and slowest valid laps are within ~1 second of each other across the entire session.
• 99.0–99.5% — 450–900ms stddev. Strong amateur. Most laps within a couple of tenths of average, occasional half-second wobbles.
• 98.0–99.0% — 900ms–1.8s stddev. Mid-pack. Identifiable "fast" and "slow" laps.
• 97.0–98.0% — 1.8–2.7s stddev. Inconsistent. Likely a few good laps and several scrappy ones.
• <97.0% — 2.7s+ stddev. Either a long strategy session or genuine instability.

These aren't hard cut-offs. A 60-second sprint at Mount Panorama has more absolute variance than a 90-second lap at Imola. The percentage normalises for lap length, but the human read is contextual.

Where the number can mislead

• Pit laps inflate the stddev. PaceBoss filters out pit-in/pit-out laps (anything >1.5× your median) before computing, so a routine pit stop doesn't count. But a slow in-lap that's still under 1.5× will count.
• Strategy variance looks like inconsistency. A driver running a longer first stint on harder tyres has slower laps mid-run, faster at the end. That's valid strategy, not inconsistency.
• Short sessions are unreliable. With three valid laps, the stddev is meaningful but high-variance. PaceBoss requires at least two valid laps to compute.

How to use it

If your consistency is below where you want it:

• Look at the pace-trend chart's shape. Is the variance random (technique) or trended (tyre degradation, fuel saving)? A flat-bottomed trace with occasional spikes = isolated mistakes. A V-shape = warmup or strategy.
• Sort the laps table by sector. If S2 has high variance and S1 and S3 don't, you have one section within S2 where you're inconsistent. PaceBoss can isolate it to the sector but not to a specific corner — for that, drive the sector against an in-game ghost or watch your replay against a steadier driver.
• Compare to your competitors. If your consistency is 99.0% and the race winner's is 99.6%, you're losing the race in the gap between those numbers — not in peak pace.

Related reading

• Theoretical best lap, explained — the related "what could you have driven?" ceiling.
• Pace bands explained — sector-level deltas that often reveal which sector is dragging consistency down.
• Find your pace deficit — turning a low consistency reading into a specific sector to attack.
• Looking ahead — the technique that drives consistency improvements before peak-pace ones.
• ACC controller settings — settings often hold consistency back before technique does.
• Glossary: pit lap — the heuristic for excluding pit laps from the consistency computation.