Lane Fairness

What data is used

Only heats (voorwedstrijden) are included in this analysis. Heats have randomly assigned lanes, which is essential — without random assignment, faster crews might systematically end up in certain lanes and we could not separate lane effects from crew quality. Finals and semi-finals are excluded because lane assignment there is based on previous results, not random.

What the bars show

Each bar shows how many seconds slower that lane is on average compared to the fastest lane. The fastest lane always shows 0.00s — it is the reference point. All other lanes are expressed as a disadvantage relative to it. The error bars show the uncertainty in that estimate — a wide bar means fewer races contributed and the estimate is less reliable.

Two separate statistical tests

There are two questions being answered, and it is important to read them together.

1. Overall test (the banner above the chart)
This asks: "Could the pattern across all lanes have arisen purely by chance?" Concretely: if all lanes were perfectly equal, how likely would it be to see lane differences at least this large just from random variation in race results? The percentage shown as a p-value is that probability. If it is low — say below 5% — we conclude that something real is going on and that lanes are not all equal on this day. If the p-value is high, the observed pattern is consistent with pure noise and we cannot draw conclusions about lane fairness.

2. Per-lane test (the symbols † * ** *** next to each lane)
This asks for each individual lane: "Could the gap between this lane and the fastest lane have arisen purely by chance?" The symbol reflects how unlikely that gap would be if all lanes were equal.

†  p < 0.10 — borderline, treat with caution
*  p < 0.05 — significant by the standard threshold
**  p < 0.01 — strong evidence
***  p < 0.001 — very strong evidence

When the two tests seem to disagree

It is possible that one or two lanes show a significant symbol, while the overall banner says the differences are not significant. This sounds contradictory but makes sense: the overall test looks at the full pattern across all lanes at once and is harder to pass. If only one lane stands out while the others are all similar, the overall test may not find enough evidence to conclude that lane differences as a whole are systematic, even though that one lane looks suspicious individually. In such cases, treat the result with extra caution.

Important caveat

A significant result means the effect is detectable, not necessarily large or practically meaningful. With many races even a tiny difference can become statistically significant. Always read the effect size in seconds alongside the significance symbol. Conversely, on a day with few races a real disadvantage may not reach significance — absence of evidence is not evidence of absence.