Lisboa V3 — every lap's line

Hover a lap number to pick its line out of the stack; click to drop it from the overlay or put it back. Italic numbers are the out-lap, the two much slower laps and the in-lap — they are drawn grey because their times are not comparable.

With all 48 laps up at once the overlap averages the colours, and because most of your laps sit close to your best the stack reads blue — which is itself the answer to “am I consistent?”. To see the fast and the slow lines apart, use Fastest 10 vs slowest 10, or zoom into a corner: the strands separate and each keeps its own colour.

The ink is not guesswork. Across the 638 m of track, a stroke this wide covers a median of 14 of the 48 laps at once (8 at the loosest station, 24 at the tightest), so each lap is drawn at the opacity that brings that many to the ink target — 16.5% each at the defaults. Change the lap count, the line width or the zoom and the figure is recomputed, which is why the readout above reports it.

How wide is the band you actually used?

Lap 14, the fastest, is the reference line. Every one of the 48 flying laps is measured perpendicular to it at 1 m stations around all 638 m, and the band is the 5th to 95th percentile of those offsets. As drawn: median 1.99 m, from 1.43 m to 3.13 m. With each lap's rigid offset removed: median 1.04 m, from 0.43 m to 2.68 m. The truth is between the two lines below — the shaded gap is the part a whole-lap displacement could account for. Tick “Measured band” above to see it on the track, and “Remove per-lap offset” to switch between the two.

• 1 · 3.13 m wide
  145 m around the lap — click to zoom there
• 2 · 2.91 m wide
  45 m around the lap — click to zoom there
• 3 · 2.65 m wide
  579 m around the lap — click to zoom there
• 4 · 2.54 m wide
  451 m around the lap — click to zoom there
• 5 · 2.35 m wide
  292 m around the lap — click to zoom there

Tightest as drawn: 1.43 m at 233 m, 1.51 m at 544 m, 1.53 m at 11 m. At that width the laps are identical to within what this receiver can separate — do not read a narrow band on a straight as consistent driving. The numbered places are ranked on the drawn band; click one to zoom there.

What this picture can and cannot tell you

The imagery has been moved, not the GPS. The saved visual alignment for this session shifts the Esri imagery 3.356 m west and 1.854 m north (-14.4 and -8.0 pixels) so it sits under the recorded track. Every coordinate you see plotted is the logger's own fix, untouched, and the lines are pixel-identical with the alignment on or off — only the photograph moves. Turn “Imagery alignment” off to see the raw registration.

That fit is subjective, made by eye against a single lap, marked provisional and under open review; a later orthophoto comparison suggests a different north component. It is also a pure translation — no rotation or scale was fitted, so any rotation or scale mismatch between photograph and track is still there, and would vary around the lap. Its size matters too: the 3.83 m shift is roughly twice the 1.99 m median width of the band it is positioning. A line that appears to run over the asphalt edge is not evidence that a wheel did — the photograph is at least as likely to be the thing in the wrong place.

About a third of the width is the receiver, not the kart. Each lap carries its own near-constant displacement: fit the best rigid shift to each one and the median is 0.88 m, reaching 1.88 m at the worst. Take those out and the band falls from 1.99 m to 1.04 m — 48% of what you see is whole-lap displacement. Tick “Remove per-lap offset” to watch it happen. A real difference in line is not a rigid translation of an entire lap, so most of what that removes is the receiver's frame wandering over the 37 minutes — but some genuine displacement goes with it. So read the two numbers as a bracket: the true driving variation is somewhere between 1.04 m and 1.99 m, and this recording cannot say where.

The receiver's own accuracy estimate — 0.71 m median, 0.80 m at the 95th percentile, from a median of 13 satellites and never fewer than 7 — is a figure for absolute precision and is not the thing that limits this picture. What limits it is that per-lap 0.88 m drift. So a whole lap sitting half a metre off another one is not a finding. What survives is shape: where a lap runs wide relative to its own line elsewhere, the rigid part cancels and the comparison holds. Read the corners against the straights, not one lap's absolute position against another's.

And the lines trace the logger's GPS antenna, a single point on the kart. The kart is wider than a point, so the area it actually swept is wider than the band drawn here.

Which laps count. The device recorded 50 boundaries. The default view draws only the 48 flying laps.

Fastest 43.622 s (lap 14), slowest flying 49.477 s (lap 34), median 44.2625 s across an even population, so the median lies between two observations. Colour is linear in lap time from the fastest lap to 44.861 s, the 90th percentile, and saturates above it: the times are right-skewed, and letting 49.477 s set the far end would squeeze the forty laps that matter into a third of the scale. The ticks under the ramp are the real lap times, so you can see the clustering and the tail for yourself. “Even spread by rank” stretches the ramp evenly over the ordering instead, which separates the cluster but is no longer proportional to time.