Driver debrief

LisboaKart, Pontinha · layout Lisboa V3 · kart 28 · 7 October 2026, 17:49–18:19 · 37 running laps in three recordings · third session with GPS on the kart

The fastest session in the log on every measure — best 43.239s, median 43.604s, 30 of 37 laps under 44 s — and the one that can least be credited to anything. It was the first session in kart 28, in the coolest air recorded, 29 days after the last one, with two stops in it. The telemetry can say where the time came from: 5 sections, led by the long corner at 229–333 m, each quicker through more speed on an unchanged route. It cannot say whose doing that was. It also settles the experiment left standing on 8 September, though not the way that page expected: the corner it targeted never lost the time it was asked to win back. That loss was the GPS trace sitting two metres from where it had sat four days earlier, and it takes most of that debrief's section findings with it.

Best lap

43.239

sheet lap 24; 0.492s under 8 Sep

Fastest 5 avg

43.286

0.501s better; the five span 0.083s

Median lap

43.604

0.660s better than 8 Sep

Laps under 44s

30 of 37

8 Sep: 9 of 47; under 43.5s 16 against 0

1 · What this debrief can and cannot say

Measured The printed sheet is transcribed correctly, and its total is not the sum of its laps. The 39 lap times average 45.498s against a printed 45.498s, with the best (43.239s) and last lap (43.302s) as printed. They sum to 1774.432s; the printed total is 1811.416s, 36.984s more. The recording shows where that time is: before sheet lap 1 the kart had already driven an out-lap of 57.3s, passing the line at 20 km/h on its way out of the pits. A clock that started 36.984s before lap 1 began started during that out-lap. These checks run every time this page is built.

Measured One session, three recordings, every sheet lap accounted for.

Recording	Sheet laps	Logger clock	Logger cut its own laps at	From its stored point	GPS fixes	Stock parser would drop	Gates vs sheet, median / worst
`a_0004`	1–7	—	131 m into the lap	8.9 m	9,693	6,057	39 / 86 ms
`a_0005`	9–16	carries on from `a_0004`	the timing line 1×, then 131 m into the lap	9.2 m	10,898	0	24 / 50 ms
`a_0006`	18–39	restarted	the timing line	2.6 m	25,625	0	16 / 65 ms

You stopped twice — your account: on lap 8 for a few seconds without touching the logger, and on lap 17 to restart it so that it would record from the proper start/finish line. The logger closed its file at each stop, and its clock agrees with you: it runs on across the first break and starts afresh after the second. GPS time shows the kart below walking pace for about 20 s and 33 s, each recording resuming within 2 m of where the last one broke off, and it reproduces the sheet's two stop laps — 8 (72.415s) and 17 (83.147s) — across those breaks to within 102 ms. Against a median lap the stops cost about 29 s and 40 s.

Measured Where the logger cut its laps, from positions. The logger's stored start/finish is a single point, the same in every recording so far, and the circuit passes it twice a lap: on the start/finish straight, and again 9.3s later on the return stretch, 131 m further round and about 9 m to one side. a_0004 and a_0005 cut their laps on the return stretch — the displaced line 8 September had, within 0.3 m on the ground — except that a_0005 counted the true line once before settling there. a_0006, recorded after the restart, cuts on the straight, a median 2.6 m from the point, as 4 September did. That confirms what the lap records and clock stamps suggested, with one correction: the displaced line is 9.3s after the true one, not about 8 s. Why the logger counts one passage rather than the other is not established.

Method Everything is timed through the same 13 gates as the last debrief. They are 4 September's section starts: a point on the ground plus the direction of travel through it. The logger's own lap line plays no part, so the two recordings with the displaced line are as usable as the third. The gates found 37 complete circuits — 7 in a_0004, 8 in a_0005 and 22 in a_0006, one for every running lap on the sheet — and their lap times agree with the sheet to a median of 17 ms, 86 ms at worst (lap 1). Section numbers are 4 September's throughout, as before.

Method — new this time Each lap is slid into one positional frame before it is timed. A lone GPS receiver puts its whole trace a metre or three away from where it would put it on another day — or twenty minutes later. Against fixed gates that moves time from one section to another; section 5 shows how much, and what it did to the last debrief. So each lap's trace is shifted east and north — no rotation, no stretch — until its sideways position at the 13 gates best matches 4 September's solo laps, and only then timed.

Session	Recording	Laps	Typical slide needed	Size	Range across laps
4 September	`a_0002`	46	0.0 m north, 0.3 m west	0.3 m	0.2–1.3 m
8 September	`a_0003`	47	1.9 m south, 0.3 m west	1.9 m	0.8–3.9 m
7 October	`a_0004`	7	1.2 m south, 0.5 m west	1.4 m	0.5–2.5 m
7 October	`a_0005`	8	2.4 m south, 0.0 m east	2.4 m	1.9–2.5 m
7 October	`a_0006`	22	3.0 m south, 0.7 m east	3.1 m	1.6–3.8 m

Sliding each recording as one block instead of lap by lap moves no section median by more than 0.040s. Differences below about 0.05s are treated as nothing on this page.

Measured The recordings are clean. 46,216 GPS fixes at nominal 25 Hz over 30.8 minutes and 26.3 km. Satellites used ranged 6–16, most often 15; the receiver's own accuracy estimate had a median of 58 cm (54–90 cm), which the slides above show to be optimistic about absolute position. In a_0004 the stock parser's guard would have discarded 6,057 of its 9,693 fixes, as it did on 4 September; the patched reader keeps them, and the lap times built on them match the sheet as well as the others'. The other two recordings lose none.

Not available No engine or pedal data, for the third session running. RPM recorded zero throughout all three recordings, and this logger has no brake or throttle channel. With a change of kart in the middle of the comparison this is the missing measurement that matters most: nothing here can tell a stronger engine from a later brake.

Pending driver input What you were working on, why kart 28, and how it felt against 19 are not recorded yet. Nothing below assumes an answer, and section 9 is provisional until there is one.

2 · The circuit, and what the section numbers mean

Every section number on this page refers to this map, and they are 4 September's, as on the last page. Blue badges are corners, grey are straights, and the yellow ticks are the gates. Numbering runs the way you drive it, from the start/finish marker.

Aerial imagery: Esri, Maxar, Earthstar Geographics and the GIS User Community. User visual calibration — provisional best visual fit: Esri imagery moved 3.356 m west, 1.854 m north. GPS coordinates and timing are unchanged; this is a subjective display alignment.

3 · The session as it ran

Kart 28 for the first time, after two sessions in 19. You had the circuit to yourself throughout — your account, and the sheet's single driver row. Sheet lap 1 is already a lap of the circuit: the kart crossed the line at 55 km/h to start it. The two laps with a stop in them are left out of every pace figure, which leaves 37 running laps in three stints.

Stint	Sheet laps	Recording	Laps	Mean	Spread	Best
Opening	1–7	`a_0004`	7	44.589s	0.702s	43.990s
After the first stop	9–16	`a_0005`	8	43.582s	0.161s	43.358s
After the second stop	18–39	`a_0006`	22	43.549s	0.250s	43.239s

Measure	4 Sep, solo	8 Sep	7 Oct	Change vs 8 Sep
Best lap	43.940s	43.731s	43.239s	-0.492s
Fastest five	44.072s	43.787s	43.286s	-0.501s
Median lap	44.336s	44.264s	43.604s	-0.660s
Mean lap	44.399s	44.446s	43.753s	-0.693s
Spread, all laps	0.300s	0.785s	0.547s	-0.238s
Spread, laps under 45.5s	0.300s	0.328s	0.366s	+0.038s

4 Sep: its last 21 laps, driven alone. 8 Sep: its 47 laps of the circuit. 7 Oct: the 37 running laps. All three were driven on an empty track.

Measured The whole distribution moved, and the middle moved most. Against 8 September the best lap is 0.492s quicker and the fastest five 0.501s, but the median improved 0.660s and the mean 0.693s. 30 of 37 laps were under 44 s against 9 of 47, and 16 were under 43.5 s, where neither earlier recording has a lap.

Measured Nothing was thrown away, and repeatability is unchanged. Raw spread reads 0.547s against 0.785s, but 8 September contains two laps with an off and this session contains none: after the opening lap no running lap was slower than 44.761s. Among laps under 45.5s the spread is 0.366s over 36 laps against 0.328s over 44 — the same repeatability at a quicker pace.

Measured The pace arrived with the second stint. The opening seven laps averaged 44.589s; 8 September's mean was 44.446s. The first lap after the first stop was already 43.542s, and the two later stints averaged 43.582s and 43.549s.

Confound Four things changed at once. The kart: 28 for the first time. The weather: across this stint the air was 22.8 °C at 17:00 falling to 22.0 °C at 18:00, humidity 62–74%, cloud 53% falling to about 30%, wind 17–18 km/h gusting to 39–41; on 8 September it was 27.2 °C falling to 25.7 °C under full cloud — so about four degrees cooler, the coolest session in the log, and on a surface that had seen 1.4 mm of rain the day before and 6.6 mm three days before where 8 September followed twelve dry days. The gap: 29 days since the last session. And two unplanned rests. Any one of these could carry the whole 0.660s. None of the gain on this page is credited to driving, because nothing here can separate it. The weather is a proxy from a model grid point a few kilometres away, and air temperature is not track temperature.

4 · The experiment from last time

The 8 September debrief set one target: the corner at 572–610 m, S12, back to a median of 2.821s or better over at least 10 laps — the time 4 September's solo laps set there — from the 3.081s it measured that day. It asked for the same kart and no other change, and it wrote its own rule: if the median did not return, the difference was not driving.

Verdict: missed as written — and void, because the loss it chased was never there.

As written: Timed exactly as on 8 September, through the same gates, this session's 37 circuits (7 in a_0004, 8 in a_0005 and 22 in a_0006) give a median of 3.117s through S12: +0.036s against 8 September's 3.081s, and 0.296s short of the target. Not one of the 37 passes reached 2.821s; the quickest was 2.969s.
The control: The target asked for kart 19. This was kart 28, so even as written the result could not have been laid on the driving.
In one frame: With every lap slid into 4 September's frame first, S12 reads 2.816s on 4 September's solo laps, 2.839s on 8 September and 2.828s now. It did not lose 0.260s on 8 September; it moved +0.023s. There was nothing to win back, so the target can be neither met nor missed.
What is left: S12 has been a 2.82–2.84s corner in all three sessions and in both karts, while the lap around it found 0.70s.

Supported interpretation By the rule written on 8 September, a median that did not return would have been read as an equipment problem. It was a measurement problem. The rule was sound; the baseline it was applied to was not.

5 · A correction to the 8 September debrief

That debrief compared its sections with 4 September's by timing both recordings through the same gates, on the assumption that the two shared one positional frame. They do not.

Measured On 8 September the whole trace sat 1.9 m south, 0.3 m west of 4 September's. One rigid slide explains 96% of the sideways differences between the two sessions at the 13 gates. Those gates face in every direction; a different driving line could not shift the kart the same compass way at all of them.

Measured That slide moves time between sections, and it predicts how much. A displaced trace meets each gate early or late by its displacement along the direction of travel. On a straight the entry and exit gates face the same way and the errors cancel. Round a corner they do not: the section gains at one gate what its neighbour loses, and the lap total is untouched. The slide fitted from sideways positions alone accounts for 98% of the path-length changes the last page reported — the 4.1 m “shorter line” through S2 (3.6 m predicted) and the 3.2 m “wider” S12 (3.2 m predicted) included.

Corner	Distance	Turns through	Change as published	Path change, measured	Path change, from the slide	Change in one frame	Change in turning time
S2	44–104 m	182°	-0.269s	-4.1 m	-3.6 m	-0.018s	-0.041s
S4	146–192 m	85°	+0.194s	+2.9 m	+2.8 m	+0.003s	-0.046s
S6	229–333 m	54°	-0.011s	-0.2 m	-0.6 m	+0.003s	+0.013s
S8	364–415 m	116°	-0.112s	-1.8 m	-1.5 m	-0.019s	+0.026s
S10	463–497 m	55°	-0.119s	-1.1 m	-1.2 m	-0.044s	+0.020s
S12	572–610 m	102°	+0.260s	+3.2 m	+3.2 m	+0.023s	+0.027s

All columns compare 8 September with 4 September's solo laps. Turning time is the time taken to rotate through the middle 60% of the corner's change of heading.

Measured A check that uses no positions agrees. Heading comes from the difference between consecutive fixes, so a constant offset cancels out of it. By that measure the time to turn S12 changed by +0.027s between 4 September's solo laps and 8 September, not +0.260s; and S2 by -0.041s, not -0.269s.

Correction What this withdraws from the 8 September page: that S2 was “where the session was actually won” (-0.269s as published, -0.018s in one frame); that S12 and S4 went backwards (+0.260s and +0.194s; +0.023s and +0.003s); that the session was “a change of line, not of speed”, since the path-length differences behind that were the slide; and the target carried into this session. In one frame no section moved by more than 0.044s between those two sessions in either direction. What stands: everything taken from the timing sheets, the two offs it located within its own recording, and its verdict on the 4 September experiment — S11 is a straight, reads -0.016s in one frame against -0.010s as published, and missed its −0.060s target either way.

Limit The slide is fitted from where each lap crosses the gates sideways, so a real change of line held at many gates at once would be partly absorbed into it. After sliding, the three sessions' typical lines sit within 0.6 m of each other at every gate, and the route through every section is the same length to within 1.1 m. That is the resolution of anything said about lines on this page.

6 · Where the time went, section by section

Each of 4 September's 13 sections in one frame: this session's 37 circuits against 8 September's 47, with 4 September's 21 solo circuits alongside. Down and blue is quicker.

Sec	Type	Distance	4 Sep solo	8 Sep	7 Oct	Change	Spread	Route Δ	Slowest Δkm/h	Fastest Δkm/h
S1	straight	0–44 m	2.537s	2.554s	2.587s	+0.032s	0.134s	+0.3 m	-0.3	-0.9
S2	corner	44–104 m	5.171s	5.153s	4.971s	-0.182s	0.119s	-0.3 m	+1.8	-0.5
S3	straight	104–146 m	2.840s	2.797s	2.825s	+0.028s	0.028s	-0.3 m	-0.1	-0.3
S4	corner	146–192 m	2.886s	2.889s	2.814s	-0.076s	0.057s	-0.1 m	+1.7	+0.0
S5	straight	192–229 m	2.293s	2.294s	2.293s	-0.001s	0.043s	+0.1 m	-1.9	+0.2
S6	corner	229–333 m	8.790s	8.793s	8.436s	-0.356s	0.135s	+0.3 m	+2.9	-1.7
S7	straight	333–364 m	2.288s	2.276s	2.281s	+0.005s	0.031s	-0.1 m	+1.1	-0.3
S8	corner	364–415 m	3.276s	3.257s	3.156s	-0.100s	0.064s	-0.5 m	+1.9	-0.5
S9	straight	415–463 m	2.929s	2.902s	2.928s	+0.026s	0.032s	+0.1 m	+1.0	-1.3
S10	corner	463–497 m	2.020s	1.976s	1.999s	+0.023s	0.048s	+0.2 m	+0.8	-1.0
S11	straight	497–572 m	4.357s	4.341s	4.252s	-0.089s	0.083s	-0.2 m	+1.9	-0.1
S12	corner	572–610 m	2.816s	2.839s	2.828s	-0.012s	0.081s	+0.2 m	+1.0	+2.2
S13	straight	610–641 m	2.135s	2.143s	2.172s	+0.029s	0.037s	+0.0 m	+0.0	-0.8

Medians. Route, slowest and fastest speed in the section are 7 Oct against 8 Sep.

Measured The gain is in 5 sections. S6 (-0.356s), S2 (-0.182s), S8 (-0.100s), S11 (-0.089s) and S4 (-0.076s): 0.804s between them. The other 8 are each within 0.032s of 8 September and sum to +0.131s. Added up, the section medians are 0.673s quicker, against 0.660s for the median lap on the sheet — two independent measurements of the same step.

Measured It is speed, on the same route. The path through every section is within 1.1 m of the other sessions', so none of this is a shorter line. What changed is how fast the kart was at the slowest point of the corners that gained: S6 +2.9 km/h, S8 +1.9 km/h, S2 +1.8 km/h and S4 +1.7 km/h. Top speed did not move: 67.8 km/h on the long straight against 67.9. S11, that straight, is 0.089s quicker without being any faster at its fastest point; the time is at its far end, where the slowest speed in the section is 56.2 km/h against 54.3. The short straights out of the corners are no quicker at all.

Supported interpretation More speed at the slowest point of 4 corners, on the same line, with no more top speed and nothing gained on the short straights, is what more grip looks like — or more readiness to use the grip there was. A cooler track, a different kart on different tyres and a driver a month fresher would each produce exactly this, and nothing here tells them apart. What it does not look like is a stronger engine: that would show in top speed and out of the corners, and it does not.

The same sections by stint, with 8 September for reference:

Laps	Circuits	Lap	S2	S4	S6	S8	S11	S12
8 September, all laps	47	44.280s	5.153s	2.889s	8.793s	3.257s	4.341s	2.839s
Opening	7	44.372s	5.049s	2.882s	8.622s	3.278s	4.381s	2.894s
After the first stop	8	43.536s	4.989s	2.807s	8.466s	3.143s	4.275s	2.843s
After the second stop	22	43.512s	4.960s	2.805s	8.395s	3.154s	4.238s	2.805s

Measured Within the session the step came after the first stop. The opening stint's median lap through the gates was 44.372s against 8 September's 44.280s, and most of its sections sit at 8 September's figures. Seven laps is a thin sample, and they are the first seven laps in an unfamiliar kart; but kart and afternoon are the same across these three rows, and the rows are not.

Three sections drawn: the two that gained most, and the corner the experiment was about. Orange is 8 September, blue is 7 October, and in the last panel the pale dashed line is 4 September. Every line has been slid into the common frame.

S6 · 229–333 m · corner 8.793s → 8.436s (-0.356s)

Medians, 8 September to 7 October: +0.3 m of route (103.2 m → 103.5 m), 51.9 → 49.8 km/h at the entry gate and 37.7 → 40.5 km/h at the slowest point. The largest gain on the lap, and none of it is route: the two lines lie on top of each other and the blue speed trace sits above the orange one through the slow part of the complex. Drawn: 8 September, 8.823s and 7 October, 8.471s — each the lap most typical of its session here in time, route and speed, not the fastest.

S2 · 44–104 m · corner 5.153s → 4.971s (-0.182s)

Medians, 8 September to 7 October: -0.3 m of route (59.0 m → 58.7 m), 57.3 → 57.2 km/h at the entry gate and 36.3 → 38.1 km/h at the slowest point. The hairpin. The same route again, taken faster at its slowest point. Drawn: 8 September, 5.167s and 7 October, 5.019s — each the lap most typical of its session here in time, route and speed, not the fastest.

S12 · 572–610 m · corner 2.839s → 2.828s (-0.012s)

Medians, 8 September to 7 October: +0.2 m of route (37.1 m → 37.3 m), 53.9 → 56.1 km/h at the entry gate and 44.6 → 45.6 km/h at the slowest point. The corner the last debrief asked you to win back. Three sessions, one route and one time; the lines that page drew apart here were two metres of GPS. Drawn: 4 September solo, 2.816s, 8 September, 2.801s and 7 October, 2.802s — each the lap most typical of its session here in time, route and speed, not the fastest.

7 · Where the remaining time is

Within this session alone: the median through each section against the fastest pass recorded there, in the common frame.

Measured The median-to-best gaps sum to 0.95s. The largest are S6 (0.172s), S2 (0.154s) and S12 (0.107s), holding 45% of the total. Adding the fastest section times gives a composite of 42.588s against an actual best circuit of 43.238s. Each “best” is the quickest of 37 passes that each carry a few hundredths of slide error in the corners, so both the gaps and the composite flatter what is there. They rank the sections; they do not promise the time.

Supported interpretation S12 is still a corner to look at, for a different reason than last time. Of the 6 corners, 4 got quicker; S10 and S12 did not. S12 reads 2.816s, 2.839s and 2.828s across the three sessions. You arrived at it faster — 56.1 km/h at its gate against 53.9 — and the time through it did not change. Within the session it did move, stint by stint: 2.894s → 2.843s → 2.805s. Its median-to-best gap (0.107s) is the 3rd largest on the lap.

8 · Findings, in priority order

The fastest session in the log, and not attributable. Best 43.239s, median 43.604s, 30 of 37 laps under 44 s. Kart, weather, a 29-day gap and two rests all changed with it, and the data cannot divide 0.660s between them.
The time is corner speed on an unchanged route. S6 (-0.356s), S2 (-0.182s), S8 (-0.100s), S11 (-0.089s) and S4 (-0.076s). Slowest speeds in those corners are up; top speed (67.8 km/h) and the short straights are not. That reads as grip, from whichever source, rather than engine.
The standing experiment is void. S12 reads 3.117s as the target was written and 2.828s once the laps share a frame, against 2.816s and 2.839s before. It never lost the time, so it could not win it back.
The last debrief's section findings were the GPS, not the driving. A 1.9 m offset between two days' traces produced its S2 gain, its S12 and S4 losses and its “change of line”. Between 4 September and 8 September no section actually moved by more than 0.044s. Every cross-session section figure from now on is taken in one frame, and anything under 0.05s is treated as nothing.
The step came after the first stop. The opening stint averaged 44.589s; the next two 43.582s and 43.549s. Same kart, same afternoon — and also the first seven laps in an unfamiliar kart, so this says when the pace arrived, not why.
The logger's lap line did not matter to any of this. All 37 running laps were timed through the gates, to within 86 ms of the sheet, whichever line the logger had picked. The stop made to restart it cost about 40s and a lap.
RPM has now recorded zero for three sessions. With the kart in question it is the one channel that would separate engine from everything else.

9 · One thing to test next session

Provisional, pending your input: choose the kart on purpose, and let the kart be the experiment. What you were working on today is not known here, so no driving change is prescribed.

Why this: 0.660s arrived with four changes at once, and the only one that can be set deliberately next time is the kart. Until one of them is held or reversed, no driving target can be scored: there is no baseline that is only driving.
Either way works: Kart 19 again tests the kart directly: its figures from 8 September are S6 8.793s and S2 5.153s, against 8.436s and 4.971s today. Kart 28 again holds the kart and lets the weather and the month vary instead; today's medians become the baseline. Taking whichever kart is free tests nothing.
Measure: Median lap, and the S6 and S2 medians, over at least ten laps, through the same gates and in one frame.
Falsifiable: In kart 19: if S6 and S2 return to 8 September's figures, the gain belonged to kart 28 or to the day; if they hold near today's, it did not belong to the kart. Differences under 0.05s decide nothing.
The logger: Leave it alone once rolling. Whichever line it counts, every lap is timed through the gates, so there is nothing to gain by stopping to reset it.
Still standing: S12, which has not moved in three sessions — and a question for you first: what are you doing differently there now that you arrive at 56.1 km/h rather than 53.9?

10 · How this was put together

Jorge Segers, Analysis Techniques for Racecar Data Acquisition, 2nd ed. (SAE International), chapter 14, “Analyzing the Driver” — contents.
Your “theoretical best” lap is a lie — but still useful — the case for treating a composite as a diagnostic.
Gates: 4 September's 13 section starts, unchanged from the last debrief. A crossing counts only within 8 m of the gate point, and a circuit only if it crossed every gate once, in order, in 40–49 s.
Frame: each lap translated east/north by least squares on its sideways position at the gates, against the median of 4 September's solo laps. The unslid times are kept only to score the target as it was written.