Driver debrief

LisboaKart, Pontinha · layout Lisboa V3 · kart 19 · 4 September 2026, 18:26–19:04 · 48 timed laps · first session with GPS on the kart

This was a personal best by 0.852 s and the largest single step in six visits, according to the karting log — that comparison comes from the log, not from this recording. The useful question is no longer whether it was quick — it was — but where the remaining time sits, and that is a question only position data can answer. Here is what the logger supports, what it merely suggests, and the one thing worth testing next time.

Best lap

43.873

sheet lap 22; a personal best by 0.852 s per the log

Fastest 5 avg

43.980

not one lucky lap

Solo phase

44.399

± 0.300s over 21 laps

Section benchmark

1.20

sum of 13 median-to-best gaps, not a lap

1 · What this debrief can and cannot say

A debrief is only worth as much as the channels behind it, so this comes first.

Measured Two independent timing systems agree. The printed Race Control sheet and the logger were matched lap by lap: logger lap 2 is sheet lap 1, and across all 48 laps the mean difference is 30 ms (largest 254 ms, on the out-lap). The logger's two extra laps are a 7.7s fragment from leaving the pits and a 52.7s in-lap, neither of which the sheet counts. The session best reads 43.873s on the sheet and 43.848s on the logger — the same lap, 25 ms apart, measured by a transponder loop and by GPS respectively.

Measured All recovered GPS fixes are retained. There are 55,549 fixes at nominal 25 Hz. Satellites used ranged 8–16, most often 14, and the receiver's own accuracy estimate had a median of 69 cm (58–110 cm). That is a precision estimate, not a bound on absolute error. GPS bias and imagery registration remain separate uncertainties. All raw GPS record checksums pass and raw GPS timestamps are monotonic. The apparent GPS time jumps arise in the parser's upper-bit correction, not in those raw timestamp fields; investigation of the parser's anchor inputs continues. Note that 31,130 of those fixes are ones the standard open-source parser silently discards; the session report explains why and how that was corrected.

Not available There is no pedal or engine data. The RPM channel recorded zero for the entire session — the pickup was not reading — and this logger has no brake or throttle channels at all. Nothing in this debrief can therefore describe when you braked, how you released the brake, or when you picked up the throttle. Where those would be the natural explanation, it is named as a question rather than a finding. Worth fixing the RPM pickup before the next session: it is the one channel that would let engine behaviour be separated from driving.

Method Sections, not corner numbers. The circuit's own corner numbering is not known here, so the lap is cut into 13 analysis sections derived from its curvature, fixed at the same distances for every lap. They are labelled S1–S13 and no official numbering is implied. Receiver accuracy estimates do not establish a hard cutoff for true line differences. These normalized-distance sections are not surveyed timing gates.

2 · The session as it ran

Kart 19, taken back deliberately after one session in 21. Hot and dry — 29–30 °C after eight rainless days — under full cloud. Four other karts on track for the first 20 minutes, which the sheet places at laps 1–27; alone from lap 28 to the end. You reported persistent grip problems and two bad offs, both in the traffic phase. No single change was under test, so no row below isolates driving from the kart, the track state or the extra practice.

Every timed lap. Orange is the traffic phase, blue the solo phase; the shaded area marks running alone. Two laps run off the top of the scale — the 51.9s out-lap and lap 2 at 51.0s.

Phase	Laps	Count	Mean	Median	Std dev	Best
Traffic, four other karts	1–27	27	45.531s	44.873s	1.933s	43.873s
Alone on track	28–48	21	44.399s	44.336s	0.300s	43.940s

Measured Solo-lap repeatability was tight. Alone on track you ran 21 laps at 44.399s with a standard deviation of 0.300s. The karting log's previous tightest is 28 August at 0.288s over 40 laps, so this is comparable repeatability roughly a second a lap faster — a cross-session comparison, from the log rather than from this recording. The whole-session spread also includes the slower early laps and changing conditions; it cannot isolate the effects of traffic, offs, or driving consistency.

3 · Where the time actually is

Each section timed on every clean lap, then the solo median compared with the fastest time you recorded there all session. The gap compares different laps and entry conditions; it is a section benchmark, not proof that the difference can be recovered repeatedly.

Sec	Type	Distance	Your fastest	Set on	Solo median	Spread	Available	Traffic − solo median
S1	straight	0–44 m	2.487s	lap 45	2.544s	0.023s	0.056s	+0.024s
S2	corner	44–104 m	4.986s	lap 24	5.130s	0.065s	0.143s	+0.024s
S3	straight	104–146 m	2.800s	lap 22	2.865s	0.039s	0.065s	-0.014s
S4	corner	146–192 m	2.799s	lap 20	2.878s	0.048s	0.079s	+0.050s
S5	straight	192–229 m	2.200s	lap 22	2.265s	0.034s	0.065s	+0.019s
S6	corner	229–333 m	8.600s	lap 22	8.848s	0.125s	0.248s	+0.005s
S7	straight	333–364 m	2.228s	lap 45	2.292s	0.037s	0.063s	+0.027s
S8	corner	364–415 m	3.315s	lap 15	3.375s	0.046s	0.060s	+0.025s
S9	straight	415–463 m	2.842s	lap 45	2.916s	0.031s	0.074s	+0.030s
S10	corner	463–497 m	1.881s	lap 19	1.961s	0.031s	0.080s	+0.004s
S11	straight	497–572 m	4.166s	lap 47	4.309s	0.068s	0.143s	+0.056s
S12	corner	572–610 m	2.763s	lap 16	2.847s	0.034s	0.084s	-0.003s
S13	straight	610–641 m	2.050s	lap 32	2.084s	0.025s	0.035s	+0.044s

Measured The section median-to-best gaps sum to 1.20s across 13 sections. This is not an achievable whole-lap gain: the section minima and medians are not one continuous lap. The top three — S6 (0.248s), S2 (0.143s) and S11 (0.143s) — hold 45% of it. S6 is also the section with the widest lap-to-lap spread (0.125s), which makes it worth examining; traffic, conditions and driving can all contribute.

Measured Traffic-phase section medians sum to 0.29s more than solo-phase medians. This is an association, not an isolated traffic penalty: practice and conditions also changed with time. Summed section medians and whole-lap means answer different questions.

4 · The divergence you spotted at 34 seconds

You noticed the speed traces separating from about 34 s in the best-versus-median comparison. Elapsed time cannot locate that on the circuit — two laps at the same second are not in the same place — so both laps were lined up by distance instead. As it happens they were only 6 m apart there, so the reading held up.

Measured It is section S11, the 75 m straight from 497 to 572 m. Sheet lap 22 took 4.200s through it; sheet lap 26 took 4.495s — a 0.295s difference on one straight. The slower lap entered it at 56.7 km/h against 61.4 km/h, and 10 m earlier, still inside the kink, it was doing 61.3 km/h against 62.4 km/h.

Supported interpretation The slower lap arrived already 4.6 km/h down. That supports testing the preceding exit. Speed can also change on the straight through throttle use, traffic, drag or engine behaviour; without working RPM or pedal channels, the cause of the remaining difference is not established.

Question for you Do you recognise that kink as a place where you sometimes get the kart straightened late? The data can show the speed you carried out of it, but not why — that would need brake and throttle traces, which this logger does not record. See it on the circuit in the driving-line replay, which highlights this exact stretch.

5 · The first corners

You asked about the opening corners specifically. S2 is the first real corner, the one that takes you from 44 to 104 m and down to about 34 km/h; S4 is the fast kink that follows.

Measured Your best lap was not your best first corner. Through S2 the best lap took 5.160s, where your fastest was 4.986s on sheet lap 24 — 0.174s quicker. The solo median is 5.130s, so the median-to-best gap here is 0.143s. S4 has a smaller gap: 0.079s, with spread 0.048s.

Supported interpretation The opening corners are worth having but they are not where the session is being decided — together they hold about 0.22s against 0.25s in S6 alone. If you want the biggest return for one change, it is not here.

6 · The two offs

Measured These metrics cannot confirm or rule out the reported offs. Among laps selected as clean (40–49 s), the largest nearest-point distance from the best lap is 5.0 m on sheet lap 16. This is not distance from the asphalt edge. The lowest speed across that selected pool is 26.7 km/h. Neither metric proves whether a kart briefly left the track; the filter also excludes slower laps.

Sheet lap	Lap time	Furthest from the best lap's line	Minimum speed
16	47.400s	5.0 m	26.7 km/h
8	47.151s	4.3 m	33.3 km/h
6	46.931s	4.0 m	37.4 km/h

Question for you The table ranks the three largest line deviations within the filtered pool; it is not a shortlist of proven offs. A recalled lap, timestamp or video would help locate the events. The current data leave both reported offs unlocated.

7 · What the session says you are capable of

Adding your fastest time in each section gives a composite lap of 43.118s. That is a benchmark, not a target — the pieces come from different laps that entered and left each section at different speeds — and the number moves with how finely the lap is cut. The theoretical-best page shows every piece on the circuit, which lap set it, where the joins do not meet, and how sensitive the total is to the method. Its difference from the logger best is 0.730s for this division. The linked sensitivity table uses the same calculations; no particular gain is guaranteed. Boundary sampling and different line lengths also affect the result.

8 · Findings, in priority order

The pace step is real and it is not one lap. Best 43.873s, fastest five 43.980s, and 35 of 48 laps under 45 s against four across the five previous sessions combined (log figures). Whatever you changed, it held for a whole stint.
S6 has the largest median-to-best gap: 0.248s. It is also your least repeatable section (0.125s spread), so the largest gap and the widest spread fall in the same place. That is the useful part — you have already driven S6 0.248s quicker than your own median under potentially different conditions; repeatability is a useful next test.
The straights are telling you about the corners before them. S11 and S9 have median-to-best gaps summing to 0.22s. Entry speed is one testable explanation; losses during the straight remain possible.
Traffic-phase section medians were 0.29s slower in sum, and your solo mean was 1.13s better. The target for the next session is not a faster best lap: it is running the solo pace while there is company.

9 · One thing to test next session

Carry more speed out of the kink at 463–497 m onto the long straight. One change, one section, measurable without any new equipment.

Why this: It is the corner feeding the longest straight on the lap, it is where your own two laps diverged by 0.295s, and the straight after it has a 0.143s median-to-best gap. This experiment tests entry speed without assuming it explains the whole gap.
What to change: Nothing else. Same kart (19), same tyres, and leave every other corner alone so the comparison means something.
Measure: Speed at 497 m and the S11 section time, on solo laps only. This session: 62.4 km/h at best, solo median S11 4.309s.
Target: Solo-lap median S11 of 4.249s or better over at least eight laps. Over a 75 m section that is about 0.9 km/h more average speed through S11; how much exit speed it takes is not derivable here. A single quick lap does not settle it.
Falsifiable: If the S11 median does not move, the hypothesis was wrong and the time is somewhere else — most likely S6, which is the next experiment.

10 · How this was put together

Debrief practice, and the caution about composite laps, follow published practitioner sources rather than invention:

Jorge Segers, Analysis Techniques for Racecar Data Acquisition, 2nd ed. (SAE International). Chapter 14, “Analyzing the Driver”, is the standard reference for separating driver performance from vehicle performance in logged data — contents.
Driver61, Prioritising Circuit Corners — why the corners leading onto the longest straights are worth the most, which is the reasoning behind the experiment above.
How optimal lap time is calculated — micro-sector summing, and why teams constrain it before believing it.
Your “theoretical best” lap is a lie — but still useful, and Occam's Racer on theoretical best laps — the case for treating the composite as a diagnostic.
A former F1 data engineer on telemetry practice and how race engineers turn data into pace — on debrief structure and on which channels carry driver information.