Barber — The Weekend in Review

Aug 15–16, 2026 · G29 Z4 M40i · Bridgestone RE-71RZ (2026 set) · HPDE, instructing Taylor · Final wrap-up — both days, all fifteen corners

DRAFT — not published. Live at this URL for review only: noindex, nofollow, not linked from anywhere public, draft:true on the event page. Nothing here is shareable until Bill says so.

Weekend PB

1:42.70

Saturday, lap 22, run 2. Old Barber mark 1:43.910 (April, Z4) — beaten by 1.21 s. May's M440i: 1:45.99.

Sunday best

1:43.39

Lap 13, first run, 77 °F. 0.69 s off the PB — and still 0.52 s under the old mark.

Weekend theoretical

1:40.70

Both days' clean laps stitched: 2.00 s under the PB. Fifteen laps contribute; Sunday owns 6 of the 15 corners.

Laps · runs

73 · 8

51 Saturday + 22 Sunday timed. 50 clean (35 / 15). Tires 13 heat cycles in, ~21 out.

The weekend in one paragraph

Two days at Barber in the Z4 produced a 1:42.70 on Saturday afternoon — a 1.21-second personal best, and a lap that still had a visible mistake in it — then a cool, short Sunday morning that drove better and lapped slower. The one technique rebuild we planned, the brake release at Charlotte's Web, took: it raised the ceiling by 0.20 s and, more importantly, it raised the floor — Sunday's six worst attempts still braked to −0.83 g where Saturday's six worst managed only −0.60 g. The Carousel, the other headline assignment, did not take at all; it went backwards and stays on the list. What kept the 1:41 away was not pace but scatter: the sum of per-corner spread grew from 3.68 s Saturday to 4.81 s Sunday across 15 clean laps instead of 41. Stitch both days' best corners together and the lap is a 1:40.70, from fifteen different laps — which is the whole finding. The car, the tires and the pressures were never the limit: 34F/32R hot held for two days, there was no falloff at 15+ heat cycles, and the fuel rule (fill before every session) made Saturday's drivetrain-malfunction limp a one-time event. On the same Saturday morning a race-prepped E36 M3 went 1:42.163 and a race-prepped Mini went 1:47.340 — the Z4 finished second of three by 0.537 s, winning 1.96 s wherever the road was straight and giving back 2.50 s wherever it had to turn.

What we set out to do vs what happened

The Sunday plan said	Result	Evidence
Minimum speed, not later braking — stop over-slowing entries	Partly	Weekend-best minimum rose in 8 of 15 corners; but Bill's minimum is still below the M3's in 9 of 15, and every one of those nine is a corner he loses time in.
Rebuild T5 Charlotte's Web — brake later and to a real threshold, then bleed the release through turn-in	Yes — the one clear win	Best T5 of the weekend came Sunday (11.266 s, −0.200 vs Saturday). The floor moved more than the ceiling: worst-six peak decel −0.60 g → −0.83 g. Throttle pickup moved from 0.5 m before the apex to 5.5 m past it.
Attack the Carousel — one sustained arc, no mid-corner bleed	No	Best window went backwards (6.570 → 6.647 s), median +0.092 s. Still 0.380 s of self-spread. It was item 1 of a five-item list and it lost the competition for attention.
Go out early, before the heat	Yes	Sunday R1 at 77 °F produced the day's best lap and three laps under 1:44. Done and packed by 10:09.
Fuel before every session	Yes — zero recurrence	Saturday's laps 19–21 were power-capped (top speed 103–105 vs 114–117 MPH). Sunday: no starve signature on any lap, and Sunday R2 was the fastest of the weekend over the T4 crest at 112.5 MPH on the lightest tank.
Hold 34F / 32R hot; do not change pressures	Yes	Held both days. No progressive within-run mid-corner speed loss in the fast rights on either day.
Expect tire falloff past 15 heat cycles	Wrong, in the good direction	Saturday's 16:01 run was the day's second-fastest on the oldest rubber; Sunday's peak lateral g (1.32) was the weekend's highest. No ceiling visible yet.
Convert to a 1:41.x	No	Repeatability, not pace: self-spread grew 3.68 → 4.81 s. Every corner's weekend best now beats the May M440i template, so there is no corner left where the ceiling is the problem.

How the two days actually ran

Day	Run	Local	Laps	Best	Amb °F	What happened
Sat	R1	08:00–08:44	14	1:44.79	77	30/29 psi cold grew to 40/39 hot → loose. Bled to 32/30 hot; the run's best lap came immediately after.
Sat	R2	09:42–10:18	12	1:42.70	86	The PB, on lap 22. Also the fuel-starvation limp — laps 19–21 came back power-capped and latched until an ignition cycle.
Sat	R3	11:27–11:41	8	1:44.59	92	Heat arriving.
Sat	R4	13:18–13:45	8	1:45.13	95	Hottest, busiest. Slowest run of the weekend.
Sat	R5	14:44–14:52	4	1:45.81	95	Short run, traffic.
Sat	R6	16:01–16:08	4	1:43.34	95	Second-fastest run of the weekend — end of the hottest day, oldest rubber, no fade. The tire evidence in one line.
Sun	R1	07:58–08:39	16	1:43.39	77	The "go early" plan, executed. Coolest track of the weekend; three laps under 1:44 in this run alone.
Sun	R2	09:51–10:09	6	1:43.41	88	Last run of the weekend. Lap 23's 1:43.41 was the final lap logged — the logger stopped at 3158 m, so it has a lap time but no complete corner trace.

Weather, from the data. Ambient came from silver.lap.source_extras.ambient_temp_c, per run: Saturday 25.1 → 34.7 °C (77 → 95 °F), climbing all day; Sunday 24.8 → 31.1 °C (77 → 88 °F), and the weekend ended before the heat did. Dry both days.

The Sunday shape you should notice. Sunday was two runs and out by 10:09. That was the right call on tires and heat, and it is also why the 1:41.x never appeared: the plan asked for six specific corners to line up inside one lap, and 15 clean laps is not many attempts at a six-part coincidence. Saturday needed 41 clean-ish laps to produce one 1:42.7. Typical clean lap was 1:46.48 Saturday and 1:46.89 Sunday — statistically the same lap. All the difference between the days lives in the tails.

The three-way race, Saturday morning

1. Mike M. — E36 M3 (race) — 1:42.163 · 2. Bill — Z4 M40i — 1:42.70 (+0.537) · 3. Rich S. — Mini (race) — 1:47.340 (+4.64)

Three laps inside about an hour on the same asphalt: Mike at 09:36, Bill at 10:07, Rich at 10:37. Distance-aligned, the M3 built a full second over Bill in the first third of the lap, the Z4 took almost all of it back through the fast middle, got its nose ahead three times around T9–T11, and lost it for good at T12. Grouped: the Z4 wins 1.96 s wherever the road is straight or nearly straight and gives back 2.50 s wherever it has to actually turn. Bill's minimum speed is lower than the M3's in 9 of 15 corners, and every one of those nine is a corner he loses time in — identical entry speeds, very different minimums. That is not the car and it is not the tires. The Mini is short of engine everywhere (102.3 MPH down the pit straight to the Z4's 116.8) and still beats both cars through T7a/T7b: 5.70 s to Bill's 6.30 and Mike's 6.06.

Two caveats that travel with these numbers. Mike's lap time is 1:42.163 from the file's own beacon markers, not the 1:42.100 the adapter stored (TDAI-143). And peak lateral g and brake points are not compared across the three cars — different loggers at different rates, and Mike's GPS-derived g channel reports a non-physical 2.12 g peak. Only where the minimum sits inside a corner transfers, because that is a driving choice.

Corner by corner — all fifteen

All speeds MPH. JTI corner labels throughout. Numbers pool both days' clean laps (35 Saturday + 15 Sunday). Per card: wk best / wk med = the fastest and the median window time across the weekend; spread = median minus best, i.e. time already demonstrated in the same car on the same tires; best min = the highest minimum speed reached all weekend, with the weekend median beside it.

How the cards are ranked. Remaining time per corner = 1.00 × own spread (fully earned — same car, same tires, same weekend) + 0.50 × gap to Mike's M3 on the PB lap (half credit: different car, so only the minimum-speed placement transfers) + 0.25 × gap to the May M440i template (quarter credit: the template car is slower, so being behind it is a warning flag, not a target). That third term came out zero for all fifteen corners — every corner's weekend best now beats the May median, including T5, which was the one exception on Saturday. So the ranking is effectively own spread plus half the M3 gap, and the May template has stopped constraining anything.

T2/T3Right · 415 m · rank 1 · 0.87 s

Best window of the weekend, worst scatter of the weekend

ON-CAMBER +9.1% COMPRESSION −8.4% grade NEW BEST SUNDAY −0.125 s SPREAD 0.490 s

wk best

10.953 s

wk med

11.442

best min

55.6 med 52.4

vs May

−0.64 s

vs M3

+0.75 s

What worked: the outside/later entry line — it produced the fastest T2/T3 of the weekend, on Sunday.

Next time at Barber: pick one fixed visual turn-in marker and use the same one every single lap. This is the biggest corner loss of the race (+0.752 s to the M3, 140% of the whole race margin) and the widest self-spread on the circuit. Mike is 5–9 MPH faster from entry to the exit of the second apex and puts his minimum at 36% of the window against Bill's 29% — slowest point too early, deficit carried through the whole double-apex. Since the two line families are only worth ~0.09 s apart, most of the 0.49 s of scatter is turn-in timing, not line choice. Sat→Sun: ceiling improved −0.125 s, median regressed +0.141 s.

T5Left · 1055 m · Charlotte's Web · rank 2 · 0.62 s

The one rebuild — and the one that took

OFF-CAMBER −1.76% COMPRESSION −5.3% grade CUE TOOK −0.200 s the lap's only threshold-brake zone

wk best

11.266 s

wk med

11.795

best min

44.6 med 40.2

vs May

−0.14 s

vs M3

+0.17 s

What worked: braking later and to a genuine threshold, then bleeding the release through turn-in — and it shows up in every attempt, not just the good ones.

Next time at Barber: keep this cue word for word, and aim the minimum at 44–45 MPH instead of 40. The floor moved more than the ceiling, which is the better outcome: Saturday's six worst T5s peaked at only −0.60 g, Sunday's six worst still reached −0.83 g, with a Sunday median of −0.90 g and one lap at −1.15 g. Throttle timing corrected too — Sunday's best six pick up 5.5 m past the apex where Saturday's best six were 0.5 m before it, so the "early throttle" half of the old failure mode is gone. T5 also stopped owing time to the May template. The remaining prize is size: Bill is 14 MPH slower than the M3 at the slowest point, the largest instantaneous deficit anywhere on the lap, and he spends the next 230 m of exit clawing it back on turbo torque over a surface that is off-camber and falling away. "Patience" stays retired — the patient attempts were the slow ones. Sat→Sun: ceiling −0.200 s, median −0.060 s. The only corner that improved on both.

T1Left · 215 m · rank 3 · 0.57 s

The narrow ask, delivered

ON-CAMBER +6.0% COMPRESSION −3.9% NEW BEST −0.115 s

wk best

4.829 s

wk med

5.208

best min

78.0 med 71.2

vs May

−0.65 s

vs M3

+0.38 s

What worked: one specific, one-sentence assignment — and it was the cleanest win of the weekend.

Next time at Barber: carry the minimum. 78 MPH is proven; the median is 71.2, and the M3 does 76.4 on its way past. V3's ask was narrow: the PB lap gave 0.16 s back at T1 while lap 50 did it properly, so put lap 50's T1 onto lap 22's lap. Sunday's best T1 (4.829 s) is quicker than either Saturday reference and the best minimum rose to 78.0. This is the argument for giving next weekend one corner at a time. Sat→Sun: ceiling −0.115 s, median +0.080 s.

T12Right · 2675 m · rank 4 · 0.51 s

The corner that ended the race — and it regressed

NEUTRAL CAMBER +0.2% +8.2% EXIT GRADE REGRESSED +0.088 s

wk best

5.146 s

wk med

5.438

best min

66.5 med 62.5

vs May

−0.05 s

vs M3

+0.44 s

What worked: nothing here — this is the one corner where both days and both comparators agree Bill is leaving time and nobody asked him to fix it.

Next time at Barber: stop over-slowing the compression. The M3 carries 72.9 MPH minimum to Bill's 64.1 — 8.8 MPH, the largest minimum-speed deficit on the lap. T12 is a compression with an 8.2% exit grade, so over-slowing costs twice: once in the corner, then again on the climb out. In the race the gap was level at T12 entry and +0.80 s by the Carousel. Note it is only barely ahead of the May template (−0.05 s) — nearly the weakest margin of the fifteen. Sat→Sun: ceiling +0.088 s, median +0.190 s. Worse on both.

T13/T14Right · 2875 m · the Carousel · rank 5 · 0.44 s

The assignment that did not land

ON-CAMBER +3.3% WENT BACKWARDS +0.077 s SPREAD 0.380 s

wk best

6.570 s

wk med

6.950

best min

72.8 med 65.9

vs May

−0.37 s

vs M3

+0.11 s

What worked: the commitment — no brake on any lap of either day, at 1.18–1.22 g. What did not: the apex accuracy, and the mid-corner bleed.

Next time at Barber: one sustained 82–83 MPH arc, no mid-corner bleed — and make it the whole first session, the sole item on the card. Mike holds 82–84 MPH through the entire arc; Bill sags to 78–80 and stays there. Combine the Carousel with T14a and the M3's margin through the complex is +0.271 s. This was rank 1 on Saturday's list and it still has the largest single-day spread on the circuit (0.385 s Sunday); it drops to rank 5 here only because pooling both days moved T2/T3, T5 and T1 above it. It is the corner the coaching corpus calls the trickiest on the property, it is the one place where the surface is helping (+3.3% on-camber, a loan Bill is not taking), and a blind apex is not something a paragraph fixes — it was item 1 of a five-item list in a 16-lap run and it lost the competition for attention. Running wide here is also what then wrecks T14a and T15. Sat→Sun: ceiling +0.077 s, median +0.092 s.

T7a/T7bRight · 1745 m · rank 6 · 0.43 s

Where a Mini with no engine beats both fast cars

ON-CAMBER +8.9% CREST −6.8% grade MEDIAN +0.147 s

wk best

6.213 s

wk med

6.527

best min

45.9 med 41.9

vs May

−0.25 s

vs M3

+0.24 s

What worked: nothing changed here — ceiling flat, median 0.147 s worse.

Next time at Barber: this is a minimum-speed corner, not a braking corner — 45.9 MPH is proven and 41.9 is typical, so the whole job is repeating the good one. The clearest cross-car evidence on the lap: Rich's Mini goes through in 5.70 s against Bill's 6.30 and Mike's 6.06, with the least power of the three. The M3 carries 50.2 MPH minimum to Bill's 43.2. Steep on-camber support (+8.9%) over a falling crest — the surface will take more than Bill is asking of it. Sat→Sun: ceiling +0.034 s, median +0.147 s.

T10Right · 2190 m · rank 7 · 0.35 s

A power section with a repeatability problem

ON-CAMBER +1.1% FLAT MIN DOWN 3.7 MPH

wk best

3.755 s

wk med

4.047

best min

90.0 med 82.4

vs May

−0.11 s

vs M3

+0.12 s

What worked: no braking on any clean lap of either day — the commitment is right.

Next time at Barber: hold it flat and stop letting the minimum sag — 90.0 MPH happened, 82.4 is normal. Only 0.2 MPH above the May template on median minimum, which is the weakest template margin of the fifteen and a hint that this corner is being coasted rather than driven. The M3 is 2.7 MPH better at the minimum. Sat→Sun: ceiling +0.094 s, median +0.156 s — and the best minimum dropped 3.7 MPH, the largest day-to-day regression in minimum speed on the lap.

T11Left · 2560 m · rank 8 · 0.32 s

Nearly solved — and the second corner whose median improved

ON-CAMBER +2.8% FLAT MEDIAN −0.040 s

wk best

2.690 s

wk med

2.983

best min

92.5 med 82.4

vs May

−0.56 s

vs M3

+0.04 s

What worked: dead level with the M3 here (+0.04 s, 84.6 vs 84.7 MPH minimum) and 11.1 MPH above the May template — the biggest template margin on the lap.

Next time at Barber: a 92.5 MPH minimum is on the table — treat T11 as flat and let the on-camber support do the work. One of only two corners whose median got faster on Sunday. The 0.29 s of spread is worth having but this is close to finished; do not spend a session on it. Sat→Sun: ceiling +0.035 s, median −0.040 s.

T7Right · 1630 m · rank 9 · 0.31 s

Level with the M3, spread with himself

ON-CAMBER +2.3% CREST LEVEL WITH M3

wk best

3.947 s

wk med

4.258

best min

74.4 med 65.8

vs May

−0.34 s

vs M3

−0.01 s

What worked: 74.4 MPH minimum against the M3's 71.0 and the May template's 59.6 — the pace here is genuinely good.

Next time at Barber: pure repetition — 0.31 s of spread on a corner with no comparator gap left means the only instruction is "do the good one again." Ten of 15 Sunday laps show a real brake event here and five do not, which is most of the scatter: decide whether T7 is a brake corner or a lift corner and then do the same thing every lap. Sat→Sun: ceiling +0.046 s, median +0.065 s. The best minimum fell 2.9 MPH.

T4Right · 900 m · the crest · rank 10 · 0.31 s

Flat, proven flat twice — and the fuel evidence lives here

OFF-CAMBER −0.5% NEW BEST −0.028 s NO STARVE SUNDAY

wk best

3.995 s

wk med

4.305

best min

85.7 med 72.4

vs May

−0.33 s

vs M3

−0.37 s

What worked: full commitment over the crest, and the fuel rule — Sunday R2 was the fastest of the weekend here at 112.5 MPH on the lightest tank of the weekend.

Next time at Barber: flat, eyes up, and nothing else — then bank the 0.31 s of spread by repeating it. One of the Z4's best corners: 0.369 s ahead of the M3, 80.1 MPH minimum to its 74.6, and the weekend's best window came Sunday. The 2026-08-15 drivetrain-malfunction limp was not a T4 problem — the three capped laps show a whole-lap power deficit (top speed 103–105 vs 114–117 MPH), i.e. a car condition, not a corner. Sat→Sun: ceiling −0.028 s, median +0.052 s; best minimum up 2.2 MPH.

T15Left · 3210 m · rank 11 · 0.30 s

The last corner — and the run to the flag is free money

ON-CAMBER +6.1% −5.0% GRADE MIN DOWN 2.7 MPH

wk best

5.580 s

wk med

5.843

best min

66.9 med 62.6

vs May

−0.11 s

vs M3

+0.08 s

What worked: the exit — the Z4 takes 0.259 s off the M3 between T15 and the line at 112.6 MPH to its 104.6.

Next time at Barber: get the exit right and the straight pays you twice — 66.9 MPH minimum is proven against a 62.6 median. The best T15 window of the entire weekend (5.580 s) belongs to Saturday lap 5, an early 1:45.63 — the single best illustration that this is a repeatability problem, not a pace problem. Strong on-camber support (+6.1%) on a falling grade. Sat→Sun: ceiling +0.095 s, median +0.056 s.

T14aRight · 3050 m · rank 12 · 0.27 s

Where the Carousel's real minimum actually lives

ON-CAMBER +7.5% CREST NEW BEST −0.022 s

wk best

5.062 s

wk med

5.256

best min

54.7 med 51.5

vs May

−0.12 s

vs M3

+0.16 s

What worked: highest peak lateral g on the lap (1.22 g Sunday) and the weekend's best window came Sunday.

Next time at Barber: treat T14a and the Carousel as one assignment — the Carousel window ends mid-deceleration, so the complex's true minimum is here. Bill's best minimum (54.7) exactly equals the M3's, but the median is 51.5 and Mike is still +0.157 s through the window; combined with the Carousel his margin over the complex is +0.271 s. Steep on-camber support at +7.5%. Sat→Sun: ceiling −0.022 s, median +0.044 s.

T9Left · 2040 m · rank 13 · 0.24 s

A Z4 corner — leave it alone

ON-CAMBER +2.2% FLAT AHEAD OF M3

wk best

4.308 s

wk med

4.549

best min

91.2 med 87.0

vs May

−0.10 s

vs M3

−0.22 s

What worked: no lift, no brake, on any clean lap of either day — 91.2 MPH minimum against the M3's 84.7.

Next time at Barber: no instruction. Full throttle, look through to T10, and spend the attention elsewhere. 0.24 s of spread with no comparator gap; the only reason it is not last on the list is honest scatter. Sat→Sun: ceiling flat (+0.002 s), median +0.095 s.

T6Right · 1400 m · rank 14 · 0.10 s

Barely a corner — and the Z4's biggest single win

NEUTRAL CAMBER FLAT +0.50 s ON THE M3

wk best

4.701 s

wk med

4.806

best min

96.1 med 94.9

vs May

−0.06 s

vs M3

−0.50 s

What worked: it is a flat-out kink (0.21–0.24 g) and Bill treats it as one — 96.1 MPH to the M3's 88.1.

Next time at Barber: nothing. This is straight-line physics and the smallest spread but one (0.104 s). The Z4's largest single-corner gain over the M3 — and there is no driving lesson in it for anybody. Sat→Sun: ceiling −0.004 s, median +0.052 s.

T8Left · 1880 m · track-out · rank 15 · 0.09 s

The lidar caution that the car does not agree with

OFF-CAMBER −1.27% FLAT CALMEST CORNER ON THE LAP

wk best

2.770 s

wk med

2.859

best min

73.7 med 69.9

vs May

−0.06 s

vs M3

−0.20 s

What worked: full commitment. Zero of 35 Saturday clean laps and zero of 15 Sunday laps show any deceleration inside the window.

Next time at Barber: full throttle, no lift, look through to T9 — and ignore the camber badge. The lidar's alarming −6.4% is camber_min at a single 5 m station; the corner's apex reads −1.27% and its mean −1.75%. What the car does: +15.2 MPH gained through the section (71.6 → 86.8 MPH), peak lateral g identical to the May template, and the lowest steering-correction index of the whole descent group. The apparent mid-window dip in lateral g is the sign change from T7b (right) into T8 (left) — a direction change, not a loss of grip. Smallest spread of the fifteen. Sat→Sun: ceiling +0.041 s, median +0.075 s.

Next time at Barber — the list

Five corners, 2.80 s, in this order

T2/T3 — 0.87 s. "One marker. Same one every lap." Fix the turn-in timing, not the line. Biggest scatter on the circuit and the biggest single loss to the M3.
T5 Charlotte's Web — 0.62 s. "Brake later, to a real threshold, then bleed the release through turn-in. Minimum 44–45, not 40." Keep the cue exactly as written — it is the only technique change that measurably moved a corner.
T1 — 0.57 s. "Carry the minimum — 78 is proven." The narrow-ask corner. Do not re-litigate the brake point.
T12 — 0.51 s. "Stop over-slowing the compression — the M3 carries nine more MPH." Costs twice, in the corner and on the 8.2% climb out.
T13/T14 Carousel — 0.44 s. "One sustained 82–83 MPH arc. No bleed." First session of the weekend, sole item on the card, nothing else competing for attention. Last time it was item 1 of five and it got nothing.

And the one structural instruction: run more clean laps before chasing the number. Fifteen clean laps is not enough attempts at a lap that needs six corners to coincide — Saturday needed 41 to produce one 1:42.7.

What to carry to Ozark / AMP

DoctrineMinimum speed, not later braking

The pattern is car-independent and it is the single most valuable thing in two days of data: arrive as fast or faster, brake harder, over-slow the minimum by 7–9 MPH, then rebuild on turbo torque. It looks fast, it feels committed, and it loses to a slower car that simply carries the minimum. Bill's minimum was below the M3's in 9 of 15 corners and he lost time in every one of those nine. At a new track, the first question in every corner is where is my minimum and how high is it — not how late can I brake.

TechniqueThe T5-style trail release

Brake later and to a genuine threshold, then bleed pressure through turn-in with real overlap instead of being fully off the pedal before the apex. At Barber that was worth 0.20 s of ceiling and — the part that transfers — it lifted the floor of the bad attempts by 0.23 g. Any off-camber compression at any track gets this treatment. The cue that does not transfer is "be patient": at T5 the patient attempts were the slow ones, and the word reinforced the actual error.

SetupPressure discipline — hot targets, not cold

34 psi front, rear two under, measured hot, bled after every session as ambient climbs. 30/29 cold in August grew to 40/39 hot and the car went loose; bleeding to 32/30 hot fixed it inside one lap. The tell for over-pressure is mid-corner speed loss in the fast right-handers — not peak lateral g, which stays flat and will lie to you.

ReliabilityThe fuel rule

Top off between every session. No exceptions, no quarter-tank folklore. The starve happened at ~6 gallons below full — about 56% of the tank — and it applies to every G-chassis car in the fleet, not just the Z4. It costs nothing and it is the entire fix. Ozark and AMP both have sustained-load sections that will find it.

MethodSacrifice-chain thinking

Corners are not independent. Running wide at the Carousel wrecks T14a and T15; over-slowing T12 costs again on the 8.2% climb out; a botched T5 entry is paid back over 230 m of exit. Before spending a session on a corner, ask what it feeds. And the corollary from this weekend: one assignment per session beats five. T1 got a single narrow sentence and improved; the Carousel got item 1 of a five-item list and went backwards.

Lessons banked

MechanismG-chassis fuel starvation, confirmedSOLVED

Saddle tank, one delivery module on the active side, and a passive venturi jet pump that is driven by pump flow — so when the active-side collector runs dry, flow stops, and when flow stops the transfer stops too. Sustained lateral load plus the light crest at T4 empties the collector, rail pressure sags, and the DME latches a reduced-power map until an ignition cycle. Triggered at ~56% of a full tank, so the street "quarter tank" number is wrong for track use. Full research note with sources: docs/research/g-chassis-fuel-starve.md.

SignatureThe pressure tell

Over-pressure understeer shows up as progressive mid-corner speed loss within a run in the fast right-handers (T9, T10, Carousel), lap over lap. It does not show up in peak lateral g, which held flat and even rose late in the day. Checked both days at 34F/32R: every apparently negative slope traced to a single traffic lap, with surrounding laps holding or climbing. No falloff.

TiresHeat cycles: the assumption was wrong, in the good direction

Entering at 13 heat cycles with falloff expected past 15, the plan was to take the shot early. Instead Saturday's 16:01 run was the day's second-fastest on the oldest rubber, and Sunday's peak lateral g (1.32) was the weekend's highest. Nothing in two days supports a heat-cycle ceiling on this set yet. Exit count ~21 is an estimate — no tire_day row was written trackside, which is a logging habit to fix, not a code change.

RepeatabilityThe remaining lap is assembly, not speed

Every one of the fifteen corners now has a weekend best that beats the May M440i template. The whole-weekend stitch is a 1:40.70 from fifteen different laps, 2.00 s under the PB. The sum of per-corner spread grew from 3.68 s Saturday to 4.81 s Sunday. There is no new speed to find at Barber right now; there is a consistent lap to assemble.

Data-loop bugs the weekend surfaced

TDAI-139 — the adapter's lakehouse flush blows the 900 s Lambda cap on every ≥30-lap 25 Hz session, silently retries, and no DLQ catches it: Aurora had 51 laps while the lakehouse showed 27. Because LapTimer exports are cumulative, every upload re-flushes the whole session and it gets worse each time. Fix is to hand off to the staged single-commit path (~13× faster). Practical effect on a track weekend: budget 30–60 minutes before per-corner analysis is possible — upload and walk away.
TDAI-143 — the AiM adapter derives lap time from the span of row timestamps rather than the beacon delta, so at 20 Hz every AiM lap is 16–67 ms optimistic. Mike's 1:42.163 was stored as 1:42.100. Found here because the virtual-race self-test failed by exactly that amount; affects every AiM member.
TDAI-144 — the lap-quality validator re-bases each track's reference median on every new member ingest, so Mike's Barber upload re-scored all 662 Barber laps and quietly moved 6 of Bill's from public to fail-partial. The new verdicts look correct, but public lap lists should not change without a human decision or at least a diff report.

Known gaps

Stated so nothing here reads as more certain than it is.

Heat-cycle exit count is an estimate (~21), not a measurement — no tire_day row exists for 8/15 or 8/16.
No throttle, brake-pressure, steering, rpm, fuel or temperature channel exists in either session — this logger recorded GPS + IMU only. Brake points are inferred from longitudinal g, throttle pickup from sustained positive g, and "steering correction" from a lateral-g wobble index. There is no tire-temperature or tire-pressure telemetry anywhere in this dataset; the pressure findings come from grip behaviour.
Sunday lap 2 (70.61 s) is excluded. It is marked valid but is a GPS artifact — a physically impossible lap time. Sunday lap 23 (1:43.41) keeps its lap time but contributes no corner data: telemetry stops at 3158 m.
The three fuel-starve laps (Saturday 19–21) are excluded from all clean-lap statistics — they are a car condition, not driving.
Peak lateral g and brake points are not compared across drivers. Different loggers at 25, 20 and 10 Hz; Mike's GPS-derived g channel reports a non-physical 2.12 g peak and is ~2× noisier per sample. Only lap times, zone times and speeds are compared across cars.
The May comparator is a different car (M440i, 10 reference laps, median per corner), so absolute cross-car speed deltas mix car and driver. It is used here only as a floor check — and it no longer binds anywhere.
Corner-window boundary caveat. T6, T8, T9, T10 and T15 have no distinct brake event and their minimum sits at a window edge; the Carousel window ends mid-deceleration, so the complex's true minimum lives in T14a. The same windows are applied to every lap and to the May baseline, so comparisons stay valid.
weather_condition is an unmapped LapTimer integer code (2→4 Saturday, 1 Sunday) and is deliberately not rendered as a condition string.