Seven Seconds Became Nine-Tenths

AI-coached track weekends · Atlanta Motorsports Park, September 2026. The sequel to The Mini Out-Cornered My M40i. It Lost by Seven Seconds Anyway.

Same weekend at Road Atlanta: The Race That Never Happened · Fourteen Years, Eight and a Half Seconds

Sunday, the T6 apex, the 1:41.440 lap. Harry's overlay shows the corner-minimum speed and the corner map right here. This is the corner my shop points to when they say I'll beg for an LSD.

Four laps, one grid. The line at zero is my April M40i. Everyone else is measured in seconds behind it.

In July I lined up Rich's race-prepped Mini Cooper S against my Z4 M40i at Road Atlanta and the BMW won by 7.318 seconds. The Mini won the corners, and the M40i won the straights by so much that the corners didn't matter.

In September I took the same idea to Atlanta Motorsports Park: 1.835 miles, sixteen turns, about 87 feet of elevation change (surveyed), and no straight long enough to hide on. Same two cars. This time I also added two more of mine: my M440i on a soaked Saturday and my E86 Z4 on a dry Sunday.

The finishing order:

	Car	Lap	Logger	Behind
1	Z4 M40i "Smurfette" (April 25)	1:36.500	GPS, ~1 Hz	—
2	Rich's Mini Cooper S (Sept 13)	1:37.444	AiM, 10 Hz	0.944 s
3	E86 Z4 3.0si, dry (Sept 13)	1:41.440	Harry's LapTimer, 25 Hz	4.940 s
4	M440i, wet (Sept 12)	1:55.570	Harry's LapTimer, 25 Hz	19.070 s

Seven seconds became nine-tenths. The rest of this post is about why, and about how much of that I can actually prove.

Why AMP shrinks the gap

On paper nothing changed. The M40i makes 382 hp and weighs approximately 3,450–3,500 lb. The Mini weighs approximately 2,350 lb race-prepped, and a stock Cooper S of that generation makes 181 hp. (Both are public spec figures. Neither car went on a scale or a dyno for this post.) The BMW has roughly 40% more horsepower per pound. At Road Atlanta that was worth 6.8% of the lap. At AMP it was worth 1.0%.

The answer is the shape of the track. At Road Atlanta the M40i arrived at the chicane at 136 mph, 16 mph up on the Mini, after leading by as much as 23 mph down the back straight, and made 3.9 seconds on the back straight alone. AMP never lets the M40i run like that. The longest uninterrupted pulls are short — about 316 m on the front straight (92 to 117 mph) and 326 m from the Carousel to T6. The other big pulls (T1 exit, T6 exit to T10) are 200–300 m each. None of them is long enough to build a real speed gap, and there's no throttle channel to call any of it full throttle, so the top-speed difference topped out at 6.7 mph (116.9 vs 110.2).

The Mini is never ahead. It's 2.39 s down braking for T10, then takes 1.45 s back on the way home. The shaded band is the 1 Hz uncertainty, more on that below.

So here's the race. They start together, and the M40i pulls away through every acceleration zone. At 1,387 m, braking for T10, it's 2.39 s up the road. The M40i won the four acceleration-led sections by 2.40 s combined.

Then the lap turns hilly and flowing, and the Mini starts eating:

Section	Who gained
Line to T1	M40i +0.64 s
T2–T3 + Carousel	tie (Mini +0.08)
T5 downhill to T6	M40i +0.60 s
T6 + exit	M40i +0.46 s
T7–T9 run	M40i +0.69 s
T10 + T11 Dip	tie (Mini +0.07)
T12 crest + T13	Mini +0.68 s
T14	Mini +0.28 s
T15–T16 + front straight	Mini +0.47 s

That last row is my favorite number in the post. The Mini beat the M40i over a section that includes the front straight, because it came through T16 4.0 mph faster and the straight wasn't long enough for the B58 to collect. It also carried 4.4 mph more over the blind T12 crest. From T12 to the line, the Mini took back 1.43 s.

The M40i's 2.4 seconds of power, minus 1.4 seconds of momentum payback, leaves you with about a second. That's AMP.

Out-cornered again: 7 of 10

At Road Atlanta the Mini beat or matched my apex speed at six of seven corners. At AMP it happened again:

Corner	Mini mph	M40i mph	Holds up at 1 Hz?
T1 (downhill hairpin)	34.6	35.0	tie, inside the error
T3 (off-camber)	55.6	52.7	Mini +2.9, yes
T4 Carousel	45.1	43.9	Mini +1.2, marginal
T6 (slowest)	38.2	36.0	Mini +2.2, yes
T10	57.2	58.9	M40i +1.7, marginal
T11 The Dip	54.5	51.0	Mini +3.5, yes
T12 (blind crest)	44.2	39.8	Mini +4.4, yes
T14	78.4	80.7	M40i +2.2, yes
T15	95.6	94.1	Mini +1.5, marginal
T16	100.9	96.9	Mini +4.0, yes

That's seven of ten for the Mini, and five of them clear by more than 2 mph (T3, T6, T11, T12, T16). I left T13 out on purpose. My April lap drove T11–T12–T13 as one carry-speed arc, which is exactly why it's my reference lap. The Mini was still braking into T13 in that window, so comparing the two isn't apples to apples. Call it a different line, not a win.

Same lesson as last time, and I'm still the one who has to learn it. The Mini's apex speeds are proof those corners go faster. I can practice that. Rich can't practice his way to 382 horsepower.

The 1 Hz confession

Same lap, two sample rates. Where 1 Hz lands on a slow corner decides what it thinks the minimum speed was.

Here's what I have to admit. My April M40i lap was logged at about one sample per second, 101 GPS fixes for the whole lap. There's no throttle, no rpm, no brake. In the Road Atlanta post I wrote off that old 1 Hz history as unable to do any of this. I wanted to know if that was actually true.

So I took Rich's 10 Hz lap, threw away nine samples out of every ten (at all ten possible starting offsets), and measured how wrong the result got compared with the real thing. That spread is the honest error bar on my M40i lap.

What survives at 1 Hz:

• The lap time and the final margin. The timer isn't the GPS sample.
• The overall shape of the gap trace.
• Top speed, within about 1–2 mph. My 116.9 might really be closer to 118.
• Section times, good to about ±0.15 s mid-lap. Every section above except the Carousel and T10 is bigger than that, so those two are honest ties.
• Apex-speed differences bigger than about 1 mph. 1 Hz reads corner minimums a bit high (+0.1 to +0.4 mph on average, up to +1 mph worst case), because a sample rarely lands on the true bottom. That bias flatters the M40i, so every "Mini carried more" above is, if anything, understated.

What doesn't survive:

• Brake points. A 1 Hz sample moves where braking seems to start by ±15–30 m. At 86 mph you cover 38 m in a second.
• Braking g, peak lateral g, trail braking. None of it.
• Anything under 1 mph at an apex, and T13's minimum.
• Exactly where the M40i's lead peaked, to better than about ±20 m.
• Any throttle or rpm story, since there's no OBD data.

So my old data can do about half of it. It can tell you who won and where, but it can't tell you how. That's why this post has no brake-point comparison for the M40i, and why the next AMP visit gets a 25 Hz logger.

The rain took 14 seconds, and mostly not where I expected

Saturday, the wet M440i just before the T13 apex. This still is from the 1:55.78 session-best lap, not the 1:55.570 overall best used for the numbers below.

The wet M440i through T6 on the 1:55.78 session-best lap: 29 mph and 0.7 g at the apex — the slow corner where it nearly matched the dry E86's numbers, more on that below.

Purple is where the rain hurt. The two green bars are where it barely did.

Saturday it rained. 13.7 mm fell on the day, with 4.4 mm in the 9 am hour alone. My first session in the M440i produced laps in the 2:24–2:59 range, which aren't really laps at all. By the last session the rain had eased to about 0.3 mm an hour and the line was drying under me. The last two laps of the day were 2.6 seconds or more quicker than the four before them, and the best was 1:55.570.

Sunday was dry, and the E86 ran 1:41.440. Both laps came off the same 25 Hz Harry's LapTimer setup, so for this pair there's no sample-rate excuse. The gap is 14.13 s.

I assumed the wet lap lost most of that under braking. It didn't:

What the dry E86 was doing	Wet M440i lost
Braking	2.08 s
Cornering	9.24 s
Power / straight	2.81 s

Two thirds of the rain's cost came mid-corner. The braking points barely moved — T3, T6, and T10 all landed within about 20 m of where they did on the dry lap. I braked in the same places and then couldn't carry the speed through. The fast corners hurt most, with T14 at −14.6 mph, T15 at −12.5 and T16 at −15.3. T14 to the line cost 3.94 s, 28% of the whole gap. Coming off T16 15 mph slower is also why the M440i's top speed was 97.5 mph to the E86's 113.2.

And then there's T6 and the hill after it, where the wet M440i basically matched the dry E86:

• T6: 29.0 mph minimum vs 30.5, brake point 968 m vs 962, peak decel −0.73 g vs −0.72 g, exit 56.6 vs 56.9 mph. The section cost 0.54 s.
• T7–T9: up the hill, the M440i was at or above the E86's speed from about 1,137 m to 1,299 m, with 68.5 mph at the T7 apex vs 64.4. The whole track-out cost 0.12 s.

A slow corner in the wet is still a slow corner. The uphill run after it is about traction and torque, and a B58 on a climb covers for a lot of missing grip.

One honesty note: this isn't a pure weather comparison. The M440i is a heavier car on Michelin PS4 street tires. The E86 is on Toyo R888Rs. So "the rain cost 14 seconds" really means rain plus car plus tire. The weather numbers are hourly reanalysis for the track's grid cell, not a rain gauge at the pit wall.

Does the E86 actually need an LSD?

My shop swears I'll "beg for an LSD" once I'm on track. T6 at AMP seemed to agree — the Z4 steps out on exit every lap. Controllable, fun, and I assumed slow. So I pulled twelve dry laps and asked the data one question: is the open diff actually limiting how hard I can drive off the corner?

The slide and the steering correction after the apex, T6, the 1:41.440 PB lap — 30 mph at the corner minimum, back up to 56 mph here.

It isn't — not yet. Coming off T6 at 30 mph, the car is already putting down 0.4 to 0.5 g of forward acceleration while still pulling 0.7 g sideways. On a straight, at those speeds, the 3.0si only makes about 0.54 g, estimated. That's nearly everything the engine has, delivered mid-exit. An open diff that was really spinning the inside wheel would show a flat ceiling well below that. It doesn't.

Drive-g reaches 75-100% of the car's own straight-line limit while lateral g is still 0.6-0.8. That's not what a traction-capped open diff looks like.

What the traces do show is a car that runs out of lateral grip everywhere, on R888Rs that deserve a better platform. My best T6 exit wasn't about traction at all. I carried two more mph and unwound the wheel sooner.

T6 isn't an outlier vs. the rest of the lap, and the car is using 0.93-0.97 of its available grip at T3, T4, T5, T9 and T14 — that's the coilover argument, not the diff argument.

T1 for comparison — 31 mph at the corner minimum, then up to 1.1 g on the way out. Lateral-limited, not drive-limited. More drive here wouldn't even be usable.

My honest math: an LSD might be worth a tenth or two. Coilovers, maybe a second. So coilovers go first. The LSD comes after, once a stiffer car starts lifting that inside rear, and Sergio gets to say I told you so.

Two honesty notes. First, this data has no gyro or wheel-speed channels, so the slide itself can't be measured — only whether drive was capped, and it wasn't. Second, the coilover second is a physics estimate from +5% sustained lateral grip, not a measurement — treat it as a range, 0.5 to 1.5 s.

The E86 got steadier as it got faster

The quiet story of the weekend is the stock-ish E86. First session: it opened at 1:55.8 and worked down to 1:42.84. Second session: 1:41.44, seven laps within 3% of best with a 0.93 s standard deviation. Third session: 1:41.62 best, and the tightest run of the whole weekend, four laps within 1.23 s and a 0.53 s standard deviation.

It finished 4.94 s behind a car with a lot more power, from April, on a lap I've been polishing for a year. I'll take that. Ask me again after the coilovers.

The fine print

How this was built: every lap is real telemetry, aligned by distance around the lap using the same method as the Road Atlanta virtual race, with each gap trace checked against the true lap-time margin. The Mini-vs-M40i race lands on 0.944 s and the M440i-vs-E86 pair lands on 14.13 s exactly.

It's a virtual race, not a same-day one. The M40i lap is from April 25, five months before the others, and it's the carry-speed lap I've written about before. My all-time AMP best is still 1:36.47 from September 2025.

Rich's lap trace doesn't quite close: the AiM export runs 22.6 m past its own start, about 0.58 s of extra track. I've used his logger's official 1:37.444 throughout, since the beacon lap time is authoritative and the GPS trace isn't. Section times carry that ~0.6% as uniform uncertainty.

The four-car chart at the top puts the M40i on the pure distance-fraction method, which at 1 Hz can wobble by up to about half a second mid-lap. The finishing gaps are exact. The two-car Mini chart uses a tighter position-matched alignment.

Weather data by Open-Meteo.com (ERA5 hourly reanalysis).

For the curious: all four speed traces. The dots are the only moments the April M40i was actually measured.

Race me: send your 10 Hz+ AMP data (Harry's LapTimer .hlptrl, or CSVs from AiM Race Studio, RaceBox, Porsche PTPA) to upload@trackdays.ai.