Simulator Session Analysis: 25 Laps at Monza

A 50-minute simulator session — 25 laps in three stints, in a 991 GT3 R (2016) at Monza — produced a best lap of 1:54.380. Two questions followed. How much time is left, and is it pace or consistency? And can the data answer at all: several of the logged channels turned out to be recorded too coarsely to use directly.

1. 01Import the session — 605,140 samples at 200 Hz, 31 of 57 channels used — logging every unit conversion and assumption made along the way.
2. 02Rebuild the channels the logger recorded too coarsely: lateral g from damper roll (fit r = 0.91) and longitudinal g from speed, with a heavier 2 Hz filter.
3. 03Classify every lap. Out-laps, in-laps and pit laps are set aside, and one lap that ran 31 m short — a likely track cut — is excluded from best-lap selection.
4. 04Compare the best lap with the next-best by distance, split flying laps into three equal sectors, and track balance and lap-time trend stint by stint.

The three best sectors were set on three different laps — 14, 12 and 16 — none of them the best lap. Added together they give a theoretical best of 1:53.588, 0.792 s quicker than anything actually driven. Overlaying lap 12 on lap 11 shows the same pattern at a smaller scale: lap 12 finishes 0.390 s behind, yet gains 0.310 s in one 200 m section and loses 0.298 s in another. The second stint was the quickest and most consistent, averaging 1:54.937 over nine flying laps. In the third, lap time rose by 268 ms per lap while the understeer gradient climbed from +0.55 to +0.68 deg/g.

• The 0.79 s gap to the theoretical best is consistency, not missing pace: every sector has already been driven, just never on the same lap.
• Laps 11 and 12 trade about 0.3 s in two specific 200 m sections, so the opportunity sits in particular corners rather than being spread around the lap.
• The third stint fell away by roughly 0.27 s per lap as understeer increased; the first two stints showed no clear trend.
• Grip use averaged 38% of a 2 g reference circle, with points concentrated along the braking and cornering axes — little braking and turning combined.
• The logged acceleration channels were in 1 g steps and speed in 3.6 km/h steps. Rebuilding them made the analysis possible, with about 0.1 g of error on longitudinal g.

Up to 0.79 s is available from repeating sector times already achieved in this session, before any new pace is found — with two 200 m sections identified as the place to start.

The first job was not finding lap time but establishing what the data could support. With the unusable channels rebuilt and every assumption recorded, one session was enough to separate consistency from pace, locate the loss to specific sections, and show a balance shift across the final stint.