Update 1.71 changed the physics for every car in GT7. Everyone has an opinion on how it feels. I happened to have telemetry from before it landed, so I went and measured what actually moved.
This page is the working. Every number in the video is here with the method that produced it, including the things I could not test.
Where the data comes from
I have been building a coaching app that records every lap you drive and every replay you watch, at 60 Hz. When the update dropped I already had Circuit Experience laps and leaderboard laps sitting in the library from before it. That is the whole reason this comparison is possible: the before was already recorded.
The core comparisons are:
- A pro's Yas Marina lap in a VW Beetle Gr.3, before and after. Same driver, same car.
- My own Porsche 911 GT3 R (992) stints at Yas, 9 laps before against 20 after, fuel-matched so weight is not a confound.
- Ten Circuit Experience pairs across Spa, Monza, Le Mans, Fuji, Barcelona, Tsukuba, Red Bull Ring, Eiger, Blue Moon Bay and Yas.
The pro's lap, same car, before and after
| Before 1.71 | After 1.71 | |
|---|---|---|
| Lap time | 1:52.694 | 1:52.328 |
| Acceleration | 12.1% slower, all five speed bands | |
| Top speed | 252.9 kph | 248.3 kph |
| Brake point | 11.9 m later on average, at all six corners | |
| Tyre sliding under braking | 55.7% | 18.3% |
His car lost 12% of its acceleration and he went three tenths quicker.
Where the time went: he gave up 0.578 s on the straights and found 1.050 s under braking and through the corners.
That split is also why the meta changed. The car at the top of the board now gives up almost nothing on the straights, 0.043 s, so it keeps far more of what it gains. That one is a different car and driver on each side, though: the old record was a Porsche 992 and the new one is a Genesis.
Braking against a real GT3
This is the part I expect to be argued with, so here is the method in full.
I used a real onboard of an AF Corse Ferrari 296 GT3 at Yas Marina with a telemetry overlay running, and read the speed frame by frame through the heaviest braking zone on the lap. By hand, because automatic reading kept dropping digits and I did not trust it. That gives 22 frames at 4 Hz covering 49.00 s to 54.25 s of the lap, from full throttle in sixth to corner entry in first.
The measurement, applied identically to the real car and to every GT7 lap: find where speed crosses 200 kph and where it crosses 100 kph, interpolate both crossings in time, then integrate speed between them to get the distance.
| 200 to 100 kph | Average g | vs real | |
|---|---|---|---|
| Real Ferrari 296 GT3 | 72.9 m | 1.62 | — |
| GT7 1.70, VW Beetle Gr.3 | 66.9 m | 1.76 | −8.2% |
| GT7 1.70, Porsche 992, 9-lap mean | 64.8 m | 1.82 | −11.1% |
| GT7 1.70, Porsche 992, top of board | 64.5 m | 1.83 | −11.6% |
| GT7 1.71, VW Beetle Gr.3 | 60.1 m | 1.96 | −17.5% |
| GT7 1.71, Porsche 992, 20-lap mean | 57.7 m | 2.04 | −20.8% |
| GT7 1.71, Genesis Gr.3, top of board | 55.4 m | 2.13 | −24.0% |
Nobody has ever claimed GT7 brakes like a real GT3, and before the patch it was already stopping 8 to 12% shorter. The finding is not that a realistic thing broke. It is that the gap widened, from 8-12% to 17-24%.
Peak deceleration agrees independently. The real car's overlay peaks at 2.20 g. My 992 was already peaking at 2.39 g before the patch and 2.57 g after it.
Averaged over the whole 200 to 100 kph stop, the real car pulls 1.62 g and every GT7 lap sits above it, from 1.76 to 2.13.
Caveats, stated plainly. Different cars, though all GT3 class. The real lap is lap 12 of a twelve-hour race with fuel aboard and worn tyres, so a qualifying lap would brake slightly shorter and close a couple of percent. One braking zone. None of that closes a 24% gap.
Steering
This is the change people are describing as understeer, and it is real, but it is not less grip.
Cornering force against steering angle, my 992, 100 to 180 kph, same laps either side:
| Cornering force | Lock needed, 1.70 | Lock needed, 1.71 | Extra |
|---|---|---|---|
| 0.30 g | 16° | 24° | +48% |
| 0.40 g | 26° | 32° | +24% |
| 0.50 g | 37° | 45° | +21% |
| 0.60 g | 54° | 64° | +20% |
| 0.65 g | 71° | 71° | 0% |
The two curves cross at about 69°. Below the crossover you turn the same amount and get less back. Through the middle of the range the two trade places. Past roughly 100° of lock the new model holds 0.73 to 0.75 g where the old one had already fallen away to 0.63-0.70.
At the 0.65 g point the two need identical lock, which is why slow corners feel completely normal and the quicker stuff feels wrong.
Polyphony state they optimised steering geometry per car, so your car will not match mine. The direction is what transfers, not the numbers.
Grip went up
Peak cornering force by speed band, my 992, same car and driver, 9 laps against 11. The steering and grip figures were cut from the first 11 post-patch laps; the braking table above includes the 9 laps captured after that:
| Speed | 1.70 | 1.71 | Change |
|---|---|---|---|
| 80–120 kph | 0.812 g | 0.837 g | +3.1% |
| 120–160 kph | 0.800 g | 0.835 g | +4.3% |
| 160–200 kph | 0.876 g | 0.888 g | +1.4% |
| 200–240 kph | 0.915 g | 0.895 g | −2.3% |
Grip is up everywhere below 200 kph and slightly down above it. The car did not get worse. It got harder to reach.
It is not Balance of Performance
Two cars in the test had zero power change and zero weight change and still lost 6.0% and 7.7% of their acceleration. Circuit Experience does not apply BoP at all.
If your PP dropped, nothing was taken off your car. Across the whole Gr.3 field, average power moved by one horsepower and weight by one kilo. The scale changed, not the cars.
What Polyphony said, against what showed up
| From the 1.71 update notes | In the telemetry |
|---|---|
| "Optimized steering geometry for each car to improve the simulation of cornering forces" | Up to 48% more lock for the same cornering force |
| "The slip ratio control and cornering brake behaviour under ABS has been adjusted" | Brake points 11.9 m later, zones 17% shorter |
| "Improved tire behavior during wheel slip" | Sliding under braking 55.7% to 18.3% |
| "Optimized road-surface and rolling resistance" | Top speed down 4.6 kph, acceleration down in every band |
| "Improved vehicle speed control under partial throttle" | Partial-throttle share 15.8% to 9.9% |
That fourth one is worth sitting with. Rolling resistance is why every car is slower in a straight line on every track, and it has been in the patch notes since day one.
What I could not test
- "Revised tire heating and wear values." Not tested. Tyre temperature is not stored per sample in this data set, and wear is not in the telemetry at all, so there was nothing to compare either half against.
- "Improved throttle response while drifting." No drift laps in the set.
- "Reworked damper characteristics." Nothing in the telemetry measures this.
Two more things I tried to measure and could not support:
- Trail braking. The share of the braking zone spent already turning did not move, 60.6% to 62.1% across 180 braking zones. Absolute trail distance fell, but only because the whole zone got shorter.
- Brake release rate. Release times fell, but by less than the zone compressed, so as a proportion of the zone it is unchanged. There is no separate technique change here.
How things are measured
Braking zones are runs of at least 10 samples with the brake above 3%, with short gaps merged. Brake points are compared zone to zone between paired laps.
Cornering force is calculated as speed multiplied by yaw rate, which is the standard derivation for a car on a curved path. It is good for comparing two laps measured the same way, and it is not used for any realism claim. The realism argument rests on braking distance, which was measured on the game and the real car by identical code.
Tyre sliding uses a normalised slip figure with a threshold found by reproducing the game's own wheelspin counter from raw data.
Acceleration is compared only at matched speeds, so a car is measured against itself at 200 kph rather than at whatever speed it happened to be doing.
Stint braking figures are means across fuel-matched laps within 2.5 s of that stint's best, applied identically to both sides. The grip figures are the 95th percentile of lateral g pooled across the stint.
Corrections
If you find something wrong here, tell me and I will fix it on this page and say what changed. Every figure above survived an adversarial audit against the raw telemetry before publication, and several earlier numbers did not survive it and were removed.