The 13 CSR2 Cars That Lost to Their Own Dyno
1,230 of 1,246 CSR Racing 2 cars beat their dyno prediction. Thirteen did not, by margins as small as 0.003 seconds. What the edges of the data show.
Almost every car in CSR Racing 2 beats the time its dyno predicts. We have quoted the figure often enough: 1,230 of the 1,246 cars with both numbers finish quicker than the game said they would.
This article is about the other sixteen, and about what the tightest margins in the database reveal.
Thirteen losses and three dead heats
Of the 1,246, thirteen cars recorded a time slower than their dyno and three matched it exactly to the thousandth.
| Shelby Cobra Daytona (E.C. 36) | +0.011s | |
|---|---|---|
| McLaren P1 GTR | +0.008s | |
| Bugatti Chiron Super Sport | +0.007s | |
| Ferrari 512 Testarossa | +0.006s | |
| McLaren P1 | +0.005s | |
| Bugatti Chiron SS 300 | +0.004s | |
| Pagani Huayra Codalunga | +0.003s |
The worst failure in the entire database is eleven thousandths of a second.
The largest shortfall anywhere on the sheet is 0.011 seconds. Not a tenth, not a hundredth — eleven thousandths, on a Shelby Cobra Daytona Coupe at Elite Customs level 36.
Twelve of the thirteen are Tier 5 cars, and the list reads like a hypercar catalogue: two McLaren P1 variants, two Bugatti Chirons, a Pagani Huayra, a Ferrari 512 Testarossa.
Why these cars and not others
The pattern is consistent with everything else we have measured. These are heavily built, top-tier cars, and our cross-tabulation found that the more finished a car is, the less a driver can add — cars above 2,500 EVO return a median 0.048 seconds against 0.184 for cars under 1,500.
Push that trend far enough and it crosses zero. A car with effectively nothing left to give will land on its prediction, and on a given run it can land a few thousandths the wrong side of it.
So these are not badly driven cars. They are cars where the dyno was already correct, and the run came in a hair under what a perfect one would have been. On this evidence CSR2's dyno is not a bad predictor at all — at the very top of the game it is close to exact.
Three cars landed exactly on the number
Three cars matched their dyno prediction to the thousandth — not faster, not slower, identical:
| Car | Dyno | Recorded | EVO |
|---|---|---|---|
| Pininfarina Battista Anniversario | 6.792 | 6.792 | 2,568 |
| Bugatti Veyron Jean Bugatti | 7.594 | 7.594 | 2,118 |
| Lamborghini Aventador Coupe | 9.462 | 9.462 | 1,219 |
The Battista is the one that makes sense. It is an electric car on 2,568 EVO, and electric cars return the least to driving of any group we have measured — a median 0.076 seconds against 0.115 for the field. There is no gearbox to mistime and almost nothing for a driver to add, so landing exactly on the prediction is close to the expected outcome.
The Aventador on 1,219 EVO is harder to explain. A car that lightly built should be returning something closer to 0.18 seconds. Either that run left time on the table, or the entry is one of the handful on this sheet that is not quite right.
How tightly packed the record sheet is
The same compression shows up when you look at the times themselves rather than the predictions.
| Measure | Count |
|---|---|
| Recorded cars with a time | 1,251 |
| Distinct times among them | 1,004 |
| Times shared by two or more cars | 199 |
| Adjacent pairs separated by exactly 0.001s | 318 |
| Adjacent pairs that are exact ties | 247 |
1,251 cars produce only 1,004 distinct times. 247 adjacent pairs are exact dead heats and another 318 are separated by a single thousandth.
Put differently: in roughly 45% of cases, the car immediately ahead of yours in the order is either tied with you or one thousandth quicker.
Five separate cars record exactly 7.159. Four record 6.759 and another four record 6.796. At that density, the ordering of the leaderboard around you is close to arbitrary, and a single better run reshuffles it.
The five cars tied on 7.159 are a good illustration of how little separates neighbours. They occupy global ranks 345 to 349, and their EVO figures run 2,311 to 2,323 — a spread of twelve points across five cars that recorded an identical time. Two of them, a McLaren Senna and a Senna GOSH Edition, carry exactly the same EVO as each other.
What this changes about how you read a rank
A rank number looks precise and mostly is not. If 247 adjacent pairs are tied outright, then being 180th rather than 150th may represent no measurable difference in car speed at all.
- Treat rank gaps under about 30 places as noise unless you check the actual times. The times are what is real; the rank is a sorting of them.
- A thousandth genuinely matters here. It is the difference between tying and being ahead, 318 times over.
- Do not chase a specific rank. Chase a time, because rank moves when other people improve and time does not.
- If your car sits in the dense 6.85 to 6.95 band, where 91 cars are packed into a tenth, small improvements move you a long way. In the sparse parts of the field they move you almost nowhere.
The honest reading of the dyno
We have spent several articles pointing out that the dyno underestimates almost every car. This is the counterweight, and it matters for fairness.
The dyno is wrong in one direction, by a median of 0.115 seconds, and it is wrong because it predicts a clean run rather than a well-driven one. But it is not sloppy. At the top of the game it is accurate to within thousandths, and in sixteen cases out of 1,246 the recorded run failed to beat it at all.
Treat it as a floor on a mid-build car and as a near-exact forecast on a finished one. Both readings come out of the same dataset, and which applies depends entirely on how far along your car is.
Sort the record database by the difference column and these sixteen cars collect at one end, with the 1,230 that beat their prediction stretching away in the other direction.