ET prediction
Drag Racing ET Prediction Software
The ET predictor is a regression model trained on the runs you've logged for each car. It corrects for density altitude, headwind, and engine temps, and it updates live as the weather changes between rounds. Bracket racers use it to dial in; sportsman racers use it to confirm setup changes.
Is the ET prediction model trained on my own runs?
Yes. Generic ET calculators assume an "average" car. Yours isn't average. The Drag Race Logbook Pro predictor builds a weighted linear regression from your own logged runs, so the slope it fits is your car's actual altitude sensitivity, the intercept is your car's actual baseline, and the residuals shrink every time you log another pass.
When you predict at a track you've raced before, runs at that track get full weight in the regression. Runs at other tracks get a smaller weight (so cross-track surface-prep noise doesn't bias the slope) and their density altitude is normalized toward the target track's elevation. The result is a prediction that reflects your car's real response to today's air at the specific strip you're standing on.
How is the drag racing ET prediction built?
When you open the predictor for a car at a track, the app pulls every run for that car, applies same-track weighting and DA normalization, and fits the regression. Then it adjusts the prediction in three steps:
- Density altitude — the regression slope predicts the air response
- Headwind / tailwind — empirical 0.010 seconds per mph of headwind, projected onto the track's compass orientation
- Engine vitals — oil temp above 250°F, coolant above 210°F, intake above 60°F, and trans above 200°F each add a small penalty in seconds
- A confidence band is built from fit quality (RMSE), effective sample size, and the ratio of same-track to cross-track runs
Reading the predictor card
The predictor shows the predicted ET in big numbers, with the confidence band below it. A wide band means low sample size — early-season at a new track, or after a major setup change. A narrow band (often under ±0.02s) means the model is confident.
The card also surfaces the headwind adjustment, the thermal adjustment, and a quick view of which runs the model is leaning on hardest. If the predicted ET feels off, those drill-downs usually tell you why: a hot oil temp pushing the model conservative, a recent setup change throwing off the fit, or simply too few runs to be trusted yet.
Does live weather drive real-time ET predictions?
Yes. The predictor doesn't freeze on the conditions you logged. It pulls live weather for the selected track from the National Weather Service (and Open-Meteo as a fallback for global coverage), computes today's density altitude, and projects your model onto it.
If a cool front rolls in between rounds and DA drops 1,200 feet, the prediction reflects it immediately. You can also snap a Kestrel or weather app photo to override the live fetch with what's actually happening at your spot in the lanes — sometimes the closest weather station is miles from the track and reads materially different conditions.
Split-ET predictions for 1/8 and 1000-foot
On top of the main ET prediction, the model also fits separate regressions on the 1/8 mile (660-foot) and 1000-foot splits when you have at least three runs that carry those values. Those split predictions are useful for two things: confirming the predictor agrees with what you saw on the 60-foot when you were halfway down the track, and noticing when a setup change is moving the back half faster than the front (or vice versa).
Sign-clamping to physical reality
There's one safety rail built into the regression: density altitude can't make your car faster. More DA means thinner air which means more ET. If your sample of runs is small and noisy enough that the raw regression fits a negative DA slope — which happens with 4–6 runs more often than you'd think — the predictor clamps that slope to zero and falls back to your weighted-mean ET adjusted for wind and gauges.
Without that clamp, the model can produce nonsense like "12.0 seconds at 7,500 ft DA" for a car that has never run quicker than 12.17 anywhere. The clamp is invisible during normal operation but it stops the bad case dead.
Bracket racing: dial-in support
Bracket racers can use the predictor's main ET output as the basis for the dial-in — most do, with a small safety margin for tree luck and surface variability. The Race Mode screen surfaces the predictor right next to the dial-in input so you can copy a value across in one tap.
Class and index racers can run the predictor purely for trend analysis: are we faster today than yesterday for a given DA? Did the setup change move the number? The card includes a delta against your rolling 10-run average so you can answer those questions without leaving the app.
Example: how it works in practice
Given
You have 23 logged runs on a Top Sportsman Camaro across two tracks. You're at Bandimere, DA today is 7,100 ft, headwind 4 mph, oil temp 245°F. Bandimere has 14 same-track runs in your history; the other 9 are from a sea-level track.
Result
The regression fits a positive DA slope of about 0.0007 seconds per foot, weighted heavily on your Bandimere runs. Cross-track runs are downweighted 0.35x and their DA is shifted toward Bandimere's elevation. Base prediction: 9.42s. Headwind adjustment: +0.04s. Thermal adjustment: 0 (oil under threshold). Final prediction: 9.46s ±0.018s. Confidence: 88%. The card shows your last three Bandimere runs (9.44, 9.47, 9.45) as the closest historical matches.
Frequently asked questions about et prediction
How many runs do I need before the predictor is useful?
A useful prediction kicks in at 3 runs with varying density altitude. ±0.05s confidence typically arrives around 8–12 runs. Tight ±0.02s confidence usually shows up around 20+ runs, especially when the runs span 1,000+ feet of DA range.
What if I've only run one track?
The predictor still works — same-track weighting just becomes the whole sample. You'll get a confident prediction at that track once you have a few runs across different DA. Predictions at OTHER tracks will be lower confidence because the model has no track-specific signal there.
Does the predictor learn from my opponents' slips?
No. The predictor is strictly per-car and per-account. Your opponents' data never enters your regression, and your data never enters anyone else's. The model that runs at predict time uses only your runs.
How is density altitude calculated?
The same way a Kestrel or Computech meter does it: from air temp, relative humidity, station pressure, and field elevation. The app pulls live values for those inputs (or reads them off a photo of a meter), then computes DA. The full method follows the standard psychrometric formulas — vapor pressure from temp + RH, then virtual temperature, then geopotential altitude via the ICAO standard atmosphere.
Can I see how the prediction was built?
Yes. The predictor card shows the slope, the intercept, the RMSE, the sample size, how many runs were same-track, and the breakdown of wind + thermal adjustments. If a number looks weird, the math is visible.
Does setup change affect the model?
It does — the model treats all your runs as one population. If you make a major setup change (new converter, gear, tune), older runs will pull the prediction toward where you used to be. Practical workaround: delete or mark the pre-change runs once you have 5+ runs on the new combination, or just trust the higher-RMSE confidence band while the new sample builds up.
Start using ET prediction in Drag Race Logbook Pro
5 manual runs + 2 logbook pages on the free plan. No credit card required. Pro is unlimited — cancel anytime.