Cup Series · Daytona & Talladega · 18 points races, 2022–2026
Superspeedway fantasy scoring is not really a race model. It is a survival model with a place-differential bonus bolted on. Whether a car is still running at the flag explains nearly four times as much of its DraftKings score as where it started — and the number of cars that survive swings from 20 to 36 between races.
Survival is worth
39.9
DK points. Cars running at the finish average 35.8; cars that do not average −4.1.
Variance explained
45%
R² of survival alone on DK points. Starting position manages 12%. Together, 54%.
DNFs per race
4–20
The observed range across 17 races. Median 9, standard deviation 4.3.
Cup points races at Daytona and Talladega, 2022 season through the 2026 Jack Link’s 500. 707 driver-races. Source: the NASCAR results database behind TheLab.
Data noteThe 2025 YellaWood 500 at Talladega has incomplete data in the database, so it is left out of anything that depends on the missing fields — every attrition figure here uses 17 races rather than 18. The 2026 Jack Link’s 500 is missing caution and lead-change data and is blank in those columns.
Plot every driver-race by starting position and DK score and the picture is not a trend line. It is two separate populations that barely overlap.
Cars that finish score a median of 36.5. Cars that wreck score a median of −5.3, and 90% of them land under 18. There is almost no path to a usable score through a DNF — the best any wrecked car has managed in five years is 44.8, and that took a 40-position place differential before the crash.
Within each cloud, starting position does its work as a straightforward tilt. Between the clouds, it does something less obvious, which the next section is about.
Every driver-race, 2022–2026
707 entries. Horizontal position is where the car started; vertical is what it scored. Hover any point for the detail.
The reason back-markers score is not that they finish better. It is that they finish at all — and that place differential pays them for the same result.
Split the field into four grid quarters and three things happen at once. Survival climbs steadily from 67.1% for cars starting in the top ten to 80.0% for cars starting 31st or worse. Mean DK climbs much harder, from 11.7 to 35.9. But finishing position given that the car survived is close to flat — 14.0 for the front quarter against 19.2 for the back, five positions across the entire grid.
So the deep starter is not out-racing anybody. They are avoiding the wreck slightly more often and getting paid 30 positions of place differential for an ordinary result. The front-row car is doing the reverse: running where the trouble is, and holding a liability rather than an option.
What changes across the grid, and what does not
Cup entries by starting quarter, 17 races with recorded finishing status. Each panel has its own scale.
Survival rate
Share of cars running at the finish. Rises steadily toward the back.
Mean DK points
Split by outcome. Both bars rise — place differential pays deep starters either way.
Finish, if it survives
Average finishing position of surviving cars. Almost flat across the grid.
| Start | Entries | Survival % | DK all | DK if survives | DK if DNF | Finish if survives | Top 5 % | Top 10 % |
|---|
Almost everything else about these races is stable. Attrition is not.
Across 17 races with recorded status, DNFs run from 4 (2023 GEICO 500) to 20 (2022 Coke Zero Sugar 400) — a five-fold spread, standard deviation 4.3 cars. Set that against how little the scoring environment itself moves: the top DK score of a race sits between 64.0 and 81.2 with a standard deviation of 5.8, and the within-race spread of DK scores barely budges at all, 22.6 to 29.1.
Two structural correlations are worth carrying into a simulator. Cautions and DNFs move together (r = 0.52), as you would expect. Lead changes move against them (r = −0.53) — a race that stays green produces 70-plus lead changes and few wrecks, while a caution-broken race freezes the field and collects cars. They are two different races, and which one you get is not knowable on Friday.
Non-finishers per race
Each bar is one race, in date order. Accidents are the darker inner segment; other mechanical retirements make up the rest.
Finishing order at these tracks is noise. The relationship between starting position and fantasy points is not.
Compute both correlations race by race and they behave completely differently. Starting position against finishing position flips sign constantly — from −0.33 to +0.32, median 0.07. It is a coin flip, race to race, exactly as it should be.
Starting position against DK points never once goes negative. In all 18 races it lands between +0.14 and +0.62, median +0.35. That is the most dependable relationship on the whole sheet, and it holds because place differential is mechanical: it does not care whether the racing was clean, it only cares where you started and where you ended.
Two correlations, computed race by race
Each row is one race. Starting position versus DK points stays positive in all 18; starting position versus finishing position crosses zero repeatedly.
Worth noting for the simulator
Attrition and the grid effect are mildly linked, not independent: across the 17 races with status, the DNF count correlates with the start-versus-DK correlation at r = 0.28 and with the winning DK score at r = 0.30. Wreck-heavy races tilt a little further toward deep starters, because the wreck itself is what generates the place differential. It is a second-order effect and 17 races is thin, but the sign is the one you would expect.
The two are treated as one track type by almost everyone, including this database. On attrition they are not close.
Daytona runs 31.8% non-finishers against Talladega’s 19.4%, and the gap is entirely accidents — 28.2% against 16.2%. It shows up hardest at the front of the grid: a car starting in the top ten survives 61.1% of the time at Daytona and 73.8% at Talladega. Start 31st or deeper and it is 72.7% against 88.4%.
The racing differs to match. Talladega produces more lead changes (63.8 to Daytona’s 44.0) with fewer cautions (5.2 to 6.4) — a bigger, faster, less consequential pack. Daytona’s wrecks also cluster: 60% of its accidents happen in the third quarter of the race, against 44% at Talladega, where they spread more evenly across the distance.
For a fantasy sheet this is one number, and it is the survival rate. Everything else about the two tracks — the start-versus-DK correlation, mean score, score spread — is close enough to treat as the same.
Daytona against Talladega
Attrition rates from the 17 races with recorded status; racing metrics from the 17 with recorded caution and lead-change data.
When the accidents happen
Share of each track’s accident retirements by quarter of race distance completed. Daytona 101 accidents, Talladega 50.
The full per-race table. Click a column heading to sort.
Superspeedway race log, 2022–2026
Correlations are computed within each race across the full field. Blank cells are values the database does not carry for that race.
| Race | Field | DNF | Accidents | Cautions | Lead ch. | Leaders | Start→DK r | Start→finish r | Top DK | Winner started |
|---|
Survival explains 45% of DK variance against starting position’s 12%. It cannot sit inside a finishing-position distribution as an error term. Give it a rate, and give Daytona and Talladega different ones — 32% and 19% are not the same number.
67.1% for top-ten starters rising to 80.0% for the back of the grid, and the split is wider at Daytona (61.1% to 72.7%) than Talladega (73.8% to 88.4%). A flat per-car wreck rate will systematically overprice the front row.
14.0 to 19.2 across four grid quarters. Once a car is still running, where it started barely predicts where it ends up — so the scoring difference should come from place differential arithmetic, not from a finishing-position edge the data does not support.
4 to 20 per race, sd 4.3, with cautions correlated at 0.52 and lead changes at −0.53. Drawing a race-level attrition parameter first, then resolving cars against it, reproduces the observed spread; a fixed per-car rate will not.
The start-versus-DK correlation was positive in all 18 races, +0.14 to +0.62. Nothing else in this dataset is that consistent. It is the one relationship worth leaning on hard.