NFL · Week 10 · up next Start Free

← Statistics · updated 8 Nov 2026

How accurate are DFS projections? Every one of ours, graded.

Tens of thousands of DraftKings projections graded against the real score, by projection bucket and position — football and baseball.

Key numbers

  • 28,401DraftKings projections graded against the real score across 121 slates
  • 74%football projections within 5 DK points of the actual 4,997 graded · mean error 3.85
  • 61%baseball projections within 5 DK points of the actual 23,404 graded · mean error 4.99
  • 6.29 → 6.12football: mean projection vs mean actual
  • 6.01 → 5.84baseball: mean projection vs mean actual

Football

Football · projected vs actual, by projection bucket (DraftKings points) mean projectionmean actual Projected 0–5: n=2649, mean projection 1.9, mean actual 1.7, 93% within 5 1.9 1.7 0–5 n=2,649Projected 5–10: n=1159, mean projection 7.2, mean actual 7.3, 64% within 5 7.2 7.3 5–10 n=1,159Projected 10–15: n=587, mean projection 12.1, mean actual 11.5, 46% within 5 12.1 11.5 10–15 n=587Projected 15–20: n=450, mean projection 17.1, mean actual 16.8, 38% within 5 17.1 16.8 15–20 n=450Projected 20–25: n=135, mean projection 21.6, mean actual 21.2, 40% within 5 21.6 21.2 20–25 n=135 4,997 graded · 12 slates · 9 Sep 2026 → 8 Nov 2026 · ribby.ai
Each pair is one projection bucket: what we projected on average (ink) and what those players actually scored (green). A bucket where the two bars match is a calibrated bucket.
Football calibration by projection bucket
ProjectednMean projMean actualWithin 5Beat projection
0–52,6491.91.793%24%
5–101,1597.27.364%42%
10–1558712.111.546%41%
15–2045017.116.838%44%
20–2513521.621.240%43%
By position (error = actual − projected)
PositionnMean projMean actualError sd
WR1,86865.76
RB1,2166.36.35.9
TE1,141444.6
QB43813.913.47.2
DST3345.55.54.7

Baseball

Baseball · projected vs actual, by projection bucket (DraftKings points) mean projectionmean actual Projected 0–5: n=7481, mean projection 3.5, mean actual 3.3, 84% within 5 3.5 3.3 0–5 n=7,481Projected 5–10: n=14294, mean projection 6.5, mean actual 6.5, 53% within 5 6.5 6.5 5–10 n=14,294Projected 10–15: n=1175, mean projection 12, mean actual 10, 33% within 5 12 10 10–15 n=1,175Projected 15–20: n=430, mean projection 16.8, mean actual 16.2, 34% within 5 16.8 16.2 15–20 n=430 23,404 graded · 109 slates · 18 Jun 2026 → 19 Oct 2026 · ribby.ai
Each pair is one projection bucket: what we projected on average (ink) and what those players actually scored (green). A bucket where the two bars match is a calibrated bucket.
Baseball calibration by projection bucket
ProjectednMean projMean actualWithin 5Beat projection
0–57,4813.53.384%43%
5–1014,2946.56.553%38%
10–151,175121033%37%
15–2043016.816.234%48%
By position (error = actual − projected)
PositionnMean projMean actualError sd
OF6,0716.16.16.8
RP5,14442.94.7
C2,1055.55.66.6
SP1,78911.2119.6
1B1,6796.66.87.3
SS1,4296.46.77
3B1,3406.26.77.2
2B1,3286.15.96.4
2B/3B4165.55.16
2B/SS3645.34.65.5

Questions this page answers

How accurate are DFS projections?

On 28,401 graded DraftKings projections (121 slates through 8 Nov 2026), 74% of football projections and 61% of baseball projections landed within 5 DraftKings points of the actual score. The spread is wide by design: a single game is noisy, and the useful question is calibration — whether players projected at 15 average 15 — which the chart shows bucket by bucket.

What does "graded" mean here?

A projection is graded when the actual DraftKings score for that player on that slate is written back next to it. Every projection the product logs before a slate locks is graded the next day; nothing is dropped after the fact.

Are projections biased high or low?

Football projections average 6.29 against 6.12 actual; baseball 6.01 against 5.84. The high buckets run hot (a 15-point projection is a ceiling more often than a midpoint), which is why Ribby's DFS desk ranks on separation, not on the raw number.

How this was measured

Every DraftKings projection the product writes before a slate locks is logged with the player, the slate and the projection; once the slate's scores are final, the actual DraftKings points are written back onto the same row. This page reads those rows live (cached six hours) and groups them by projection bucket and by position. Rows without an actual score — a scratched player, a postponed game — are excluded, not imputed.

"Within 5" is the share of graded rows whose actual score sits within 5 DraftKings points of the projection. "Beat projection" is the share that scored at or above it. Buckets under 30 rows are not drawn. Mean error is the mean absolute difference, reported for transparency and never used to tune the model: the product's own law is that a projection earns weight on lineup results, not on per-player error.

Data: our own projection log, football and baseball. Football scores are DraftKings NFL Classic; baseball scores are DraftKings MLB Classic. The window starts when the log started (9 Sep 2026 football, 18 Jun 2026 baseball) and ends on the newest graded slate.

Cite this

These numbers are free to quote with a link (CC BY 4.0). Suggested: Ribby, “How Accurate Are DFS Projections? The Graded Ledger”, ribby.ai, updated 8 Nov 2026. https://ribby.ai/stats/dfs-projection-accuracy

https://ribby.ai/stats/dfs-projection-accuracy

The numbers, with a pulse.

Ribby files two fact-checked columns a day off these tables — read the Wire — and answers about YOUR roster inside the app.

Start Free
Live demo Start Free — No Card