Calibration

How accurate are our projections?

Most fantasy services tell you their projections are good. We show you.

Below is the actual measured accuracy of our projections on a completed week, side-by-side with what FantasyPros, ESPN, and Sleeper said for the same players. These are the projections exactly as a subscriber receives them, final adjustments included — not an internal version that never ships.

In season

Season 2025, Week 17

The projections we published to subscribers for this week, scored after the games against what actually happened. Not a re-run and not a re-fit — these are the numbers that were on the board before kickoff.

The board we served, 2025 Week 17

Player-week-stat outcomes measured

2795

interval accuracy is measured on 2251 of these — the rest are projections we publish without a range, either because the stat is too rare to calibrate one or because the range would be too narrow to state

across 382 of 910 players projected

50% prediction interval

58.2% contained the actual outcome

Target: 50.0% · +8.2 pp

80% prediction interval

82.6% contained the actual outcome

Target: 80.0% · +2.6 pp

90% prediction interval

91.1% contained the actual outcome

Target: 90.0% · +1.1 pp

Reading the pp figure: it is the measured rate minus the target, so a positive figure means the intervals contained the outcome more often than we promised and a negative one means they contained it less often.

These measure the per-stat ranges as published. A player's fantasy-point floor and ceiling are built from them and then adjusted for whether he was expected to play, and that adjustment is not scored here.

Backtest

Season 2025, Week 5

A completed week, projected using only what was known before it was played. Not the week you are currently being projected for — that one has no results to score against yet.

Walk-forward backtest, 2025 Week 5

Player-week-stat outcomes measured

1578

interval accuracy below is measured on 1542 of these — the rest are projections we publish without a range, either because the stat is too rare to calibrate one or because the range would be too narrow to state

across 277 of 308 players projected

50% prediction interval

57.5% contained the actual outcome

Target: 50.0% · +7.5 pp

80% prediction interval

81.6% contained the actual outcome

Target: 80.0% · +1.6 pp

90% prediction interval

88.8% contained the actual outcome

Target: 90.0% · -1.2 pp

A "50% prediction interval" is the range we publish saying "the actual outcome should land inside this band 50% of the time." If our intervals are well-calibrated, the empirical rate matches the target. The same applies to the 80% and 90% intervals.

Why this matters. Point projections (a single number per player) have no calibration story to tell. Without a width, you can't tell a confident projection from a hopeful one. We publish full prediction intervals because they let you act on the projection — start a player when the floor is high, fade him when the ceiling is what you're chasing.

By stat

Empirical coverage, per stat

Cells are colored green when the empirical rate is within 2 percentage points of the target, yellow within 5 pp, red beyond. Larger samples (high "n") are more trustworthy than small ones.

Stat	n	Mean actual	Mean projected	50% PI	80% PI	90% PI
Receptions	198	2.66	2.35	53%	81%	89%
Receiving TDs	198	0.22	0.17	—	—	—
Receiving yards	198	28.98	23.80	61%	82%	89%
Receiving first downs	151	1.63	1.76	52%	84%	90%
Fumbles lost	142	0.08	0.07	—	—	—
Rushing TDs	89	0.24	0.25	—	—	—
Rushing yards	89	29.58	28.05	54%	80%	91%
Rushing first downs	84	1.58	2.03	44%	83%	93%
Interceptions	30	0.50	0.57	63%	87%	90%
Passing TDs	30	1.37	1.14	43%	77%	77%
Passing yards	30	199.67	177.12	70%	93%	97%
Passing first downs	27	10.81	9.53	63%	89%	93%
Team defense — blocked kicks	25	0.08	0.10	—	—	—
Team defense — fumble rec.	25	0.68	1.03	—	—	—
Team defense — INTs	25	0.64	0.75	64%	88%	88%
Team defense — pts allowed	25	24.56	23.22	44%	80%	96%
Team defense — sacks	25	2.20	2.27	52%	76%	92%
Team defense — safeties	25	0.04	0.03	—	—	—
Team defense — TDs	25	0.08	0.14	76%	88%	92%
FGs made (≤19 yd)	23	0.00	0.00	—	—	—
FGs made (20-29 yd)	23	0.39	0.37	65%	65%	78%
FGs made (30-39 yd)	23	0.52	0.53	52%	87%	96%
FGs made (40-49 yd)	23	0.22	0.46	61%	96%	100%
FGs made (50+ yd)	23	0.35	0.34	—	—	—
Extra points made	22	2.73	2.54	64%	91%	91%

Receptions

n = 198

81% 80% PI

50% PI

53%

90% PI

89%

Mean actual

2.66

Mean projected

2.35
Receiving TDs

n = 198

— 80% PI

50% PI

—

90% PI

—

Mean actual

0.22

Mean projected

0.17
Receiving yards

n = 198

82% 80% PI

50% PI

61%

90% PI

89%

Mean actual

28.98

Mean projected

23.80
Receiving first downs

n = 151

84% 80% PI

50% PI

52%

90% PI

90%

Mean actual

1.63

Mean projected

1.76
Fumbles lost

n = 142

— 80% PI

50% PI

—

90% PI

—

Mean actual

0.08

Mean projected

0.07
Rushing TDs

n = 89

— 80% PI

50% PI

—

90% PI

—

Mean actual

0.24

Mean projected

0.25
Rushing yards

n = 89

80% 80% PI

50% PI

54%

90% PI

91%

Mean actual

29.58

Mean projected

28.05
Rushing first downs

n = 84

83% 80% PI

50% PI

44%

90% PI

93%

Mean actual

1.58

Mean projected

2.03
Interceptions

n = 30

87% 80% PI

50% PI

63%

90% PI

90%

Mean actual

0.50

Mean projected

0.57
Passing TDs

n = 30

77% 80% PI

50% PI

43%

90% PI

77%

Mean actual

1.37

Mean projected

1.14
Passing yards

n = 30

93% 80% PI

50% PI

70%

90% PI

97%

Mean actual

199.67

Mean projected

177.12
Passing first downs

n = 27

89% 80% PI

50% PI

63%

90% PI

93%

Mean actual

10.81

Mean projected

9.53
Team defense — blocked kicks

n = 25

— 80% PI

50% PI

—

90% PI

—

Mean actual

0.08

Mean projected

0.10
Team defense — fumble rec.

n = 25

— 80% PI

50% PI

—

90% PI

—

Mean actual

0.68

Mean projected

1.03
Team defense — INTs

n = 25

88% 80% PI

50% PI

64%

90% PI

88%

Mean actual

0.64

Mean projected

0.75
Team defense — pts allowed

n = 25

80% 80% PI

50% PI

44%

90% PI

96%

Mean actual

24.56

Mean projected

23.22
Team defense — sacks

n = 25

76% 80% PI

50% PI

52%

90% PI

92%

Mean actual

2.20

Mean projected

2.27
Team defense — safeties

n = 25

— 80% PI

50% PI

—

90% PI

—

Mean actual

0.04

Mean projected

0.03
Team defense — TDs

n = 25

88% 80% PI

50% PI

76%

90% PI

92%

Mean actual

0.08

Mean projected

0.14
FGs made (≤19 yd)

n = 23

— 80% PI

50% PI

—

90% PI

—

Mean actual

0.00

Mean projected

0.00
FGs made (20-29 yd)

n = 23

65% 80% PI

50% PI

65%

90% PI

78%

Mean actual

0.39

Mean projected

0.37
FGs made (30-39 yd)

n = 23

87% 80% PI

50% PI

52%

90% PI

96%

Mean actual

0.52

Mean projected

0.53
FGs made (40-49 yd)

n = 23

96% 80% PI

50% PI

61%

90% PI

100%

Mean actual

0.22

Mean projected

0.46
FGs made (50+ yd)

n = 23

— 80% PI

50% PI

—

90% PI

—

Mean actual

0.35

Mean projected

0.34
Extra points made

n = 22

91% 80% PI

50% PI

64%

90% PI

91%

Mean actual

2.73

Mean projected

2.54

Targets: 50.0% / 80.0% / 90.0%. Colors: green ≤ 2pp off, yellow ≤ 5pp, red beyond. Only stats with n ≥ 5 are shown.

Interval coverage is withheld for 8 of the 25 stats below (FGs made (50+ yd), FGs made (≤19 yd), Fumbles lost, Receiving TDs, Rushing TDs, Team defense — blocked kicks, Team defense — fumble rec., Team defense — safeties), and the three PI columns read “—” on those rows. A coverage figure here is a statement about bands that sit inside one another, and for these stats some of the projections we published state percentiles that are not in order — so the figure would not be a measurement of what this page describes. Every other column on those rows is measured over exactly the same outcomes as the rest of the table.

Head to head

Mean absolute error vs. other public projections

For every (player, stat) where both we and the competitor projected — and where the actual outcome is known — we compute the absolute miss (|projection − actual|) and average it. Lower is better. Each cell shows our error and theirs on that competitor's own rows, because a competitor publishes fewer players than we do and averaging over different sets of players is not a comparison. n is the shared row count.

Stat	Rows we project	vs FantasyPros	vs ESPN	vs Sleeper
Receptions	198 MAE 1.27	1.33 vs 1.60 better · n=91	1.32 vs 1.39 better · n=182	1.33 vs 1.36 better · n=181
Receiving TDs	198 MAE 0.29	0.37 vs 0.39 better · n=69	0.30 vs 0.30 even · n=182	0.31 vs 0.31 even · n=179
Receiving yards	198 MAE 17.42	19.88 vs 20.24 better · n=91	18.42 vs 18.53 better · n=182	18.49 vs 18.20 worse · n=181
Receiving first downs	151 MAE 0.87	—	—	—
Fumbles lost	142 MAE 0.14	0.22 vs 0.25 better · n=25	0.15 vs 0.15 even · n=126	0.16 vs 0.16 even · n=107
Rushing TDs	89 MAE 0.34	0.41 vs 0.41 even · n=30	0.37 vs 0.36 worse · n=74	0.37 vs 0.39 better · n=75
Rushing yards	89 MAE 18.54	23.38 vs 26.59 better · n=36	20.43 vs 20.73 better · n=74	20.07 vs 22.16 better · n=76
Rushing first downs	84 MAE 1.11	—	—	—
Interceptions	30 MAE 0.68	0.61 vs 0.66 better · n=12	0.78 vs 0.76 worse · n=24	0.78 vs 0.80 better · n=24
Passing TDs	30 MAE 0.95	1.21 vs 1.14 worse · n=12	1.08 vs 1.10 better · n=24	1.08 vs 1.06 worse · n=24
Passing yards	30 MAE 45.26	48.81 vs 63.12 better · n=12	45.90 vs 40.42 worse · n=24	45.90 vs 44.54 worse · n=24
Passing first downs	27 MAE 2.82	—	—	—
Team defense — blocked kicks	25 MAE 0.17	—	—	—
Team defense — fumble rec.	25 MAE 0.67	—	—	—
Team defense — INTs	25 MAE 0.75	—	—	—
Team defense — pts allowed	25 MAE 7.51	—	—	—
Team defense — sacks	25 MAE 1.40	—	—	—
Team defense — safeties	25 MAE 0.07	—	—	—
Team defense — TDs	25 MAE 0.20	—	—	—
FGs made (≤19 yd)	23 MAE 0.00	—	—	—
FGs made (20-29 yd)	23 MAE 0.57	—	—	0.57 vs 0.55 worse · n=23
FGs made (30-39 yd)	23 MAE 0.61	—	—	0.61 vs 0.59 worse · n=23
FGs made (40-49 yd)	23 MAE 0.48	—	—	0.48 vs 0.51 better · n=23
FGs made (50+ yd)	23 MAE 0.48	—	—	0.48 vs 0.46 worse · n=23
Extra points made	22 MAE 1.00	1.17 vs 1.13 worse · n=12	—	1.00 vs 0.95 worse · n=22

Receptions

198 rows we project · MAE 1.27

vs FantasyPros

1.33 vs 1.60
better · n=91

vs ESPN

1.32 vs 1.39
better · n=182

vs Sleeper

1.33 vs 1.36
better · n=181
Receiving TDs

198 rows we project · MAE 0.29

vs FantasyPros

0.37 vs 0.39
better · n=69

vs ESPN

0.30 vs 0.30
even · n=182

vs Sleeper

0.31 vs 0.31
even · n=179
Receiving yards

198 rows we project · MAE 17.42

vs FantasyPros

19.88 vs 20.24
better · n=91

vs ESPN

18.42 vs 18.53
better · n=182

vs Sleeper

18.49 vs 18.20
worse · n=181
Receiving first downs

151 rows we project · MAE 0.87

vs FantasyPros

—

vs ESPN

—

vs Sleeper

—
Fumbles lost

142 rows we project · MAE 0.14

vs FantasyPros

0.22 vs 0.25
better · n=25

vs ESPN

0.15 vs 0.15
even · n=126

vs Sleeper

0.16 vs 0.16
even · n=107
Rushing TDs

89 rows we project · MAE 0.34

vs FantasyPros

0.41 vs 0.41
even · n=30

vs ESPN

0.37 vs 0.36
worse · n=74

vs Sleeper

0.37 vs 0.39
better · n=75
Rushing yards

89 rows we project · MAE 18.54

vs FantasyPros

23.38 vs 26.59
better · n=36

vs ESPN

20.43 vs 20.73
better · n=74

vs Sleeper

20.07 vs 22.16
better · n=76
Rushing first downs

84 rows we project · MAE 1.11

vs FantasyPros

—

vs ESPN

—

vs Sleeper

—
Interceptions

30 rows we project · MAE 0.68

vs FantasyPros

0.61 vs 0.66
better · n=12

vs ESPN

0.78 vs 0.76
worse · n=24

vs Sleeper

0.78 vs 0.80
better · n=24
Passing TDs

30 rows we project · MAE 0.95

vs FantasyPros

1.21 vs 1.14
worse · n=12

vs ESPN

1.08 vs 1.10
better · n=24

vs Sleeper

1.08 vs 1.06
worse · n=24
Passing yards

30 rows we project · MAE 45.26

vs FantasyPros

48.81 vs 63.12
better · n=12

vs ESPN

45.90 vs 40.42
worse · n=24

vs Sleeper

45.90 vs 44.54
worse · n=24
Passing first downs

27 rows we project · MAE 2.82

vs FantasyPros

—

vs ESPN

—

vs Sleeper

—
Team defense — blocked kicks

25 rows we project · MAE 0.17

vs FantasyPros

—

vs ESPN

—

vs Sleeper

—
Team defense — fumble rec.

25 rows we project · MAE 0.67

vs FantasyPros

—

vs ESPN

—

vs Sleeper

—
Team defense — INTs

25 rows we project · MAE 0.75

vs FantasyPros

—

vs ESPN

—

vs Sleeper

—
Team defense — pts allowed

25 rows we project · MAE 7.51

vs FantasyPros

—

vs ESPN

—

vs Sleeper

—
Team defense — sacks

25 rows we project · MAE 1.40

vs FantasyPros

—

vs ESPN

—

vs Sleeper

—
Team defense — safeties

25 rows we project · MAE 0.07

vs FantasyPros

—

vs ESPN

—

vs Sleeper

—
Team defense — TDs

25 rows we project · MAE 0.20

vs FantasyPros

—

vs ESPN

—

vs Sleeper

—
FGs made (≤19 yd)

23 rows we project · MAE 0.00

vs FantasyPros

—

vs ESPN

—

vs Sleeper

—
FGs made (20-29 yd)

23 rows we project · MAE 0.57

vs FantasyPros

—

vs ESPN

—

vs Sleeper

0.57 vs 0.55
worse · n=23
FGs made (30-39 yd)

23 rows we project · MAE 0.61

vs FantasyPros

—

vs ESPN

—

vs Sleeper

0.61 vs 0.59
worse · n=23
FGs made (40-49 yd)

23 rows we project · MAE 0.48

vs FantasyPros

—

vs ESPN

—

vs Sleeper

0.48 vs 0.51
better · n=23
FGs made (50+ yd)

23 rows we project · MAE 0.48

vs FantasyPros

—

vs ESPN

—

vs Sleeper

0.48 vs 0.46
worse · n=23
Extra points made

22 rows we project · MAE 1.00

vs FantasyPros

1.17 vs 1.13
worse · n=12

vs ESPN

—

vs Sleeper

1.00 vs 0.95
worse · n=22

Each pairing is computed only over the rows where that competitor published a projection — a small "n" means narrower coverage on that stat, and our own number is recomputed on exactly those rows so the two are like for like. The left column is our error across every row we project, which is a wider and easier population; it is reported for context and is deliberately not set against a competitor.

Some sources store a zero where they mean "no projection for this player", which is not the same claim as projecting zero — scoring it as a forecast would let a blank count as a miss. We drop any value too small to be a projection on that stat's own scale, and here that withheld 162 from FantasyPros and 19 from Sleeper. Those rows are excluded from both sides of the pairing, ours included.

Boom rate

How often do the booms we forecast actually happen?

A "boom" is a big week for the position — a top-tier weekly fantasy score (full PPR). We publish a boom chance for every player we project at QB, RB, WR and TE. At K and DEF we publish none — we state a boom bar only where a second, independent source has checked it, so every K and DEF row on the rankings board carries a dash under Boom. The rates below are measured at every position we have a boom band for, K and DEF included. Here's how those forecasts held up on the players whose results we have: if we say a group of players has a 30% chance to boom, about 30% of them should.

We forecast (average boom chance)

10.6%

Actually boomed

13.8%

across 275 player-weeks, 33 not yet scored

Position	Boom is	n	We forecast	Actually boomed	Gap boomed − forecast
QB	25+ pts	30	9%	13%	+4 pp
RB	20+ pts	59	11%	8%	-3 pp
WR	18+ pts	85	11%	14%	+3 pp
TE	13+ pts	52	13%	21%	+8 pp
K	14.5+ pts	24	6%	12%	+6 pp
DEF	12.5+ pts	25	12%	12%	+0 pp

QB

Boom is 25+ pts · n = 30

+4 pp gap

We forecast

9%

Actually boomed

13%
RB

Boom is 20+ pts · n = 59

-3 pp gap

We forecast

11%

Actually boomed

8%
WR

Boom is 18+ pts · n = 85

+3 pp gap

We forecast

11%

Actually boomed

14%
TE

Boom is 13+ pts · n = 52

+8 pp gap

We forecast

13%

Actually boomed

21%
K

Boom is 14.5+ pts · n = 24

+6 pp gap

We forecast

6%

Actually boomed

12%
DEF

Boom is 12.5+ pts · n = 25

+0 pp gap

We forecast

12%

Actually boomed

12%

Reading the Gap: it is what actually boomed minus what we forecast, so a positive gap means we under-forecast — it happened more often than we said — and a negative gap means we over-forecast.

Preliminary: based on a single backtest week, so the per-position rates are still thin and will sharpen as more weeks are measured — this page will show whether the calibration holds. Boom thresholds are an early cut and may be retuned.

How we measure

Coverage and MAE are computed from the same data file we use internally — no cherry-picking. Every player who appeared in this week's actuals and also had a projection from us (or from a competitor) is included in the tally for the stats they recorded. Stats with fewer than 5 observed outcomes are dropped; results below that threshold aren't trustworthy.

We measure the projections we ship. Our numbers pass through a final set of adjustments before they reach you, and the run scored above has been through exactly those — so what you read here is the accuracy of the product, not of an earlier step in producing it. When those adjustments change, this page moves with them.

We rely on calibrated uncertainty — projections that come with honest prediction intervals, not just point estimates. As more weeks land, this page will accumulate them, and you'll be able to see whether the calibration holds up across the season. If it doesn't, you'll see that here too.

Where we get our college numbers

A rookie has no NFL record to project from, so his first-season numbers lean on what he did in college. Our college numbers come from CollegeFootballData.com, a freely accessible, independently run college-football data project worth knowing about if you follow the draft.