Report card

Week 10 report card: how GridIron's own projections did

Quick answer: GridIron's Week 10 projections missed by an average of 6.0 PPR points across 189 graded players. On the 187 of them with earlier games this season we missed by 6.0, better than the 6.4 a simple last-4-games average would have missed by. 25% of projections landed within 3 points of the actual score and 48% within 5.

Players graded
189
Average miss
6.0 pts
Within 5 pts
48%
Season average miss
5.6 pts

How did we grade Week 10?

We publish our own projections every week, so we grade them every week too, in public. Week 10 is in the books: 189 players graded, an average miss of 6.0 points, and the hits and misses are below, the ugly ones included.

Who is graded: every QB, RB, WR and TE we projected as available for at least 5 PPR points who recorded a stat line in Week 10. The projection graded is the last version we published before that player's kickoff. Players we projected who didn't play are listed separately and are not in the average.

The miss is the gap between projected and actual PPR points, ignoring direction, and the average miss is the mean of those gaps. The baseline is the player's average over his last 4 games played before Week 10: the number you'd get with no model at all. A projection that can't beat it isn't earning its keep.

Projections come from /projections/{week}?source=gridiron and actual points from the weekly stats on /players, both in the GridIron API.

How accurate were our Week 10 projections by position?

Week 10 GridIron projections vs actual PPR points, by position
PositionPlayersOur average missLast-4 average missWithin 5 pts
QB316.96.935%
RB506.66.848%
WR715.96.649%
TE374.74.754%
All1896.06.448%

The last-4 column covers the 187 players with at least one earlier game this season; on those same players our average miss was 6.0.

TEs were the easiest position to project in Week 10 (average miss 4.7) and QBs the hardest (6.9). Positions score on different scales, so the fairer read is each row against its own last-4 column.

What were our biggest hits in Week 10?

Among the 36 players we projected highest in Week 10, these landed closest to the number. Getting a 20-point player within a point or two is the hard part; getting a 5-point player close is mostly arithmetic.

Closest projections among our top 36, Week 10
PlayerPosTeamProjectedActualDiff
CeeDee LambWRDAL21.820.2-1.6
Drake LondonWRATL15.316.9+1.6
Josh AllenQBBUF21.323.3+2.0
Derrick HenryRBBAL15.919.0+3.1
Bijan RobinsonRBATL21.117.7-3.4
James CookRBBUF18.321.7+3.4

What were our biggest misses in Week 10?

A miss is a miss, so here are the largest in each direction. Where the snap counts show an obvious reason (a player who left early, or one whose role suddenly grew) we note it; otherwise we don't guess.

Biggest over-projections, Week 10 (scored less than projected)
PlayerPosTeamProjectedActualDiffContext
Drake MayeQBNE19.03.0-16.0Played 17% of snaps; his average before this week was 96%.
Jalen CokerWRCAR12.10.3-11.8—
Kyren WilliamsRBLAR16.85.2-11.6—
Chris OlaveWRNO16.86.0-10.8—
Brock BowersTELV14.63.8-10.8Played 19% of snaps; his average before this week was 86%.
Biggest under-projections, Week 10 (scored more than projected)
PlayerPosTeamProjectedActualDiffContext
George PickensWRDAL13.243.5+30.3—
Christian McCaffreyRBSF20.646.3+25.7—
Jahmyr GibbsRBDET22.944.2+21.4—
Emmett JohnsonRBKC15.335.1+19.8—
Patrick MahomesQBKC16.635.4+18.8—

Across every graded player, actual points averaged 0.2 above the projection, so the week ran close to even overall.

Who did we project who didn't play?

We projected these players as available for at least 5 points and they recorded no stats in Week 10: Shedeur Sanders (QB, CLE, 13.5), Carson Wentz (QB, MIN, 12.3), Gardner Minshew (QB, ARI, 11.0), Joe Fagnano (QB, BAL, 9.0), Marcus Mariota (QB, WAS, 7.6), Jordan Addison (WR, MIN, 7.6), Spencer Rattler (QB, NO, 6.9), Sam Howell (QB, DAL, 6.5), Aidan O'Connell (QB, LV, 6.3), Riley Leonard (QB, IND, 5.9), Phil Mafah (RB, NYG, 5.8), Tua Tagovailoa (QB, ATL, 5.8) and 2 more. They aren't in the average miss above, but each one was a projection a lineup may have trusted, so we list them. Projections are published before game-day inactives, so a late scratch lands here; the injury status at kickoff is the check that catches it.

How have our projections done in 2026 so far?

Every 2026 week with GridIron projections, graded the same way. "Ours, same players" is our average miss on just the players who also have a last-4 baseline, so the two columns beside it compare like with like.

GridIron projections, average miss by week (2026)
WeekPlayersOur average missOurs, same playersLast-4 average missWithin 5 pts
Week 91875.25.25.656%
Week 101896.06.06.448%
Season3765.65.66.052%

We've beaten the last-4 baseline in 2 of 2 graded weeks this season, and over the season our average miss is 5.6 against 6.0 on the same players.

What are we watching next week?

  • QBs were the hardest call this week (6.9 average miss). If the same position tops the list again next week, that's a pattern, not noise.
  • We beat the baseline this week (6.0 against 6.4). We came closer than the last-4 average for 52% of the players graded.
  • 14 projected players didn't play. Projections are built before inactives, so re-check statuses before each kickoff.
  • One week is a small sample. A single game swings on a touchdown or a long play; the season table is the better guide to how far to trust the numbers.

How can I grade GridIron projections myself?

Every number on this page is reproducible from the API. Fetch our projections for the week, sweep /players for the actual points, and compare:

import requests

API = "https://api.gridirondata.com/api/v1"
H = {"x-api-key": "YOUR_KEY"}
SEASON, WEEK = "2026", 10

# Our projections for the week (one call).
rows = requests.get(f"{API}/projections/{WEEK}", headers=H,
                    params={"source": "gridiron", "scoring": "ppr", "season": SEASON}).json()["projections"]

# Actual PPR points: one paginated sweep of /players.
players, cursor = {}, None
while True:
    params = {"limit": 100, **({"cursor": cursor} if cursor else {})}
    body = requests.get(f"{API}/players", headers=H, params=params).json()
    players.update({p["player_id"]: p for p in body["players"]})
    cursor = body.get("next_cursor")
    if not cursor:
        break

misses = []
for r in rows:
    gp = r["gridiron_projection"] or {}
    stats = players.get(r["player_id"], {}).get("seasons", {}).get(SEASON, {}).get("weekly_stats", {})
    row = stats.get(str(WEEK))
    if gp.get("available") is False or gp.get("projected_points") is None or not row:
        continue
    misses.append(abs(row["fantasy_points_ppr"] - gp["projected_points"]))

print(f"{len(misses)} players, average miss {sum(misses) / len(misses):.1f} PPR")

That's one call for the projections and about 18 for the player sweep. A free key gets the current week's projections; earlier weeks and the weekly stats take Basic ($5/month). Our projections may be shown to your own users under the Pro + Commercial license. See the GridIron projections guide and how they're built.

Frequently asked questions

What is a good average miss for weekly fantasy projections?

Weekly fantasy scoring is noisy, so even good projections miss by several points on average for players projected for 5 or more PPR. The useful test isn't the raw number but whether a projection beats a simple baseline on the same players, week after week.

Why grade against a last-4-games average?

Because it's what you'd use with no model at all: each player's average over his last four games played. If a projection can't beat that, it isn't adding anything. It's also easy to check yourself.

Which projections are graded?

GridIron's own projections only, the last version published before each player's kickoff, for QBs, RBs, WRs and TEs projected as available for at least 5 PPR points who played. Players who didn't play are listed separately and left out of the average.

How do I get GridIron projections?

From /projections/{week}?source=gridiron (or the gridiron_projection object on every /projections row). A free key gets the current week; Basic gets every week; Pro adds the per-stat lines and game context.