Home / Insights / MLB Model vs Market
Model vs Market · Week in Review

MLB projections vs. the market

Week of Jul 13 to 19, 2026 (2026-W29) · model clean_v2 · generated Aug 15, 2026

A report, not advice. Every projection below was logged before first pitch in our immutable record and graded after the game. nothing is added or removed later. We publish no pick, lean or edge; this page reports how our numbers did, including the misses. The market's line is shown for context only.

How we did

Graded MLB games, Jul 13 to 19, 2026 · reports, not picks
Games graded
43
2026-W29
Moneyline record
29-14
67.4% higher-prob side
our >50% side won
Moneyline Brier
0.240
better than 0.250 coin flip
lower is better
Totals avg miss
3.82
runs |proj − actual|
43 games
Totals vs market
n/a
line capture just began
0 lined games this week

We can't show model-vs-market on totals yet: opening totals weren't captured on this week's graded games, so graded-and-lined games = 0. A model-vs-market number only counts games where we have both sides. In our held-out research backtest the model trails the market by 0.06 runs (3.52 vs 3.46); the live number turns on as lined games accrue.

Where the week's results came from

A fixed set of splits, published every week whether it flatters us or not · slices under n=5 show their count and no rate
Moneyline splitRecordn
Model's side was the home team21–1233
Model's side was the road team8–210
Model's side was the market favorite14–519
Model's side was the market underdog9–514
Coin-flip calls (50-55%)19–928
Moderate calls (55-65%)10–515
Strongest calls (65%+)not scored: below the n=5 floor (n=0)
Games with a total under 8.5not scored: no opening total captured for these games (n=0)
Games with a total 8.5-9.5not scored: no opening total captured for these games (n=0)
Games with a total above 9.5not scored: no opening total captured for these games (n=0)
Totals splitOur avg missMarket'sn
Projection sat above the linenot scored: no opening total captured for these games (n=0)
Projection sat below the linenot scored: no opening total captured for these games (n=0)
Lines under 8.5not scored: no opening total captured for these games (n=0)
Lines 8.5-9.5not scored: no opening total captured for these games (n=0)
Lines above 9.5not scored: no opening total captured for these games (n=0)
Confidence bucketWe saidIt happenedn
Coin-flip calls (50-55%)53.0%67.9%28
Moderate calls (55-65%)57.5%66.7%15
Strongest calls (65%+)not scored: below the n=5 floor (n=0)

League base rates for the same games: home teams went 23–20 (n=43); market favorites went 19–14 (57.6%, n=33). Every model result above sits inside that backdrop.

The three largest totals misses account for 21.6% of the week's total absolute totals error (35.4 of 164.3 runs across 43 games).

Strongest call that landed

DET @ LAA (Jul 18): we had the away side at 64%; the away side won.

Final DET 7, LAA 0; opening price on that side -192; the archived probables started.
Strongest call that missed

DET @ LAA (Jul 19): we had the away side at 64%; the home side won. It stays in the record.

Final DET 2, LAA 3; opening price on that side -155; the archived probables started.

One week is a small sample. These are counts of where the results came from, not conclusions about skill; the season-long record is the number that matters.

The biggest misses

We show them on purpose · totals, Jul 13 to 19, 2026

The point of a public track record is that the misses stay in. Our three largest totals misses this week:

GameProjectedActualOff by
WSH @ OAK11.22715.8
SD @ KC9.72111.3
CWS @ TOR9.418.4

Every graded game

Projected total vs. actual, and whether our higher-probability moneyline side won · 43 games
GameProj totalActualMissHome win%ML side
NYM @ PHI9.454.50.56Miss
MIA @ MIL8.235.20.56Hit
TEX @ ATL8.9167.10.56Hit
SF @ SEA8.271.10.56Miss
STL @ ARI9.190.10.56Miss
TB @ BOS8.880.80.54Hit
SD @ KC9.9133.10.54Hit
MIN @ CHC9.572.50.54Miss
CWS @ TOR8.7167.30.54Miss
BAL @ HOU8.954.00.49Hit
DET @ LAA8.635.60.49Hit
WSH @ OAK11.22715.80.49Hit
CIN @ COL10.891.80.49Hit
LAD @ NYY9.136.10.54Miss
TEX @ ATL9.0134.00.54Miss
MIN @ CHC8.980.90.54Hit
PIT @ CLE8.880.80.54Hit
CIN @ COL11.8131.20.54Hit
CWS @ TOR9.418.40.56Hit
MIA @ MIL8.2145.80.56Hit
STL @ ARI9.181.10.54Hit
BAL @ HOU9.363.30.49Hit
TB @ BOS8.2134.80.49Miss
NYM @ PHI8.871.80.56Hit
DET @ LAA8.271.20.35Hit
SD @ KC9.472.40.54Hit
WSH @ OAK11.3164.70.54Hit
SF @ SEA8.271.20.59Hit
SD @ KC9.72111.30.56Miss
STL @ ARI9.4155.60.56Hit
TB @ BOS8.371.40.56Hit
MIA @ MIL8.744.70.54Hit
NYM @ PHI9.272.20.54Miss
BAL @ HOU8.871.80.54Miss
CWS @ TOR8.535.50.54Miss
WSH @ OAK11.174.10.49Hit
SF @ SEA8.290.80.59Hit
CIN @ COL10.6154.40.49Hit
MIN @ CHC8.8112.20.54Hit
LAD @ NYY8.935.90.54Hit
TEX @ ATL8.7134.30.54Hit
PIT @ CLE8.080.00.54Miss
DET @ LAA8.853.90.36Miss

What the market says

A market we deliberately don't model: the market's own record
MLB run line
42.8% of −1.5 favorites cover

Across 11,972 final regular-season games with a standard run line, the moneyline favorite laying −1.5 covered 42.8% of the time and won outright 58.3%. We publish no run-line projection, because the market prices this one efficiently, so this is the market's record, not ours.

Market benchmark, not a Nebula projection.

One number: Mean absolute error (MAE)

A rotating look at how we grade ourselves

Our totals number is graded on average miss: |projected runs − actual runs|. In our held-out research backtest the model's MAE is 3.52 runs versus the market's 3.46: we trail by 0.06. We say so plainly.

Keep reading

The full record and the live board