The accuracy record

Projections are only worth what they get right. This page is the machine's report card: our projections against every public system we can score on the same players with the same yardstick, and against the simple forecasts anyone could make without a model. Wins and losses both. It updates as seasons close.

Three headline results: lower error than all nine public systems on 2026 hitters so far, including ZiPS, Steamer, ATC, and THE BAT. Lower pooled error than PECOTA, ZiPS, and Steamer on pitchers across 2021 to 2025, winning 12 of 14 scored system-seasons. And in the long test, projecting each past season using only what was knowable beforehand, our hitter projections beat the standard Marcel forecast in 46 of 46 seasons since 1980, and our pitcher projections in 15 of 15 since 2011.

Every table uses RMSE, root-mean-square error: how far the projection missed what the player actually did, on the projection's own scale. Lower is better. Comparisons are always on the same players, weighted by playing time. Debiased columns remove each system's overall league-level miss first.

2026 hitters vs the field

Preseason projections frozen before Opening Day, scored against the season so far (through 2026-08-21). Final scoring comes in October.

We lead all nine systems on raw error, and the lead is statistically decisive against six of them. Against the Depth Charts, Steamer, and Steamer600 the lead survives bias adjustment but is not yet decisive; the October scoring settles it.

Pitchers 2021 to 2025 vs the field

PECOTA 2021-2025, Steamer 2021-2025, ZiPS 2022-2025. These are the public systems whose historical preseason pitcher files we hold; more will be added as their archives are obtained.

Season by season, all 14 scored system-seasons

Year	System	Pitchers	PITCH+ RMSE	System RMSE	PITCH+ RMSE (debiased)	System RMSE (debiased)	PITCH+ r	System r	Raw winner	Debiased winner
2021	PECOTA	544	11.05	11.79	11.03	11.75	0.393	0.242	PITCH+	PITCH+
2021	Steamer	540	11.0	11.64	10.98	10.74	0.4	0.402	PITCH+	Steamer
2022	PECOTA	519	10.97	11.74	10.97	11.29	0.4	0.323	PITCH+	PITCH+
2022	ZiPS	507	10.89	12.58	10.89	11.66	0.409	0.315	PITCH+	PITCH+
2022	Steamer	519	10.94	12.91	10.93	11.1	0.401	0.352	PITCH+	PITCH+
2023	PECOTA	527	10.37	10.51	10.36	10.51	0.418	0.383	PITCH+	PITCH+
2023	ZiPS	517	10.35	10.79	10.34	10.67	0.432	0.375	PITCH+	PITCH+
2023	Steamer	523	10.37	10.27	10.37	10.27	0.421	0.43	Steamer	Steamer
2024	PECOTA	515	10.65	12.38	10.65	11.2	0.351	0.245	PITCH+	PITCH+
2024	ZiPS	509	10.59	11.66	10.59	11.06	0.347	0.257	PITCH+	PITCH+
2024	Steamer	514	10.57	11.53	10.57	10.74	0.355	0.325	PITCH+	PITCH+
2025	PECOTA	522	10.67	10.92	10.67	10.85	0.335	0.319	PITCH+	PITCH+
2025	ZiPS	513	10.63	10.66	10.63	10.65	0.335	0.343	PITCH+	PITCH+
2025	Steamer	520	10.65	10.59	10.65	10.42	0.338	0.395	Steamer	Steamer

We won 12 of the 14 scored system-seasons: all 5 against PECOTA, all 4 against ZiPS, and 3 of 5 against Steamer. Steamer beat us in 2023 and 2025; more in the losses section below.

The long record

Anyone can fit the past. The test that matters is projecting each season with a model that only knew the seasons before it, then scoring it against what happened, next to Marcel, last year's number carried forward, and the league average.

Hitters, 1980-2025

Pitchers, 2011-2025

Hitters: 46 of 46 seasons better than Marcel. Pitchers: 15 of 15. Not one lucky decade, not one lucky era: every single season.

Team wins

Player projections rolled up to team win totals, tested over 180 team-seasons.

The final team model beats both naive baselines, but the margin over regressed last-year wins is about half a win of RMSE (10.97 vs 11.43), and before calibration the roll-up only edges that baseline (11.18 vs 11.43). Team win totals are mostly luck plus roster talent; nobody's team projections are dramatically better than a good regression, and we will not pretend ours are. No public system's team projections have been scored head to head yet; that starts with the 2026 scoreboard below.

The 2026 scoreboard

Actual is the live 2026 winning percentage, updated with the daily refresh; final scoring comes when the season ends. The closest projection on each row is bold.

Where we lose

No projection system wins everywhere, and a page that only shows wins is advertising, not evidence. The current losses:

• Steamer's pitcher projections beat ours outright in 2023 and 2025, and also in 2021 once each system's overall bias is removed. Steamer is the best public pitching system we have scored and the gap is real in the seasons it won.
• ZiPS 2025 pitching was effectively a tie on error, with ZiPS holding the higher correlation that year.
• On 2026 hitters our raw-error lead over the Depth Charts, Steamer, and Steamer600 is not yet statistically decisive after bias adjustment. It is a lead, not a verdict.
• Our team-level edge over a simple regressed baseline is modest, roughly half a win of RMSE, and our team numbers have not yet been scored against any public system. That scoring starts this October and will be posted here.
• The 2026 hitter table is season to date, not final, and the pitching head-to-head covers the three public systems whose historical preseason files we hold.

When any of these change, in either direction, this page changes with them.

Every number on this page comes from scoring frozen projections against what actually happened, with past seasons projected using only what was knowable beforehand. Same players, same yardstick, playing-time weighted, losses included.