Our Win Projections Beat the Standings in 10 of 12 Weeks, but Not by Much
BBMI’s weekly projected-win snapshots from July 7 through Sept. 22, checked against all 30 final records and against a projection built from the standings alone.
averaged across all 12 weeks
The simplest way to project a team’s final record is to assume it keeps winning at its current rate. A team that is 50-42 is on pace for 88 wins, so you call it 88. We’ll call that method pace. It needs no model, and it is the first thing a projection system has to beat.
BBMI’s projected wins beat pace in 10 of the 12 weekly snapshots we archived this season. The margin was small. Averaged across all 12 weeks, BBMI’s projections landed 0.13 wins per team closer to the final totals than pace did. The biggest edge came in the first snapshot, July 7, when BBMI missed teams’ final totals by an average of 3.86 wins and pace missed by 4.16. Pace came out slightly ahead on Aug. 25 and Sept. 22.
On July 7, BBMI had San Diego finishing with 79.4 wins and Boston with 76.4. The Padres won 91 and the Red Sox 87, the two largest misses in the review. In the other direction, the Athletics (projected 73.4, finished 64) and Seattle (83.8, finished 76) came up well short. The closest early calls were Milwaukee (102.3, finished 103), the Cubs (89.2, finished 89) and the White Sox (84.1, finished 84).
As a group, the July 7 projections did not lean high or low. Fifteen teams were projected too high and 15 too low, and 20 of the 30 finished within five wins of their projection.
How projected wins are calculated
A projected win total starts with the wins a team has already recorded. We then simulate the rest of its actual schedule game by game, using the model’s win probability for each matchup, and repeat the whole simulation 10,000 times. The projection is recorded wins plus the average number of additional wins across those 10,000 seasons. A team with 60 wins that averages another 25.4 in the simulations is projected for 85.4.
That is why the projections carry decimals. As the season goes on, completed games become fixed results and fewer games remain to simulate, so the projections have less room to move.
Every team, every week
Each of the six division charts below follows one team per line across the 12 weekly snapshots. The diamond at the right edge is that team’s actual final win total. A projection that was right all along runs flat and ends at its diamond. A line that climbs or drops to meet its diamond belongs to a team the model had to catch up with.
BAL
BOS
NYY
TB
TOR
ATL
MIA
NYM
PHI
WSH
CWS
CLE
DET
KC
MIN
CHC
CIN
MIL
PIT
STL
ATH
HOU
LAA
SEA
TEX
ARI
COL
LAD
SD
SF
Values are projected final win totals, not power-ranking scores or schedule-adjusted ratings. All six charts use the same scale.
The review starts July 7 because 2026 is BBMI’s first MLB season, and we did not consider the model ready until late May or early June. The weekly archive begins July 7, so that is where the scoring begins.
BBMI against pace, week by week
Average miss is mean absolute error: for each of the 30 teams, take the gap between the projection and the final win total, ignore whether it was too high or too low, and average across teams. Lower is better. Pace is the team’s winning percentage on the snapshot date multiplied by 162. BBMI edge is pace’s average miss minus BBMI’s; a positive number means BBMI was closer. BBMI teams within 3 wins counts BBMI projections that landed within three wins of the final total. For scale, final win totals this season ran from 58 (Colorado) to 103 (Milwaukee).
Both lines fall steeply, and most of that drop reflects the shrinking schedule rather than forecasting skill. On July 7 the average team had about 71 games left. By Sept. 22 it had fewer than six. Every completed game is one less game to get wrong, which is why pace improved almost as much as BBMI did: 4.16 to 0.86, against BBMI’s 3.86 to 0.90.
The column that measures the model is the edge. BBMI was closer than pace in 10 of 12 weeks, by between 0.04 and 0.30 wins per team. The edge averaged 0.20 wins per team across the first six snapshots and 0.06 across the last six. That pattern is consistent with the model’s read on team quality mattering most when the most games remain. It is still a modest edge from one season. The 12 snapshots are not 12 independent tests, since the same 30 teams and many of the same remaining games appear week after week, and we have not tested the gaps for statistical significance.
Miami shows what the edge looks like on a single team. On July 7 the Marlins were 50-42, an 88-win pace. BBMI projected 86.2, which meant expecting .517 baseball the rest of the way instead of .543. Miami went 4-12 over the next three weeks and finished with 80. Both methods were too high, and BBMI was 1.8 wins closer.
The model does not always separate from the standings. The Athletics were 41-50 on July 7, a 73.0-win pace, and BBMI projected 73.4. Neither method saw a 64-win finish coming.
San Diego and Boston finished far above their July projections
The Padres were the largest miss in the review and the one the model never fully caught. BBMI projected 79.4 wins on July 7, 82.6 on Aug. 4, 85.6 on Sept. 1 and 89.5 on Sept. 22. San Diego finished with 91, above every weekly projection in the archive.
The Padres struggled into July. They lost eight straight from June 27 to July 4, their longest skid in 13 years, and were three games under .500 and 11th in the National League late in the month. The second half went the other way. When San Diego clinched a playoff spot, MLB.com’s assessment was that the Padres had baseball’s best offense in September on top of what was already its best bullpen. On Sept. 20, Fernando Tatis Jr. hit his 25th home run in a win over Miami, San Diego’s 13th in 14 games. Each weekly projection expected the rest of the schedule to go roughly as the model’s game-by-game probabilities said it would, and each week the Padres won more than that.
Boston’s turnaround came earlier and faster. The Red Sox were last in the AL East before the streak. After a July 1 loss to Washington, they won 15 straight, tying the franchise record set by the 1946 team, and went 27-3 from July 3 through Aug. 7. They did it with Garrett Crochet and Roman Anthony out for extended stretches. Sonny Gray, who posted a 1.97 ERA over 45 innings in June and July, and Willson Contreras drove the run, and Aroldis Chapman closed, moving to first on MLB’s all-time strikeout list for relievers during the streak.
BBMI went from 76.4 projected wins on July 7 to 87.6 on Aug. 4, an 11.2-win move in four weeks. Boston finished with 87. The Aug. 4, Sept. 1 and Sept. 22 projections (87.6, 87.8 and 86.7) were all within a win of that.
The Athletics and Seattle finished well below them
The Athletics’ slide was already underway at the July 7 snapshot. They spent 44 days alone or tied for first in the AL West earlier in the season, then lost 17 of 20 heading into the All-Star break, including the last nine. Nick Kurtz and Zack Gelof were on the injured list, the staff’s 5.21 ERA was the worst in the American League, and the club fired pitching coach Scott Emerson during the break. The A’s went 4-17 between the July 7 and Aug. 4 snapshots, moving from 41-50 to 45-67, and BBMI’s projection fell from 73.4 to 65.4. They finished with 64, 9.4 wins below the July call.
Seattle’s projection came down more gradually. BBMI had the Mariners at 83.8 wins on July 7 and 77.2 on Aug. 11. On Aug. 13, Logan Gilbert threw six scoreless innings in a 1-0 win at Yankee Stadium that ended a six-game losing streak. It was Seattle’s sixth win in 21 games. The Mariners finished with 76, the same number as BBMI’s final weekly projection.
Miami followed the same path on a smaller scale, going from 50-42 to 54-54 between the July 7 and July 28 snapshots. Its projection fell from 86.2 to 80.2 over that span, and the Marlins finished with 80.
These records line up with the projection changes, but this review does not separate how much of each change came from games already played, roster and injury updates, the remaining schedule, or changes to the model. That would take a separate analysis.
Milwaukee’s first projection was its best
BBMI’s first archived projection for the Brewers was 102.3 wins, against a final 103. The projection did not hold there. It fell to 98.2 by Aug. 4 and stayed below 100 until the final weekly snapshot, which had 101.1. The July call on Milwaukee was closer than the September one.
The league-wide numbers had a similar wobble. The average miss rose from 1.65 wins on Sept. 1 to 1.77 on Sept. 8 before falling again. Fewer games left makes a projection more accurate on average, not in every week.
The full scorecard
Four of the 12 snapshots for every team, sorted by the July 7 miss, from the teams BBMI projected too low to the teams it projected too high.
Snapshot columns are BBMI’s projected final win totals on that date. July 7 error is the July 7 projection minus actual wins: negative means BBMI projected too few wins, positive too many. The average size of this error across all 30 teams, ignoring sign, is 3.86 wins.
What one season shows
With about 71 games left per team, BBMI’s July 7 projections were within five wins of the final total for 20 of 30 teams and did not lean high or low as a group. The large misses were teams whose seasons turned sharply after that date, and the model trailed San Diego all the way to the end. Against a projection built from the standings alone, BBMI was closer in 10 of 12 weeks, by an average of 0.13 wins per team.
That suggests the model added some information beyond the standings this season.
One season is not enough to show how durable that edge is. The next test is whether it holds across seasons and against stronger benchmarks than pace.
Source and method: BBMI MLB rankings and weekly archive, retrieved Sept. 27, 2026, Central time. The review covers 360 team-snapshot observations: 30 teams across 12 weekly archive labels, July 7 through Sept. 22. It scores the one-decimal simulation values behind the page’s “Proj W–L” column, before the page rounds them to whole wins, and does not score the separate schedule-adjusted ranking values. Each week is averaged across its own 30 teams. Current and final values are not scored as forecasts.
Dates: Chart and table dates follow the archive selector and are not always the simulation date. The July 7 label holds a July 8 simulation, July 14 holds July 13, and the other weekly labels hold next-day simulations. The accompanying CSV records both. Some early snapshots were extracted later from historical commits, so this review does not independently establish when every value first appeared to readers.
Actual results: Final win totals come from BBMI’s final records and match the Sept. 27 season-simulation payload, which has zero games remaining, for all 30 teams. The Yankees and Orioles each played 161 games. The projected W–L display forces a 162-game total; we score their actual 93 and 79 wins without adding a result. These are BBMI’s published final records, not a separate reconciliation against an official league feed.
Benchmark: Pace is 162 × wins ÷ (wins + losses), using the standings in the same simulation payload as each forecast. Team records and paces cited in the text come from those same payloads. Pace is a simple benchmark, not another forecasting service. The review does not test prediction-interval calibration or statistical significance. Game and player details come from the linked reports. Team logos: MLB.
Related research
Schlittler Pulls Clear; Misiorowski and Yamamoto Lead the NL Race
September 16: Cam Schlittler leads the AL runs-saved model at 99.8%. Jacob Misiorowski leads the NL at 57.4%, with Yoshinobu Yamamoto at 37.2% and Chris Sale at 5.4%. Updated inputs, 500 simulated finishes.
Cam Schlittler 99.8% AL · Jacob Misiorowski 57.4% NL · 500 simulated seasonsAwards · MVPCrow-Armstrong Pulls Ahead in the September MVP Update
September 16: Pete Crow-Armstrong leads the NL value model at 97.0%, up from 66.2%. Yordan Alvarez again finishes first in all 500 AL simulations. Updated batting, pitching and Statcast fielding inputs; the same player-value framework.
Yordan Alvarez 100.0% AL · Pete Crow-Armstrong 97.0% NL · 500 simulated seasonsAwards · Cy YoungSchlittler, Misiorowski Hold Cy Young Leads as Sale Surges
Cam Schlittler and Jacob Misiorowski answered late-August pressure with wins, and the model still has both atop their leagues at 67.0% and 58.8%. The movement is behind them: Parker Messick has climbed to 20.2%, Dylan Cease to 11.8%, and Chris Sale to 11.8% on the back of a nine-inning shutout of the Dodgers. Plus the July and August calls, graded.
Schlittler 67.0% AL · Misiorowski 58.8% NL · 500 simulated seasonsSee it live
This is the “we show our work” side. The Prediction Record is the other half — every call locked before the game and graded in public.