Gigantes 6, Toros 5 (LIDOM)

Game of the Night: Gigantes 6, Toros 5 (LIDOM)

Continued from page 1

• Drama Index 4.62 ★★★★★ · 2 lead changes · peak leverage 7.70 (bottom 9th)
• Top WPA: Gilberto Jimenez +0.61 (batter) · Jimmy Cordero +0.37 (pitcher) · Jesús Sánchez +0.32 (batter)
• Bottom WPA: Joe Record -0.61 (pitcher) · Emiliano Teodo -0.41 (pitcher) · Mike Rosario -0.24 (batter)

Watch: Gameday (MLB.com)

Inning	Matchup	Result	Win probability	WPA	LI
Top 9th	Gilberto Jimenez vs. Joe Record	Single	GIG 20% → 82%	+0.61	6.27
Bot 9th	Mike Rosario vs. Jimmy Cordero	Groundout	TOR 24% → 0%	-0.24	7.70
Bot 7th	Jesús Sánchez vs. Hansel Robles	Single	TOR 58% → 76%	+0.19	2.77

Transaction Wire

Continued from page 1

• Orioles: RHP Colin Selby back from the 60-day IL

More moves (4)

• Orioles: RHP Albert Suárez elected free agency
• Orioles: SS José Barrero elected free agency
• Reds: DH Eugenio Suárez elected free agency · MLB Trade Rumors
• Reds: RHP Pierce Johnson elected free agency · MLB Trade Rumors

The Week at a Glance

Continued from page 1

• By the numbers: 3 games · 23 runs · 2 HR · 1 extra-inning game · 1 one-run game · 0 shutouts · 1 walk-off
• Extremes: longest: Brewers 3, Rays 4 (12), Tue Oct. 27 · most runs: Brewers 10, Rays 1, Mon Oct. 26
• Team of the week: Brewers 2-1, +11 run differential, streak W1
• Walk-offs: Brewers 3, Rays 4 (12), Tue Oct. 27: Ben Williamson single

Game of the Week: Brewers 3, Rays 4 (12)

• Tue Oct. 27 · Drama Index 4.43 ★★★★★ · 2 lead changes · stakes: postseason · 12 innings

Inning	Matchup	Result	Win probability	WPA	LI
Top 12th	Cooper Pratt vs. Ian Seymour	Home Run	MIL 50% → 84%	+0.34	2.29
Bot 12th	Liam Hicks vs. Shane Drohan	Single	TB 34% → 65%	+0.31	5.48
Bot 12th	Jonny DeLuca vs. Shane Drohan	Single	TB 65% → 81%	+0.16	4.48

Slugfest Superlatives

	Player	Detail	When
Best start	Jacob Misiorowski (MIL)	+0.33 WPA over 21 batters, innings 1-6	vs. TB, Wed Oct. 28
Best relief outing	Aaron Ashby (MIL)	+0.34 WPA over 7 batters, innings 9-11, entered at LI 2.22	vs. TB, Tue Oct. 27
Biggest hit	Cooper Pratt (MIL)	home run, top 12th, 2-2 → 3-2, +0.34 WPA	vs. TB, Tue Oct. 27
Toughest week at the plate	Jonathan Aranda (TB)	-0.37 WPA over 11 plate appearances	3 games

Leaderboards

Batter	Team	WPA	PA	G
Cooper Pratt	Brewers	+0.61	14	3
Liam Hicks	Rays	+0.27	9	2
Chandler Simpson	Rays	+0.21	12	3
David Hamilton	Brewers	+0.04	11	3
Christian Yelich	Brewers	+0.04	13	3

Pitcher	Team	WPA	BF	G
Aaron Ashby	Brewers	+0.34	7	1
Jacob Misiorowski	Brewers	+0.33	21	1
Bryan Baker	Rays	+0.26	10	2
Tyler Wells	Rays	+0.25	11	3
Robert Gasser	Brewers	+0.22	19	1

Team	W-L	RS	RA	Diff
Brewers	2-1	17	6	+11
Rays	1-2	6	17	-11

The Bracket

• World Series: Rays at Brewers, Brewers win 4-1 (3 games this week)

How our forecast works

A forecast here is the chance that the home team wins a game, worked out from team ratings, home field and, in MLB, the two starting pitchers.

It is not betting advice, and it does not know lineups, injuries or weather.

How it works

Every forecast is built in the steps below. Each step adds rating points to the home team’s side of the gap and the chance is worked out again, so the last step is exactly the chance that is published and recorded. Nothing is rounded between steps; a chance is rounded to a whole percent only when it is printed.

Step 1: a rating for every club

Every club has one number, its rating, and the average is 1500. When nothing else is counted, a club that is D points ahead of its opponent has a chance of 1 / (1 + 10^(−D / 400)) of winning. This is an Elo rating.

Rating gap	0	25	50	100	200
Chance of the stronger club	50%	54%	57%	64%	76%

A club the model has not seen yet starts at 1500.

Step 2: home field

The home team gets H extra points on top of the gap before the chance is worked out. H is 24 points in MLB and 32 points in the winter leagues. Between two equal clubs that gives the home team 53.4% in MLB and 54.6% in the winter leagues.

Home-field advantage is real but small. In the games the fit learned from, the home team won 53.9% in MLB and 54.9% in the winter leagues. A game at a neutral site, such as the Caribbean Series, has no home-field step: neither club gets the points.

Step 3: the starting pitcher

In the winter leagues the starters are not counted: a winter pitcher has about ten starts to judge him by, and the announced starters are often missing from the data. In MLB the starter adds points, in five steps.

• FIP puts a pitcher’s season in one number, the runs per nine innings he would have allowed from what he controls: (13 × HR + 3 × (BB + HBP) − 2 × K) / IP + a constant. The constant makes the league’s FIP equal the league’s ERA, so FIP reads like ERA. We work it out ourselves from the pitcher’s numbers.
• This season and last season are put together. Last season’s numbers count at a weight of 0.5.
• Regression to the mean pulls the number toward the league’s: 300 innings of a league-average pitcher are added to his record. A pitcher with 20 innings keeps 6% of his own FIP, one with 200 innings keeps 40%.
• The club’s usual starter is the average of that number for the starters of its last 20 games. A club’s rating already holds its usual rotation, so only the difference from the usual starter counts.
• The difference becomes points: (usual FIP − tonight’s FIP) × the league’s average innings per start (5.19) / 9 is runs per game, and each run per game is worth 150 points, about 20 percentage points of chance.

Example: a starter at 2.80 against a usual starter at 3.90 is 1.10 runs per nine innings, × 5.19 / 9 = 0.634 runs per game, × 150 = 95.1 points for his club.

A starter who is not announced, or an opener (a pitcher who started under half of the games he pitched in), counts as the club’s usual starter: 0 points.

Step 4: one game, step by step

Tiburones at Cardenales is the first game of Nov. 1, 2026 in the record, the latest day with a forecast. Each step adds its points to the home team’s side of the gap and the chance is worked out again.

• Start: 50%. Every game begins as a coin flip.
• Ratings: Cardenales is 41.7 points ahead of Tiburones. Chance 56%.
• Home field: +32.0 points for Cardenales. The gap is now +73.7. Chance 60%.

The same walk as one line:

Built from: 50% → ratings 56% → home field 60% · starters not counted in winter ball

From the ratings and home field alone, Cardenales is expected to win 60%. If it wins, its rating moves by +3.2 points and Tiburones’s by −3.2. If it loses, by −4.8 and Tiburones’s by +4.8. K is 8.

Step 5: after the game

A rating moves by K × (result − expectation). The result is 1 for a win and 0 for a loss. The expectation is the chance from the ratings and home field alone: starters do not move ratings, so a club is not credited twice for a good pitcher. The opponent moves by the same amount the other way. K is 5 in MLB and 8 in the winter leagues.

Example with K = 5: a home team expected to win 60% gains 5 × (1 − 0.6) = 2 points when it wins and loses 5 × 0.6 = 3 points when it loses.

Step 6: a new season

At a season’s first game every rating gives back a share r of its distance to 1500. r is 33% in MLB and 50% in the winter leagues. Clubs change over the winter and the model knows nothing about the moves, so it forgets part of what it learned. The winter leagues have their own r.

Example: with r at 33%, a rating of 1560 goes back to 1540.2.

Step 7: a series

The chance of winning a playoff series (so far a winter league’s final) is counted exactly from the chances of the games still to be played, in the order of home games the schedule gives, for example 2-3-2 in a best of seven. Every way the games can go is added up, so the number is not a simulation.

A series gets a number only when the schedule lists its length, its state and every game it can still have, “if necessary” games included. When the schedule does not say, nothing is guessed.

The Caribbean Series is a tournament at one site, and each of its clubs comes from a league: it carries the rating of that league’s champion, and a club with no league of ours starts at 1500. Home field goes only to the club of the host country, in its own games; every other game is on neutral ground. A semifinal and the final are one game each, so the title chances are a product of single games: a club’s chance to win its semifinal, times, for each club it could meet in the final, the chance that club gets there and the chance to beat it, added up. Nothing is simulated.

Step 8: checking the forecasts

Every forecast is checked against the result, and the check is counted from the record alone: the forecasts as they were written before the game, never worked out again with today’s numbers.

• The Brier score is the average of (forecast − result)², with the result 1 for a home win and 0 for a loss. 0 is perfect and 0.25 is a coin flip. A forecast of 60% scores 0.16 when it comes true and 0.36 when it does not.
• Two simple rules to beat: the coin flip, which always says 50% (0.25), and “the home team always wins”, which always says the home team’s share of wins in the fit (53.9% in MLB and 54.9% in the winter leagues).
• Calibration sorts the games into bands by the favorite’s chance and compares each band’s average forecast with how often the favorite won.
• A backtest scores the model on seasons the fit never saw.

Only finished games count. A forecast counts if it was written before the first pitch, and the last one written before the first pitch is the one kept. A cancelled or tied game stays in the record and is not scored. After a game only its result is added: nothing in the record is deleted or changed.

Step 9: where K, H, r, R and W come from

A fit tries a grid of values and keeps the combination with the smallest Brier score on the training seasons: first the team’s numbers (K, H, r), then the pitcher’s (R, W). Each season in the fit is replayed the way a live run is: everyone starts at 1500, and the Brier score counts only after the same settling, two seasons for MLB and one for the winter leagues.

MLB is fitted on Retrosheet’s games and the winter leagues on MLB Stats API results; the seasons used are in the table. The winter leagues share one K, one H and one r.

Number	MLB	Winter
K (points per surprise)	5	8
H (home field, points)	24	32
r (pull toward 1500)	33%	50%
R (innings of regression)	300	–
W (points per run)	150	–
Last season’s weight	0.5	–
Innings per start	5.19	–
Home team won	53.9%	54.9%
Fitted on	2000–2019 and 2021	2016–2023
Tested on	2022–2025	2024–2025
Fitted	Oct. 29, 2026	Oct. 29, 2026

The fit repeats itself: MLB whenever the site’s history is rebuilt, the winter leagues after each winter season ends. Every forecast in the record carries the version of the numbers it was made with.

Calibration

Do the favorites win as often as we say? Games are put in bands by the favorite’s chance, and each band shows the average chance we gave and how often the favorite won.

Favorite’s chance	Games	Average forecast	Favorite won	Share
50-55%	8	52.0%	4	50%
55-60%	5	57.4%	4	80%
60-65%	1	61.8%	1	100%
65-70%	0	–	–	–
70% and up	0	–	–	–

A band holds favorite chances from its low edge up to, not including, its high edge. A game at exactly 50% has no favorite: it is counted in the first band and never as a win for the favorite.

What the model does not know

• Lineups and rest days: who plays tonight and who sits.
• Injuries and roster moves. A rating learns only from results, so a trade or an injury shows up only after the games.
• Bullpens: who is rested or tired, and who pitches after the starter.
• Weather, ballpark and umpire.
• Moves over the winter. A new season only brings every rating part of the way back to 1500.
• Winter-league starters: they are not counted.
• Betting odds and other people’s forecasts: none are used. Only MLB Stats API results and Retrosheet’s history go in.
• How a game is going. The win probability and the Drama Index in the Daily describe a game while it is played, with a different model; a forecast is made before the first pitch.

Data and sources

Results, schedules, announced starters and pitchers’ numbers come from the MLB Stats API.

The MLB model is fitted on Retrosheet’s play-by-play history, which the site builds into its own data, and the winter leagues’ model on MLB Stats API results.

Each forecast in the record carries the version of the numbers it was made with. The numbers now in use:

• MLB: version 4070baad, fitted Oct. 29, 2026.
• Winter: version b099bc3c, fitted Oct. 29, 2026.

The information used here was obtained free of charge from and is copyrighted by Retrosheet. Interested parties may contact Retrosheet at “www.retrosheet.org”.

News

• Braves News: Dana Brown, Vidal Brujan back, more — Yahoo Sports
• Skirka to speak at OBA’s Best Ever Coaches Clinic — Canadian Baseball Network
• Yankees’ Spencer Jones, girlfriend get engaged at Empire State Building: ‘Forever starts today’ — Yahoo Sports
• McFarland: Baseball Sask’s Kilshaw reflects on dream year — Canadian Baseball Network

Transactions & Rumors

• Rays Exercise Club Option On Drew Rasmussen — MLB Trade Rumors
• Elly De La Cruz’s Luka Doncic trolling adds twist to Dodgers trade rumors — Yahoo Sports
• Orioles Make Multiple Roster Moves — MLB Trade Rumors
• Some Offseason Trade Targets for the Guardians — Yahoo Sports

Analysis · Prospects (4)

Analysis

• Looking at 3 Yankees’ Prospects With Momentum Going Into 2027 — Just Baseball
• Flores Blooms — Baseball Musings
• In 58th Season, Brewers Finally Make Fans’ Dreams Come True — Just Baseball

Prospects

• 2026 Houston Astros Org Prospect Team — Prospects Live