MLB · MODEL PERFORMANCE

The whole record, one page

Both halves of the record in one place: every graded pick with its edge and result, and every projection measured against the Vegas line on the same games. Live since May 23, 2026, the start of the simulation era. Locked before first pitch, graded at posted prices, nothing removed.

✓ LOGGED BEFORE FIRST PITCH2026 season, from 5/23

1,695

Games priced

60.0%

F5 ML win rate

Oct 28, 2026

Graded through

SEASON

F5 MLFIRST-5 MONEYLINE · FIRED PICKS

60.0%

138W – 92L · 41P · 95% CI 53.6–66.1%

published · 2026 · break-even 58.6%

+2.0%

ROI at posted prices

+5.5u

Units won

F5 RLFIRST-5 RUN LINE · ±0.5

55.8%

126W – 100L · 1P · 95% CI 49.2–62.1%

published · 2026 · break-even 52.4%

+6.4%

ROI · −110 juice

+14.5u

Units won

LINE QUALITYF5 ML

45.4%

Share of games BBMI landed closer than the market, on the game total

published · 1,695 paired games · May 23 – Oct 28 ’26

3.57

BBMI game-total MAE

3.45

Market game-total MAE

Calibration by edge

published · F5 ML · 230 decided

F5 ML: pick rules and edge bands

First-five moneyline, home side only. The rows below are every pick split by how far apart the two teams were projected to be after five innings. In force today. Premium fires at home & EV > 0. The published rule offers no Premium ★ on this product, though 57 picks carry that tier, the first on Jul 30, 2026, from a marker written before the current star rule existed. The rows below split the picks by how far apart the model had the two teams, in runs, whichever side it favored. The first row of buttons cuts them by the EV the posted price gave, and the second scopes them to the side the model leaned.

PROJECTED RUN MARGIN	TIER AS POSTED	N	WIN %	ROI
AS POSTED · THE TIER EACH PICK ACTUALLY FIRED UNDER
	Premium	181	59.7%	+3.6%
	Premium ★ Best Bet	49	61.2%	-3.9%
All picks as posted	TOTAL	230	60.0%	+2.0%
BY EDGE INTERVAL · DESCRIPTIVE, SAME PICKS
0.00 to 0.50		50	44.0%	-9.9%
0.50 to 1.00		115	66.1%	+9.5%
1.00 to 1.50		46	60.9%	-2.5%
1.50 to 2.00		11	54.5%	-16.3%
2.00+		8	75.0%	+24.1%
All decided	TOTAL	230	60.0%	+2.0%

Posted tiers versus current edge bands

These are the same picks as the two rows above, cut by what the model projected instead of by the tier they were sold at, so the totals agree. The positive-EV requirement took effect on Aug 24, 2026, the date on the rules card; picks posted before it fired on the projected margin alone, which is why some of them sit below zero EV.

How pick counts and ROI are calculated

Each row is a range of projected run margin in runs, counted the same way whichever side it favored. All picks with Either selected is every decided pick this product posted; the buttons cut that set by the EV its price gave and by the side the model leaned. N counts DECIDED picks. Every ROI here divides realized units by every pick STAKED, pushes included, which is the definition the product card uses and the reason the total below matches it rather than reading 41 pushes higher. So a row’s N and the count its ROI was divided by are not always the same number.

Season to date

F5 ML

Win rate versus break-even

Cumulative win rate after each week, against the break-even this product is priced at. F5 ML prices at 58.6%, which is well above the -110 figure the spread products use, so the two are not comparable by eye.

13 weeksbreak-even 58.6%now 60.0%

BY WEEK ▼	PICKS	WIN %	ROI	CUMULATIVE
10/23 ’26	2	0.0%	-66.7%	60.0%
10/16 ’26	1	0.0%	-50.0%	60.5%
10/9 ’26	1	100.0%	+38.5%	60.8%
10/2 ’26	2	100.0%	+84.7%	60.6%
9/25 ’26	11	90.9%	+51.8%	60.3%
9/18 ’26	19	47.4%	-12.2%	58.7%
9/11 ’26	18	55.6%	+1.0%	59.8%
9/4 ’26	20	60.0%	+8.0%	60.2%
8/28 ’26	19	52.6%	-6.2%	60.3%
8/21 ’26	26	65.4%	+11.9%	61.3%
8/14 ’26	41	53.7%	-12.1%	60.4%
8/7 ’26	35	51.4%	-14.9%	64.3%
7/31 ’26	35	77.1%	+25.0%	77.1%

BBMI vs Vegas

Our picks: model vs market probability

The model’s win probability against the market’s, taken from its two posted prices with the vig removed. Scored over 230 of 271 logged games; the rest are ungraded or ended tied, and a tie pushes this bet rather than proving either side wrong.

230 of 271 logged games · on the moneyline

WHICH LANDED CLOSER125 / 105 / 0 tie

Per game, whose probability sat nearer the result

BBMI54.3%

VEGAS45.7%

BRIER SCOREVegas by 0.0010

Squared probability error, lower is better

BBMI0.2386

VEGAS0.2377

AVERAGE MISS, PCT POINTSBBMI by 0.7 pts

How far each probability sat from the result

BBMI47.3

VEGAS48.0

FAVORITE WONBBMI by 0.0 pts

Share of games each side’s favorite won

BBMI60.4%

VEGAS60.4%

Sport-wide game-total figures

Below: the published sport-wide figures, which are measured on the GAME TOTAL across every priced game and do not move with the window.

Every game the sport priced: line quality

No picks here, line quality head to head, on every game the sport priced. These are the published figures across 1,695 paired games, May 23 – Oct 28 ’26, and they are measured on the GAME TOTAL: the cohort carries no market margin line, so no sport-level margin pair exists to show.

1,695 paired games · May 23 – Oct 28 ’26 · on the game total

WHICH LANDED CLOSERpublished · 770 / 921 / 4 tie

BBMI45.4%

VEGAS54.3%

MAE, RUNSVegas by 0.13 runs

Average miss on the game total, in runs

BBMI3.57

VEGAS3.45

PROJECTED WINNER ACCURACYVegas by 0.7 pts

Share of games each projection’s implied winner was right

BBMI56.6%

VEGAS57.3%

BRIER SCOREVegas by 0.005

Probability accuracy: rewards honest odds, not bold ones

BBMI0.247

VEGAS0.242

Why bias is shown separately

BIAS is published beside these and is not shown: it is a SIGNED lean compared on absolute value, so the smaller number is the smaller lean and not the better projection, which is the opposite of how every row above reads. Both sides lean the same way, low.

Calibration

Brier 0.247 · Published · full 1,695-game sample

Predicted versus actual win rates

Predicted win probability against the actual win rate, across every priced game rather than only the ones a pick fired on. This panel does not follow the product pill: the cohort is the sport’s. Read from the published panel rather than recomputed: a second derivation of a published number is a second number.

HOME WIN P	N	PRED	ACTUAL	GAP
< 35%	80	30.0%	51.2%	+21.3
35–45%	277	40.8%	40.4%	-0.4
45–55%	634	50.0%	51.1%	+1.1
55–65%	536	59.3%	58.2%	-1.1
≥ 65%	168	70.2%	65.5%	-4.7

How to read GAP and Brier

A calibrated 60% should win about 60% of the time. GAP is actual minus predicted, so a positive number means the model was too cautious in that bucket. Brier 0.247 over 1,695 games; lower is better, and 0.25 is what a coin flip scores, which is worth reading against a sport this close to one.

Products & rules

thresholds read from config

Current tier thresholds

The threshold at which each tier fires. Thresholds are versioned in config and read from it, never restated here.

PRODUCT	THE BET	PREMIUM	PREMIUM ★	IN FORCE FROM
F5 ML	First-five moneyline, HOME SIDE ONLY. Fires when the home price gives positive EV.	home & EV > 0	not offered	Aug 24, 2026
F5 RL	First-five spread, both sides. The sign of the edge picks the side.	\|edge\| ≥ 1.25	\|edge\| ≥ 1.75	Aug 5, 2026
F9 ML	Full-game moneyline, HOME SIDE ONLY. Fires when the home price gives positive EV.	home & EV > 0	home & EV ≥ 5%	Aug 24, 2026
F9 RL	Takes the +1.5 on the side the market has as the underdog. Covers if that side loses by one or wins.	away +1.5 cover ≥ 1.50 · home +1.5 cover ≥ 1.25	cover ≥ 2.25	Jul 5, 2026

Why break-even rates differ

The four break-evens differ because the four products are priced differently. A moneyline favorite needs a far higher hit rate to break even than a spread at -110, so comparing their win rates directly says nothing.

When each rule took effect

IN FORCE FROM is the day the definition in the two tier columns took effect, read from the registry that the tier labeler branches on, so the date and the rule cannot drift apart. Picks posted before it were tiered under the rule live that day and are reported that way in AS POSTED, never refitted to today’s.

Rules change; posted picks stay fixed

These are the definitions in force today. They change through the season as the model is refined, and a change applies from its date forward only: a pick is never refitted to a rule that did not exist when it fired, and the log is never edited to match a newer one. Both moneyline products took their current definition on Aug 24, 2026, when the home-side requirement and the positive-EV requirement came in together. Picks posted before that day could take either side and could price below zero EV, which is why the log holds some of each and why they are counted in the record exactly as they fired.

All sports →

The log

271 rows · mlb-f5-ml.json

F5 ML picks: model versus market

Every F5 ML pick, with what the model projected beside what the market posted. On this product the market posts a PRICE, not a line, so the market column is odds and there is no line to land closer to.

WIN P: home leads after five

WIN P is the HOME side’s chance of leading after five innings, with ties refunded. A first-five moneyline pushes when the score is level after five, so home lead, tie and away lead are three outcomes and this column is the first of them, whichever side the model took.

THE PICK
					BBMI FAVORS				FINAL
2026-10-28	Milwaukee Brewers @ Tampa Bay Rays	+125	+0.21	14.8%	Tampa Bay Rays	41%	Premium	LOSS	2–0
2026-10-27	Milwaukee Brewers @ Tampa Bay Rays	-109	+0.21	4.4%	Tampa Bay Rays	49%	Premium	PUSH	2–2
2026-10-24	Tampa Bay Rays @ Milwaukee Brewers	-115	+0.25	3.1%	Milwaukee Brewers	44%	Premium	LOSS	4–3
2026-10-17	Tampa Bay Rays @ Cleveland Guardians	+102	+0.20	6.5%	Cleveland Guardians	44%	Premium	PUSH	1–1
2026-10-16	Tampa Bay Rays @ Cleveland Guardians	+100	+0.15	0.2%	Cleveland Guardians	40%	Premium	LOSS	3–0
2026-10-15	Milwaukee Brewers @ Los Angeles Dodgers	-160	+0.72	3.3%	Los Angeles Dodgers	54%	Premium	PUSH	4–4
2026-10-13	Cleveland Guardians @ Tampa Bay Rays	-130	+0.40	3.3%	Tampa Bay Rays	48%	Premium	WIN	0–3
2026-10-08	Cleveland Guardians @ Chicago White Sox	-105	+0.10	3.1%	Chicago White Sox	44%	Premium	WIN	3–4
2026-10-06	Milwaukee Brewers @ San Diego Padres	-135	+0.73	8.2%	San Diego Padres	52%	Premium	WIN	2–3
2026-10-01	Philadelphia Phillies @ Atlanta Braves	+105	+1.29	34.4%	Atlanta Braves	60%	Premium	WIN	0–5
2026-09-30	Boston Red Sox @ New York Yankees	-145	+0.59	1.5%	New York Yankees	52%	Premium	WIN	0–2
2026-09-30	Chicago Cubs @ San Diego Padres	-145	+0.60	1.9%	San Diego Padres	49%	Premium	WIN	1–4
2026-09-29	Boston Red Sox @ New York Yankees	-147	+0.75	3.5%	New York Yankees	53%	Premium	WIN	0–2
2026-09-27	Texas Rangers @ Minnesota Twins	-105	+0.19	5.1%	Minnesota Twins	45%	Premium	WIN	1–5
2026-09-27	Atlanta Braves @ Miami Marlins	-130	+0.32	0.5%	Miami Marlins	46%	Premium	PUSH	1–1
2026-09-27	Los Angeles Angels @ Seattle Mariners	-175	+1.20	5.7%	Seattle Mariners	57%	Premium	WIN	0–3
2026-09-26	Cleveland Guardians @ Kansas City Royals	-125	+0.54	5.0%	Kansas City Royals	47%	Premium	LOSS	5–3
2026-09-26	Los Angeles Angels @ Seattle Mariners	-180	+1.48	7.7%	Seattle Mariners	60%	Premium	WIN	0–4
2026-09-25	Chicago Cubs @ Boston Red Sox	+108	+0.40	12.1%	Boston Red Sox	47%	Premium	PUSH	3–3
2026-09-25	Texas Rangers @ Minnesota Twins	+105	+0.26	10.6%	Minnesota Twins	44%	Premium	WIN	2–8
2026-09-25	St. Louis Cardinals @ Milwaukee Brewers	-160	+0.79	2.3%	Milwaukee Brewers	53%	Premium	WIN	1–4
2026-09-25	Baltimore Orioles @ New York Yankees	-115	+0.98	16.3%	New York Yankees	56%	Premium	WIN	0–1
2026-09-24	St. Louis Cardinals @ Pittsburgh Pirates	-163	+0.88	5.0%	Pittsburgh Pirates	56%	Premium	PUSH	1–1
2026-09-24	Chicago White Sox @ Kansas City Royals	+100	+0.28	8.2%	Kansas City Royals	45%	Premium	LOSS	9–1
2026-09-24	Tampa Bay Rays @ New York Yankees	-163	+0.71	1.5%	New York Yankees	52%	Premium	LOSS	3–0

All picks published before first pitch and graded at the price posted. Past performance is not a guarantee.