Receipts · Public Record

The Ledger

This is our report card. Before every game, our model calls who wins and how sure it is — and we lock that call in, stamped with the time, before the game starts. Once it's over, we grade it against what actually happened. Every hit and every miss stays here, in the open. No edits, no hindsight.

How we did this week

last 7 days · games we called right
College Football5–1MLB49–30

Biggest moves this week

vs a week ago

MLB

Playoff odds

Premier League

Title odds

La Liga

Title odds

Serie A

Title odds

Bundesliga

Title odds

Ligue 1

Title odds

EFL Championship

Promotion odds

WNBA

Title odds

How our call has moved since preseason

month by month

College Football

Playoff odds
TeamPreAugTrend
Notre Dame82%82%
Oregon65%65%
Indiana63%63%
Miami57%57%

MLB

Playoff odds
TeamJulAugSepTrend
Braves100%100%100%
Rays100%100%100%
Dodgers100%100%100%
Brewers100%100%100%

NFL

Playoff odds
TeamPreTrend
Baltimore Ravens91%
Buffalo Bills85%
Los Angeles Rams82%
Kansas City Chiefs80%

Premier League

Title odds
TeamPreAugSepTrend
Arsenal44%49%49%
Man City29%33%33%
Chelsea6%8%8%
Liverpool12%5%6%

La Liga

Title odds
TeamPreAugSepTrend
Barça54%61%62%
Real Madrid24%28%28%
Atlético Madrid5%4%4%
Villarreal9%3%3%

Serie A

Title odds
TeamPreAugSepTrend
Inter Milan46%50%49%
Como18%15%15%
Roma10%9%12%
Juventus9%11%10%

Bundesliga

Title odds
TeamPreAugSepTrend
Bayern82%85%85%
Dortmund8%8%8%
RB Leipzig2%2%2%
Stuttgart3%2%2%

Ligue 1

Title odds
TeamPreAugSepTrend
PSG46%40%40%
Lens19%20%20%
Marseille8%8%8%
Lille6%7%7%

EFL Championship

Promotion odds
TeamPreAugSepTrend
Southampton72%78%77%
Middlesbrough66%71%70%
Millwall51%59%60%
Swansea28%41%47%

NBA

Title odds
TeamPreTrend
Thunder22%
76ers17%
Spurs11%
Pistons10%

WNBA

Title odds
TeamJulAugSepTrend
Lynx45%39%39%
Valkyries12%13%13%
Aces11%12%12%
Dream7%11%11%

F1

Title odds
TeamJulAugTrend
Andrea Kimi Antonelli97%98%
Lewis Hamilton2%1%
George Russell1%1%
Charles Leclerc0%0%

Soccer — the three-way record

win · draw · loss

How to read these

Three outcomes, not two. Picking one of home, draw or away is a harder question than picking a winner, so these hit rates are not comparable to our football or baseball numbers. The Brier is summed over all three outcomes and runs 0–2, where the two-way sports' runs 0–1.

A draw is a miss, not a push. If we called a home win and it finished level, the hit rate counts that against us in full. The Brier is the fairer read of the same match: it scores every number we published, so rating the draw 30% and being wrong costs far less than rating it 8% and being wrong.

We rarely call a draw. Under this model the draw peaks around 29% while the two win chances split the rest, so it is almost never the single likeliest result. Every drawn match is therefore an automatic miss, which is what caps the hit rates below. The number that actually tests the draw model is on each league's card: what we rated those matches to draw, against how many did.

Premier League

three-way call · win, draw or loss
Right outcome
50%
10 of 20
Always pick home
50%
the bar
Brier
0.603
0–2 scale, lower better
Best possible
80%
we never call draws

Results so far 10 home · 4 draw · 6 away|Draws we rated them 23%, 20% finished level

  • Aston Villa FCvArsenal FCsaid A 49% · was A
  • Manchester United FCvIpswich Town FCsaid H 77% · was H
  • Chelsea FCvBrighton & Hove Albion FCsaid H 57% · was H
  • Leeds United FCvBrentford FCsaid H 42% · was D
  • Sunderland AFCvFulham FCsaid H 41% · was H
  • Tottenham Hotspur FCvNewcastle United FCsaid H 53% · was A

Called, not yet played

locked before kickoff
  • 4 SeptIpswich Town FCvLiverpool FCaway
  • 5 SeptNewcastle United FCvAFC Bournemouthhome
  • 5 SeptNottingham Forest FCvTottenham Hotspur FCaway
  • 5 SeptManchester City FCvCoventry City FChome
  • 5 SeptBrighton & Hove Albion FCvLeeds United FChome
  • 5 SeptBrentford FCvSunderland AFChome
  • 5 SeptFulham FCvCrystal Palace FChome
  • 5 SeptHull City AFCvAston Villa FCaway
  • 6 SeptEverton FCvManchester United FCaway
  • 6 SeptArsenal FCvChelsea FChome

home / draw / away %, then our call

La Liga

three-way call · win, draw or loss
Right outcome
57%
17 of 30
Always pick home
53%
we beat it
Brier
0.557
0–2 scale, lower better
Best possible
77%
we never call draws

Results so far 16 home · 7 draw · 7 away|Draws we rated them 25%, 23% finished level

  • FC BarcelonavRayo Vallecano de Madridsaid H 66% · was H
  • CA OsasunavGetafe CFsaid H 45% · was H
  • RC Celta de VigovAthletic Clubsaid H 54% · was A
  • RC Deportivo La CoruñavValencia CFsaid H 43% · was H
  • Real Madrid CFvMálaga CFsaid H 67% · was H
  • Sevilla FCvClub Atlético de Madridsaid H 38% · was A

Called, not yet played

locked before kickoff
  • 3 SeptReal Sociedad de FútbolvRC Celta de Vigohome
  • 4 SeptReal Betis BalompiévReal Madrid CFhome
  • 5 SeptAthletic ClubvClub Atlético de Madridhome
  • 5 SeptRayo Vallecano de MadridvReal Racing Club de Santanderhome
  • 5 SeptVillarreal CFvRC Deportivo La Coruñahome
  • 6 SeptValencia CFvFC Barcelonaaway
  • 6 SeptMálaga CFvLevante UDhome
  • 6 SeptDeportivo AlavésvCA Osasunahome
  • 6 SeptRCD Espanyol de BarcelonavSevilla FChome
  • 7 SeptGetafe CFvRC Celta de Vigohome
  • 7 SeptElche CFvReal Sociedad de Fútbolhome

home / draw / away %, then our call

Serie A

three-way call · win, draw or loss
Right outcome
80%
16 of 20
Always pick home
40%
we beat it
Brier
0.481
0–2 scale, lower better
Best possible
95%
we never call draws

Results so far 8 home · 1 draw · 11 away|Draws we rated them 26%, 5% finished level

  • Atalanta BCvBologna FC 1909said H 43% · was H
  • US LeccevAS Romasaid A 49% · was A
  • SS LaziovGenoa CFCsaid H 42% · was H
  • Cagliari CalciovFC Internazionale Milanosaid A 55% · was A
  • SSC NapolivComo 1907said A 38% · was A
  • Juventus FCvParma Calcio 1913said H 53% · was H

Called, not yet played

locked before kickoff
  • 4 SeptGenoa CFCvComo 1907away
  • 5 SeptACF FiorentinavTorino FChome
  • 5 SeptFC Internazionale MilanovSSC Napolihome
  • 5 SeptAS RomavAtalanta BChome
  • 6 SeptFrosinone CalciovVenezia FChome
  • 6 SeptParma Calcio 1913vAC Monzahome
  • 6 SeptBologna FC 1909vUS Sassuolo Calciohome
  • 6 SeptJuventus FCvAC Milanhome
  • 7 SeptCagliari CalciovUS Leccehome
  • 7 SeptUdinese CalciovSS Laziohome

home / draw / away %, then our call

Bundesliga

three-way call · win, draw or loss
Right outcome
56%
5 of 9
Always pick home
78%
the bar
Brier
0.520
0–2 scale, lower better
Best possible
78%
we never call draws

Results so far 7 home · 2 draw · 0 away|Draws we rated them 22%, 22% finished level

  • FC AugsburgvFC Schalke 04said H 49% · was H
  • SC FreiburgvSV Werder Bremensaid H 50% · was H
  • Borussia DortmundvHamburger SVsaid H 60% · was H
  • SV 07 ElversbergvBayer 04 Leverkusensaid A 47% · was H
  • 1. FSV Mainz 05vSC Paderborn 07said H 51% · was D
  • 1. FC KölnvTSG 1899 Hoffenheimsaid A 40% · was H

Called, not yet played

locked before kickoff
  • 4 SeptVfB Stuttgartv1. FC Kölnhome
  • 5 SeptSV Werder BremenvRB Leipzigaway
  • 5 SeptTSG 1899 HoffenheimvBorussia Dortmundaway
  • 5 SeptBayer 04 Leverkusenv1. FC Union Berlinhome
  • 5 SeptBorussia MönchengladbachvSV 07 Elversberghome
  • 5 SeptSC Paderborn 07vSC Freiburgaway
  • 5 SeptFC Schalke 04vFC Bayern Münchenaway
  • 6 SeptHamburger SVv1. FSV Mainz 05home
  • 6 SeptEintracht FrankfurtvFC Augsburghome

home / draw / away %, then our call

Ligue 1

three-way call · win, draw or loss
Right outcome
39%
7 of 18
Always pick home
33%
we beat it
Brier
0.667
0–2 scale, lower better
Best possible
61%
we never call draws

Results so far 6 home · 7 draw · 5 away|Draws we rated them 24%, 39% finished level

  • AS Monaco FCvOlympique de Marseillesaid H 43% · was H
  • Stade Rennais FC 1901vLe Mans FCsaid H 60% · was H
  • Paris FCvOGC Nicesaid H 52% · was H
  • Stade Brestois 29vToulouse FCsaid H 41% · was D
  • Olympique LyonnaisvLe Havre ACsaid H 53% · was D
  • FC LorientvES Troyes ACsaid H 55% · was A

Called, not yet played

locked before kickoff
  • 3 SeptToulouse FCvLille OSChome
  • 4 SeptOlympique LyonnaisvAJ Auxerrehome
  • 4 SeptParis Saint-Germain FCvAS Monaco FChome
  • 5 SeptRacing Club de LensvFC Lorienthome
  • 5 SeptLe Havre ACvStade Brestois 29home
  • 5 SeptOGC NicevLe Mans FChome
  • 6 SeptES Troyes ACvRC Strasbourg Alsaceaway
  • 6 SeptAngers SCOvStade Rennais FC 1901away
  • 6 SeptOlympique de MarseillevParis FChome

home / draw / away %, then our call

EFL Championship

three-way call · win, draw or loss
Right outcome
36%
13 of 36
Always pick home
42%
the bar
Brier
0.657
0–2 scale, lower better
Best possible
67%
we never call draws

Results so far 15 home · 12 draw · 9 away|Draws we rated them 27%, 33% finished level

  • Birmingham City FCvSouthampton FCsaid A 38% · was D
  • Stoke City FCvNorwich City FCsaid H 38% · was H
  • Swansea City AFCvWatford FCsaid H 44% · was H
  • Portsmouth FCvDerby County FCsaid A 38% · was A
  • Sheffield United FCvBolton Wanderers FCsaid H 50% · was H
  • West Ham United FCvWolverhampton Wanderers FCsaid H 40% · was H

Called, not yet played

locked before kickoff
  • 2 SeptMillwall FCvWrexham AFChome
  • 2 SeptQueens Park Rangers FCvCardiff City FChome
  • 2 SeptWest Bromwich Albion FCvCharlton Athletic FChome
  • 2 SeptBurnley FCvMiddlesbrough FCaway
  • 5 SeptPreston North End FCvBlackburn Rovers FChome
  • 5 SeptStoke City FCvCharlton Athletic FChome
  • 5 SeptLincoln City FCvSouthampton FCaway
  • 5 SeptQueens Park Rangers FCvMiddlesbrough FCaway
  • 5 SeptSheffield United FCvNorwich City FChome
  • 5 SeptPortsmouth FCvCardiff City FChome
  • 5 SeptWest Bromwich Albion FCvWatford FChome
  • 5 SeptWest Ham United FCvDerby County FCaway

home / draw / away %, then our call · showing 12 of 22 locked

Game predictions

Before every game, the model locks in a win probability. Once the game is final, we score it — no edits, no hindsight. Here's how the calls have held up, sport by sport.

New here? What these numbers mean
Right side —
how often the team we favored actually won. Just win/loss on our picks.
Brier score —
our grade for the probabilities themselves, not just which side we picked. It rewards being right and being honest about how sure we were: a confident call that hits scores well, a confident call that misses gets punished hard. Saying 50% on everything (a coin flip) scores 0.25 and a flawless forecaster scores 0 — so the lower the number, the sharper our odds have been.
The bar we hold ourselves to —
a coin flip is the wrong thing to beat. Home teams win a bit more than half their games, so anyone can score about 0.250 by predicting the home-field rate every single time, while knowing nothing whatsoever about the two teams. That number — not 0.25 — is the bar shown under each Brier score, and beating it by a little is the honest description of what a good baseball model does.
Calibration —
whether our confidence is honest: when we say a team has a 70% chance, do teams like that really win about 70% of the time?

MLB game predictions

the board →

Games called

676

locked before first pitch

Now final

652

played & scored

Right side

58%

our pick won 376 of 652

Brier score

0.245

must beat 0.249 — a model that knows nothing

Calibration

predicted vs actual

When the model gives a team, say, a 60% chance — do teams like that actually win about 60% of the time? Each row is how often our picks in that range really won.

30–40%
55%
40–50%
43%
50–60%
57%
60–70%
66%

actual predictedWell-calibrated → the marker sits at the edge of the bar.

Recently graded

Locked predictions

home win prob · graded when final
  • Sep 16:40 PMMets@RaysSean Manaea vs Freddy PeraltaTB 56%
  • Sep 16:40 PMBlue Jays@GuardiansSpencer Miles vs Gavin WilliamsSP CLE +2CLE 56%
  • Sep 16:40 PMPadres@RedsRandy Vásquez vs Nick LodoloSP SD +1CIN 46%
  • Sep 16:40 PMGiants@PiratesLogan Webb vs Paul SkenesSP PIT +1PIT 57%
  • Sep 16:45 PMMariners@Red SoxBryan Woo vs TBDSP SEA +1BOS 58%
  • Sep 16:45 PMBraves@NationalsAJ Smith-Shawver vs Jake IrvinSP ATL +1WSH 47%
  • Sep 17:40 PMBrewers@CubsRobert Gasser vs Matthew BoydSP CHC +2CHC 53%
  • Sep 17:40 PMMarlins@RoyalsTyler Phillips vs Randy DobnakSP KC +3KC 51%
  • Sep 17:40 PMTigers@TwinsTroy Melton vs Andrew MorrisMIN 47%
  • Sep 18:05 PMAthletics@RangersBrady Basso vs MacKenzie GoreTEX 57%
  • Sep 18:10 PMWhite Sox@AstrosSean Burke vs Ronel BlancoSP CWS +3HOU 46%
  • Sep 18:40 PMOrioles@RockiesKyle Bradish vs Gabriel HughesSP COL +1COL 50%
  • Sep 19:38 PMYankees@AngelsGerrit Cole vs Grayson RodriguezSP NYY +1LAA 43%
  • Sep 19:40 PMPhillies@D-backsJesús Luzardo vs Eduardo RodriguezSP PHI +1AZ 50%
  • Sep 110:10 PMCardinals@DodgersMichael McGreevy vs Eric LauerSP STL +3LAD 56%
  • Sep 212:40 PMPadres@RedsCasey Mize vs Brandon WilliamsonSP SD +3CIN 44%
  • Sep 22:35 PMAthletics@RangersJacob Lopez vs Cody BradfordTEX 57%
  • Sep 23:10 PMOrioles@RockiesTrevor Rogers vs Tomoyuki SuganoSP BAL +2COL 46%
  • Sep 26:40 PMBlue Jays@GuardiansDylan Cease vs Joey CantilloSP TOR +5CLE 49%
  • Sep 26:40 PMGiants@PiratesLanden Roupp vs Jared JonesSP SF +1PIT 54%
  • Sep 27:40 PMMarlins@RoyalsEury Pérez vs Noah CameronKC 49%
  • Sep 27:40 PMBrewers@CubsJacob Misiorowski vs David PetersonSP MIL +4CHC 48%
  • Sep 27:40 PMTigers@TwinsDrew Anderson vs Dean KremerSP DET +1MIN 47%
  • Sep 29:38 PMYankees@AngelsCam Schlittler vs Reid DetmersLAA 44%

Reading a row: the percentage is our chance that the named team — the home team — wins. The bolded team is our pick. Every line is locked before first pitch and graded automatically once the game is final. SP is what tonight's starters are worth, and it always names the team it helps: “SP PHI +5” means the pitching matchup moves the game five points to Philadelphia. Each starter is measured against his own club's rotation rather than the league, so an ace on an excellent staff can show a small number, or none at all. A row with no SP tag was locked before the starters were posted, so it's team-level only.

MLB home-win probabilities come from a log5 matchup on each club's regressed Pythagorean rating plus a fixed home-field edge — the same engine that runs the season simulation — then adjusted for the announced starting pitchers. Each starter is rated on FIP (the earned-run rate implied by his strikeouts, walks and home runs, which strips out defense and batted-ball luck), using only his innings as a starter, regressed 70 innings toward the league's rotation baseline, and measured against his own rotation's norm, since the club's run prevention already assumes an average start from it. That gap is credited 65% to the starter and capped at 10 points. A quarter of the rotation's own FIP-versus-runs gap is applied to the club's expected runs allowed on the same logic. Every weight was fit out of sample, on games played after the stats used to predict them — and the bullpen was tested the same way and left at zero because it didn't pay.

Full MLB record, week by week →

College football predictions

the board →

Games called

49

locked before kickoff

Now final

6

played & scored

Right side

83%

our pick won 5 of 6

Brier score

0.159

must beat 0.222 — a model that knows nothing

Calibration

predicted vs actual

When the model gives a team, say, a 60% chance — do teams like that actually win about 60% of the time? Each row is how often our picks in that range really won.

40–50%
0%
50–60%
100%
60–70%
0%
90–100%
100%

actual predictedWell-calibrated → the marker sits at the edge of the bar.

Recently graded

Locked predictions

home win prob · graded when final

Reading a row: the percentage is our chance that the named team — the home team — wins. The bolded team is our pick. Every line is locked before kickoff and graded automatically once the game is final.

CFB home-win probabilities come from each team's margin — its power rating minus the opponent's, plus a home-field edge — mapped through the model's margin spread. The same ratings that run the season sim.

Full College Football record, week by week →

Every prediction is written before the game and cannot be edited. Brier score is the mean squared error against the 0/1 outcome, lower is better; we judge it against the home-field base rate rather than a coin flip, because the base rate is free. Nothing on this page is backdated.