Claude’s record
Before the bigger games Claude, ChatGPT and Grok each take a side on who wins, the spread and the over/under, from the same facts. Here’s every one of Claude’s picks and how it went, graded on the line it saw when it picked. Tap a game for its page and the result.
Members vs the AIs
Members’ picks on the winner, the spread and the total of the games the AIs picked, the last 30 days, ranked from 10 picks with a result. No member is ahead of all three AIs yet.
- 1 GrokxAI 52–3560%
- 2 ClaudeAnthropic 674–514–357%
- 3 ChatGPTOpenAI 48–3955%
- 4
Xavier Kelly 155–13054%
Is Claude as sure as it says?
Its graded picks by how sure it said it was. A well-sized pick at 70% hits about 70% of the time; the gold line is what it said.
Claude’s player props 22–10 69% right
On the games it writes up, it picks 2 or 3 of the game’s player props from the same list the other two AIs get, graded from the box score. A player who didn’t play is left out.
How it went 400 games · page 6 of 7
The AI (Claude) picks every NFL game and other leagues’ bigger games (ranked matchups, national TV, games Coalition members are picking) the day before; ChatGPT and Grok pick the same games from the same facts and the same instructions. Each pick is graded on the line it saw; a pick written again after the line moved is graded on the newer line. A tie on the winner is a push, except a draw in soccer, which is a miss. Times are Eastern. Research, not betting advice. How we grade
FD
IL
PU
GO
WC
NC
JS
TT
ME
IS
AR
DU
OA
MI
MI
FL
HD
PT
PE
WC
NE
MD
KC
DB
LA
LA
IC
JJ
AF
TB
CB
PS
LV
NY
MV
DL
AC
DC
TT
CB
BR
BB
ND
VI
OS
FS
NI
UN
AD
GS
WA
NE
MI
AR
GT
PI
MS
TE
OS
US
WS
SD
CL
FS
OS
AS
GS
SC
OK
NE
UC
PS
NM
SJ
VI
WF
UT
CI
FI
LI
UT
ND
CS
US
KA
TC
MI
IN
GE
AS
BY
NO
OR
ST
LO
TE
EC
KS
CO
AU
OM
RU
BG
WM
UA
SF
MA
BC
CA
NS