← arena

results 50 games

9,484 decisions · 4,231 failed attempts · $5.54
50games
9,484decisions put to a model
23,911moves played
14.23Mtokens
$5.54logged cost
3.3hwall clock

standings

modelwinsrateinterval (adj. 4 models)medianp90slowestfailedrateout/callcost
deepseek-v4-flash19/5038%23.1–55.6%49.1s174.4s1866s3,63560%1899$1.27
gpt-5.6-luna19/5038%23.1–55.6%5.4s8.6s31s20%302$1.23
gemini-3.7-flash7/5014%5.8–30.2%2.9s4.2s26s30%186$1.43
hy35/5010%3.5–25.4%13.9s41.8s137s59121%1107$1.62
Chance is 25% with 4 seats. Intervals are Bonferroni-adjusted — testing every model against chance is several tests, and an unadjusted 95% interval crowns a winner about one run in four by luck alone. Timings here are whole-decision wall clock including retries and backoff, not model latency — these logs predate the split, so a model that retried often reads far slower than it is.

per model

median seconds per decision, retries included

deepseek-v4-flash49.1s
hy313.9s
gpt-5.6-luna5.4s
gemini-3.7-flash2.9s

failed attempts

deepseek-v4-flash60%
hy321%
gemini-3.7-flash0%
gpt-5.6-luna0%

output tokens per decision

deepseek-v4-flash1899
hy31107
gpt-5.6-luna302
gemini-3.7-flash186

cost

hy3$1.62
gemini-3.7-flash$1.43
deepseek-v4-flash$1.27
gpt-5.6-luna$1.23

failures

4,231 failed attempts against 9,484 successful decisions. Every one was retried until the model answered — no move was ever substituted.
failurecountshare
timed out on our deadline4,18899%
unparseable reply411%
connect error10%
remoteprotocol error10%

retry depth

deepest chain: 17 attempts
attempt 12,298
attempt 21,144
attempt 3490
attempt 4201
attempt 556
attempt 616
attempt 79
attempt 87

seat balance

Turn order is worth real points, so seats rotate per match. These should be even.
modelseat 1seat 2seat 3seat 4
deepseek-v4-flash13121213
gemini-3.7-flash12131312
gpt-5.6-luna12121313
hy313131212

game length

median 148 turns, range 58–355; median 450 moves. Only 40% of moves went to a model — forced moves with a single legal action skip the API.

every record

Each game replays move by move from its own log. Raw files are served as text.
gameturnswinnerraw
m000_g0234gemini-3.7-flashmoves · decisions · log
m001_g0194gemini-3.7-flashmoves · decisions · log
m002_g091gpt-5.6-lunamoves · decisions · log
m003_g0127gpt-5.6-lunamoves · decisions · log
m004_g0189hy3moves · decisions · log
m005_g0134hy3moves · decisions · log
m006_g0218deepseek-v4-flashmoves · decisions · log
m007_g0122hy3moves · decisions · log
m008_g0154gpt-5.6-lunamoves · decisions · log
m009_g0212gemini-3.7-flashmoves · decisions · log
m010_g0105deepseek-v4-flashmoves · decisions · log
m011_g073gpt-5.6-lunamoves · decisions · log
m012_g0165gpt-5.6-lunamoves · decisions · log
m013_g0100gpt-5.6-lunamoves · decisions · log
m014_g0156gemini-3.7-flashmoves · decisions · log
m015_g0132gpt-5.6-lunamoves · decisions · log
m016_g0157gpt-5.6-lunamoves · decisions · log
m017_g0110deepseek-v4-flashmoves · decisions · log
m018_g0131gemini-3.7-flashmoves · decisions · log
m019_g0150gpt-5.6-lunamoves · decisions · log
m020_g0248deepseek-v4-flashmoves · decisions · log
m021_g0183deepseek-v4-flashmoves · decisions · log
m022_g0128gpt-5.6-lunamoves · decisions · log
m023_g0146deepseek-v4-flashmoves · decisions · log
m024_g0194gemini-3.7-flashmoves · decisions · log
m025_g0214gpt-5.6-lunamoves · decisions · log
m026_g0106gpt-5.6-lunamoves · decisions · log
m027_g070gpt-5.6-lunamoves · decisions · log
m028_g0147gpt-5.6-lunamoves · decisions · log
m029_g0355deepseek-v4-flashmoves · decisions · log
m030_g0143hy3moves · decisions · log
m031_g0137gpt-5.6-lunamoves · decisions · log
m032_g0198deepseek-v4-flashmoves · decisions · log
m033_g0207deepseek-v4-flashmoves · decisions · log
m034_g058gemini-3.7-flashmoves · decisions · log
m035_g0204gpt-5.6-lunamoves · decisions · log
m036_g0114hy3moves · decisions · log
m037_g0139deepseek-v4-flashmoves · decisions · log
m038_g0241deepseek-v4-flashmoves · decisions · log
m039_g0160deepseek-v4-flashmoves · decisions · log
m040_g0180deepseek-v4-flashmoves · decisions · log
m041_g0129gpt-5.6-lunamoves · decisions · log
m042_g0102deepseek-v4-flashmoves · decisions · log
m043_g0100gpt-5.6-lunamoves · decisions · log
m044_g0151deepseek-v4-flashmoves · decisions · log
m045_g0214deepseek-v4-flashmoves · decisions · log
m046_g0187deepseek-v4-flashmoves · decisions · log
m047_g0113gpt-5.6-lunamoves · decisions · log
m048_g0184deepseek-v4-flashmoves · decisions · log
m049_g0109deepseek-v4-flashmoves · decisions · log