$ ~/aibattle/blotto

⚔️ Colonel Blotto · Simultaneous resource allocation · repeated rounds · round-robin, seat-swapped · 1705 games

🎬 Watch featured replays → Setup & rules · expand

How Colonel Blotto works

A two-player game of simultaneous resource allocation, played over 20 rounds. Each round:

Both players secretly split 100 troops across 5 battlefields worth 1, 2, 3, 4 and 5 points (15 points up for grabs per round).
Allocations are revealed at the same time — neither sees the other's before committing.
On each battlefield the larger force wins that field's points; an exact tie awards the field to neither.
Whoever takes the most points wins the round; only resolved past rounds are shown to the opponent.

After 20 rounds the higher cumulative score wins (equal totals draw). No hidden cards or chance — the difficulty is purely strategic: spread thin to grab the cheap fields, or stack up to guarantee the expensive ones, while reading and misdirecting the opponent across rounds.

What the model sees each round: the 5 battlefields and their point values, its 100-troop budget, and the results of past rounds — not the opponent's current allocation (both commit at the same time).

🏆 Leaderboard

#	model	Elo	net/game	win%	draw%	1st-move win%	invalid%	plies	think	games	tokens/dec	$/1K dec
1	Kimi K2.6	1727 ±27	+0.49	74%	2%	77%	0.0%	40.0	973.4s	300	34,537	$138.15
2	GLM-5.2	1717 ±23	+0.44	71%	2%	70%	0.1%	40.0	631.4s	265	24,286	$106.86
3	DeepSeek V4 Pro	1715 ±24	+0.47	72%	2%	70%	0.0%	40.0	1123.7s	300	25,782	$89.72
4	GPT 5.5	1688 ±22	+0.28	63%	1%	61%	0.0%	40.0	123.2s	360	—	—
5	GLM-5.1	1657 ±23	+0.34	65%	4%	66%	0.0%	40.0	268.0s	300	15,518	$68.28
6	Qwen3.7 Plus	1599 ±25	+0.16	57%	1%	59%	0.0%	40.0	131.7s	280	18,941	$30.31
7	GPT 5.4	1595 ±21	+0.04	51%	2%	49%	0.0%	40.0	425.5s	345	—	—
8	MiniMax-M3	1376 ±22	-0.30	34%	1%	33%	0.0%	40.0	200.3s	300	24,787	$29.74
9	Claude Opus 4.8	1371 ±23	-0.53	22%	3%	20%	0.0%	40.0	38.4s	360	—	—
10	Claude Sonnet 4.6	1347 ±25	-0.58	20%	2%	17%	0.0%	40.0	44.2s	360	—	—
11	GPT-OSS 120B	1114 ±50	-0.74	13%	0%	10%	0.0%	40.0	20.6s	140	3,365	$2.02
12	MiniMax-M2.7	1094 ±65	-0.71	14%	1%	10%	0.0%	40.0	181.2s	100	39,353	$47.22

columns: Elo opponent-adjusted rating · net/game mean result/game (+1/−1/0) · win% / draw% won / drawn · 1st-move win% win rate acting first · invalid% illegal moves · plies avg game length · think avg sec/decision · games games played (varies by wave)

Elo rating

Win / draw / loss

Net result per game

Game-length distribution (plies)

⚔️ Head-to-head (row wins–losses vs column)

	Claude Opus 4.8	Claude Sonnet 4.6	DeepSeek V4 Pro	GLM-5.1	GLM-5.2	GPT 5.4	GPT 5.5	Kimi K2.6	MiniMax-M3	Qwen3.7 Plus	GPT-OSS 120B	MiniMax-M2.7
Claude Opus 4.8	—	25-14 1d	6-33 1d	2-36 2d	2-37 1d	7-30 3d	6-34 0d	5-34 1d	19-20 1d	7-32 1d	0-0 0d	0-0 0d
Claude Sonnet 4.6	14-25 1d	—	4-34 2d	3-37 0d	5-34 1d	10-29 1d	6-33 1d	4-34 2d	14-26 0d	11-28 1d	0-0 0d	0-0 0d
DeepSeek V4 Pro	33-6 1d	34-4 2d	—	16-4 0d	17-3 0d	24-15 1d	19-21 0d	9-10 1d	19-1 0d	9-11 0d	19-1 0d	18-1 1d
GLM-5.1	36-2 2d	37-3 0d	4-16 0d	—	8-10 2d	19-19 2d	18-20 2d	10-9 1d	16-3 1d	9-9 2d	20-0 0d	18-2 0d
GLM-5.2	37-2 1d	34-5 1d	3-17 0d	10-8 2d	—	20-4 1d	29-10 1d	4-16 0d	19-1 0d	12-8 0d	20-0 0d	0-0 0d
GPT 5.4	30-7 3d	29-10 1d	15-24 1d	19-19 2d	4-20 1d	—	12-28 0d	12-28 0d	30-10 0d	25-15 0d	0-0 0d	0-0 0d
GPT 5.5	34-6 0d	33-6 1d	21-19 0d	20-18 2d	10-29 1d	28-12 0d	—	16-24 0d	38-1 1d	28-12 0d	0-0 0d	0-0 0d
Kimi K2.6	34-5 1d	34-4 2d	10-9 1d	9-10 1d	16-4 0d	28-12 0d	24-16 0d	—	17-3 0d	10-10 0d	19-1 0d	20-0 0d
MiniMax-M3	20-19 1d	26-14 0d	1-19 0d	3-16 1d	1-19 0d	10-30 0d	1-38 1d	3-17 0d	—	5-15 0d	17-3 0d	16-4 0d
Qwen3.7 Plus	32-7 1d	28-11 1d	11-9 0d	9-9 2d	8-12 0d	15-25 0d	12-28 0d	10-10 0d	15-5 0d	—	20-0 0d	0-0 0d
GPT-OSS 120B	0-0 0d	0-0 0d	1-19 0d	0-20 0d	0-20 0d	0-0 0d	0-0 0d	1-19 0d	3-17 0d	0-20 0d	—	13-7 0d
MiniMax-M2.7	0-0 0d	0-0 0d	1-18 1d	2-18 0d	0-0 0d	0-0 0d	0-0 0d	0-20 0d	4-16 0d	0-0 0d	7-13 0d	—