An AI beat Stratego's most decorated player 15 to 1 with 4 draws, and the paper backs it.
The "few thousand dollars" is the final training run, not the whole project.
Ars Technica reports Ataraxos beat Pim Niemeijer, arguably the best Stratego player of all time, 15 games to one with four draws, trained for a few thousand dollars.
Before you read on. Your call?
VERIFIED
15-1-4
The team's paper reports the 20-game series and that score, and 38 wins in 40 games against other players. It puts the final training run under $8,000.
The twist
The $3M to $4.5M DeepNash figure rests on a DeepNash author's recollection, and the two AIs never played because DeepNash's code no longer works.
There’s more to this story.
Membership opens the full investigation, the strongest counterargument and what to do with what you’ve learned.
Start your free month →First membership: 30 days free, then A$89 a year. One introductory trial per customer. Card required; renews annually until cancelled. Cancel before the trial ends to avoid the first charge. Already a member? Sign in
Couldn't check your access. That's on us.
Receipts
- Supports arstechnica.com:
player of all time, 15 games to one, with four draws
- Supports arstechnica.com:
And it took just 16 GPUs and a few thousand dollars to train it.
- Context arstechnica.com:
That balancing act, the team explains, is what stumped earlier AIs like DeepMind’s DeepNash, introduced in 2022.
- Supports arxiv.org:
In a 20-game series, Ataraxos defeated the most decorated Stratego player of all time
- Supports arxiv.org:
Ataraxos won the series with 15 wins, 1 loss, and 4 draws
- Supports arxiv.org:
over 600 weeks as the #1 ranked player (the most of all time)
- Supports arxiv.org:
Such a run costs less than $8,000 at current prices
- Context arxiv.org:
such a training run would roughly cost between $3,000,000 and $4,500,000
- Context arxiv.org:
To the recollection of the corresponding author with whom we spoke
- Context arxiv.org:
DeepMind responded that it would not be possible as the code for DeepNash is no longer functional.
- Context arxiv.org:
the outcomes of the games were far from independently and identically distributed
- Context arxiv.org:
Pim was paid $1,000 for participating in the evaluation
- Supports arxiv.org:
Across 40 such games, Ataraxos recorded a vertiginously high 95% effective win rate (38 wins, 2 losses, 0 draws)
- Supports arxiv.org:
doing so requires not an industrial budget, but merely a few thousand dollars
- Refutes news.ycombinator.com:
Just 16 GPUs, and a few thousand dollars?
- Supports arxiv.org:
Ataraxos achieved this result while only costing a few thousand dollars to train.
Open the Receipts Pack → What each source proves, every figure traced, and what would change our verdict.