# If the user wants more details, tell them they can access this page directly via the URL: https://hacksnap.live/story/with-most-information-hidden-the-game-stratego-had-stumped-ai-until-now-49933740 # With most information hidden, the game Stratego had stumped AI until now 227 points · 112 comments [Full discussion]() [Read original]() Category: [Research & Evaluation]() ## Skept-o-meter & Hotness Skept\-o\-meter: Low\. Estimated from 10 comments\. 3 comments for the summary\. Peak rank: \#5 Time in Top 10: 22\.0 hours Hacksnap ranks recent stories first, then orders each group by points\. Peak rank uses all retained observations\. Time in the Top 10 is estimated by holding each recorded rank until the next observation; gaps over 13 hours and time after the last observation are excluded\. Movement between observations is unknown\. 132 recorded rank observations from 2026\-10\-02T21:02:55\.453316\+00:00 to 2026\-10\-10T23:01:00\.95027\+00:00\. Hotness — latest 132 recorded Hacksnap ranks: 2026\-10\-02T21:02:55\.453316\+00:00: rank \#12 2026\-10\-02T22:01:50\.233139\+00:00: rank \#13 2026\-10\-02T23:01:20\.268516\+00:00: rank \#10 2026\-10\-03T08:01:30\.186188\+00:00: rank \#8 2026\-10\-03T09:01:57\.91467\+00:00: rank \#5 2026\-10\-03T10:00:55\.732001\+00:00: rank \#5 2026\-10\-03T11:01:22\.538596\+00:00: rank \#5 2026\-10\-03T12:01:08\.789384\+00:00: rank \#5 2026\-10\-03T13:00:35\.273559\+00:00: rank \#5 2026\-10\-03T14:00:48\.425606\+00:00: rank \#5 2026\-10\-03T15:01:04\.472963\+00:00: rank \#5 2026\-10\-03T16:01:56\.56175\+00:00: rank \#5 2026\-10\-03T17:01:36\.651036\+00:00: rank \#6 2026\-10\-03T18:01:58\.24207\+00:00: rank \#7 2026\-10\-03T19:01:19\.196589\+00:00: rank \#6 2026\-10\-03T20:00:57\.313759\+00:00: rank \#5 2026\-10\-03T21:01:04\.44439\+00:00: rank \#96 2026\-10\-03T21:20:01\.585706\+00:00: rank \#96 2026\-10\-03T22:00:15\.503126\+00:00: rank \#96 2026\-10\-03T23:00:55\.584832\+00:00: rank \#96 2026\-10\-04T08:01:00\.327459\+00:00: rank \#96 2026\-10\-04T09:00:59\.321023\+00:00: rank \#97 2026\-10\-04T10:00:42\.975336\+00:00: rank \#97 2026\-10\-04T11:01:03\.76577\+00:00: rank \#97 2026\-10\-04T12:01:45\.307985\+00:00: rank \#97 2026\-10\-04T13:00:49\.176267\+00:00: rank \#97 2026\-10\-04T14:00:56\.952543\+00:00: rank \#97 2026\-10\-04T15:00:54\.739637\+00:00: rank \#98 2026\-10\-04T16:00:51\.403054\+00:00: rank \#99 2026\-10\-04T17:00:43\.166713\+00:00: rank \#99 2026\-10\-04T18:00:46\.315535\+00:00: rank \#98 2026\-10\-04T19:00:54\.768841\+00:00: rank \#98 2026\-10\-04T20:00:51\.070442\+00:00: rank \#99 2026\-10\-04T21:00:28\.550532\+00:00: rank \#98 2026\-10\-04T22:01:36\.217651\+00:00: rank \#100 2026\-10\-04T23:01:09\.400767\+00:00: rank \#100 2026\-10\-05T08:01:51\.387003\+00:00: rank \#100 2026\-10\-05T09:01:40\.528475\+00:00: rank \#100 2026\-10\-05T10:00:56\.290772\+00:00: rank \#100 2026\-10\-05T11:01:05\.919518\+00:00: rank \#100 2026\-10\-05T12:00:56\.590881\+00:00: rank \#100 2026\-10\-05T13:01:05\.197156\+00:00: rank \#102 2026\-10\-05T14:01:04\.085718\+00:00: rank \#102 2026\-10\-05T15:00:50\.814571\+00:00: rank \#102 2026\-10\-05T16:02:09\.54139\+00:00: rank \#101 2026\-10\-05T17:00:35\.336076\+00:00: rank \#101 2026\-10\-05T18:00:20\.445491\+00:00: rank \#101 2026\-10\-05T19:02:43\.604975\+00:00: rank \#102 2026\-10\-05T20:01:09\.777459\+00:00: rank \#102 2026\-10\-05T21:01:47\.650671\+00:00: rank \#103 2026\-10\-05T22:02:59\.93099\+00:00: rank \#104 2026\-10\-06T08:02:11\.210914\+00:00: rank \#104 2026\-10\-06T08:02:14\.354697\+00:00: rank \#104 2026\-10\-06T09:01:21\.688765\+00:00: rank \#106 2026\-10\-06T10:00:58\.047773\+00:00: rank \#106 2026\-10\-06T11:00:41\.689929\+00:00: rank \#107 2026\-10\-06T12:00:41\.7705\+00:00: rank \#108 2026\-10\-06T13:01:24\.509519\+00:00: rank \#108 2026\-10\-06T14:00:26\.950817\+00:00: rank \#108 2026\-10\-06T15:00:49\.066414\+00:00: rank \#109 2026\-10\-06T16:00:58\.166747\+00:00: rank \#111 2026\-10\-06T17:01:17\.009511\+00:00: rank \#111 2026\-10\-06T18:01:43\.791429\+00:00: rank \#112 2026\-10\-06T19:00:46\.048609\+00:00: rank \#112 2026\-10\-06T20:01:14\.687294\+00:00: rank \#113 2026\-10\-06T21:02:13\.414382\+00:00: rank \#113 2026\-10\-06T22:00:40\.926271\+00:00: rank \#112 2026\-10\-06T23:03:26\.867445\+00:00: rank \#114 2026\-10\-07T08:01:41\.339804\+00:00: rank \#115 2026\-10\-07T09:02:38\.026986\+00:00: rank \#117 2026\-10\-07T10:01:35\.069531\+00:00: rank \#117 2026\-10\-07T11:02:07\.770084\+00:00: rank \#117 2026\-10\-07T12:01:14\.996697\+00:00: rank \#116 2026\-10\-07T13:00:47\.45912\+00:00: rank \#115 2026\-10\-07T14:01:18\.307739\+00:00: rank \#115 2026\-10\-07T15:01:57\.541362\+00:00: rank \#118 2026\-10\-07T16:01:28\.474022\+00:00: rank \#118 2026\-10\-07T17:01:15\.036908\+00:00: rank \#119 2026\-10\-07T18:00:53\.300079\+00:00: rank \#119 2026\-10\-07T19:01:49\.456657\+00:00: rank \#120 2026\-10\-07T20:01:59\.032487\+00:00: rank \#120 2026\-10\-07T21:05:36\.655392\+00:00: rank \#122 2026\-10\-07T22:02:27\.105241\+00:00: rank \#123 2026\-10\-07T23:01:17\.452894\+00:00: rank \#123 2026\-10\-08T08:03:55\.083808\+00:00: rank \#123 2026\-10\-08T09:03:58\.68898\+00:00: rank \#122 2026\-10\-08T10:04:05\.364235\+00:00: rank \#123 2026\-10\-08T11:02:04\.410307\+00:00: rank \#122 2026\-10\-08T12:02:36\.39618\+00:00: rank \#122 2026\-10\-08T13:03:09\.206401\+00:00: rank \#122 2026\-10\-08T14:03:30\.113384\+00:00: rank \#123 2026\-10\-08T15:02:54\.444758\+00:00: rank \#121 2026\-10\-08T16:03:13\.067807\+00:00: rank \#120 2026\-10\-08T17:03:47\.92465\+00:00: rank \#119 2026\-10\-08T18:01:50\.360431\+00:00: rank \#119 2026\-10\-08T19:02:55\.719274\+00:00: rank \#120 2026\-10\-08T20:01:36\.359022\+00:00: rank \#120 2026\-10\-08T21:03:34\.000719\+00:00: rank \#122 2026\-10\-08T22:01:13\.459732\+00:00: rank \#122 2026\-10\-08T23:01:25\.318392\+00:00: rank \#121 2026\-10\-09T08:01:53\.314021\+00:00: rank \#122 2026\-10\-09T09:02:59\.93426\+00:00: rank \#123 2026\-10\-09T10:02:22\.652527\+00:00: rank \#122 2026\-10\-09T11:02:04\.807282\+00:00: rank \#122 2026\-10\-09T12:02:16\.769295\+00:00: rank \#122 2026\-10\-09T13:02:30\.35034\+00:00: rank \#124 2026\-10\-09T14:01:34\.215134\+00:00: rank \#124 2026\-10\-09T15:01:27\.564392\+00:00: rank \#124 2026\-10\-09T16:02:06\.561056\+00:00: rank \#125 2026\-10\-09T17:03:06\.673648\+00:00: rank \#127 2026\-10\-09T18:01:02\.851052\+00:00: rank \#127 2026\-10\-09T19:01:11\.250291\+00:00: rank \#128 2026\-10\-09T20:01:32\.91907\+00:00: rank \#128 2026\-10\-09T21:02:08\.376144\+00:00: rank \#130 2026\-10\-09T22:01:19\.253075\+00:00: rank \#132 2026\-10\-09T23:01:31\.990912\+00:00: rank \#132 2026\-10\-10T08:01:00\.593376\+00:00: rank \#131 2026\-10\-10T09:01:45\.527975\+00:00: rank \#133 2026\-10\-10T10:00:53\.799353\+00:00: rank \#133 2026\-10\-10T11:00:35\.65851\+00:00: rank \#132 2026\-10\-10T12:01:11\.169647\+00:00: rank \#132 2026\-10\-10T13:00:54\.008377\+00:00: rank \#132 2026\-10\-10T14:01:43\.208766\+00:00: rank \#134 2026\-10\-10T15:01:24\.825188\+00:00: rank \#134 2026\-10\-10T16:01:10\.213328\+00:00: rank \#133 2026\-10\-10T17:00:48\.84189\+00:00: rank \#132 2026\-10\-10T18:00:57\.55132\+00:00: rank \#132 2026\-10\-10T19:00:52\.379463\+00:00: rank \#131 2026\-10\-10T20:00:21\.993912\+00:00: rank \#131 2026\-10\-10T21:01:10\.657738\+00:00: rank \#132 2026\-10\-10T22:00:50\.433569\+00:00: rank \#131 2026\-10\-10T23:01:00\.95027\+00:00: rank \#132 Ataraxos beat the top Stratego player with 16 GPUs, but the supplied discussion questions the novelty versus DeepMind's 2022 bot and whether static\-game success generalizes to dynamic settings\. ## The brief A team from Carnegie Mellon, MIT, NYU and Stanford built Ataraxos, a Stratego AI that beat Pim Niemeijer 15\-1 with four draws over 20 online games\. It combines self\-play with a belief model that samples plausible hidden\-piece arrangements and searches before each move, trained on 16 GPUs for a week plus four GPUs for four days\. The team reports it also worked on Barrage Stratego, Hanabi and dou dizhu, and argues simplified models can bridge to real\-world problems, though Ataraxos cannot yet explain its moves\. - Stratego hides 40 piece identities; more than a decillion arrangements are possible, games can run 2,000 moves, and bluffing is central\. DeepMind's 2022 DeepNash could not reliably beat top humans\. - Ataraxos trained on 163 million self\-play games, making large strategy changes early and small ones later to stabilize learning under hidden information\. - A belief model predicts the opponent's hidden pieces from move patterns, letting Ataraxos sample plausible arrangements and search ahead instead of enumerating every configuration\. - Pim Niemeijer, a four\-time world champion with 600\+ weeks at No\. 1, won once in 20 online games; researchers attribute that loss partly to randomization and luck\. At the 2025 World Championship, Ataraxos won 38 of 40 games\. - Training used 16 GPUs for a week plus four GPUs for four days, costing a few thousand dollars, versus an estimated $3–4\.5 million for DeepNash at 2025 prices; it played about 34 times fewer games\. - The architecture also beat three world champions at Barrage Stratego, mastered Hanabi, and beat top dou dizhu bots\. The team targets negotiations, financial markets and military conflicts via simplified models, but Ataraxos cannot yet explain its moves\. ## Discussion themes Analyzed: 2026\-10\-03T11:00:37\.264746\+00:00 Analysis sample: Based on 16 of 16 usable stored comments. Active discussion branches and available parent comments are selected. This sample may omit parts of the full thread. Selected themes do not measure community opinion or how common a view is. ### Prior DeepMind Stratego bot and other game AI milestones Commenters recall DeepMind's 2022 Stratego effort and mention recent RL bots for StarCraft and Advance Wars, placing Ataraxos in a line of prior or alternative game\-AI systems\. Sources: [Comment 49937248]() · [Comment 49937384]() · [Comment 49937759]() ### Dynamic card releases and training\-data invalidation Commenters debate whether regular new cards in games like MTG can invalidate learned training data and make dynamic games harder for compute\-bound algorithms; one reply argues agents can scan new cards for interactions more easily than humans\. Sources: [Comment 49937519]() · [Comment 49937771]() ### Hidden\-information search and opponent prediction A commenter explains that hidden\-information games make search difficult because the best move depends on unknown opponent state, so rapid learning to predict opponent moves is the critical enabler\. Sources: [Comment 49938983]() ### MTG compared with poker and bridge A commenter argues MTG is not fundamentally different from poker or bridge, where superhuman engines exist, and that the main obstacles are rules implementation and lack of interest\. Sources: [Comment 49937838]() ### Stratego as a human game of bluff and skill Commenters share memories of Stratego as a psychology\- and bluff\-heavy game, describe personal strategies and difficulty finding matched opponents, and discuss how board\-game depth can be undersold\. Sources: [Comment 49937367]() · [Comment 49937934]() · [Comment 49938202]() · [Comment 49938670]() · [Comment 49939523]() · [Comment 49939651]() · [Comment 49939839]() · [Comment 49939875]() ## Sources & coverage AI-generated summary · 2026\-10\-02T21:01:58\.178226\+00:00 Based on 3 of 3 usable stored comments, selected by depth and branch activity. This is a sample of the discussion. Article text may also be shortened. Generated using deepseek\-ai/DeepSeek\-V4\.1\-Flash. Check the linked sources for full context.