https://chipzen.ai/stack Original writing on competitive AI bots and the arena they compete in - from the team building Chipzen. en-US Fri, 04 Sep 2026 21:24:59 GMT https://chipzen.ai/stack/telling-11-poker-bots-apart https://chipzen.ai/stack/telling-11-poker-bots-apart Sun, 23 Aug 2026 00:00:00 GMT We put 247,946 heads-up hands from 11 LLM-coded poker bots on Kaggle, hole cards on every hand. A plain logistic regression can tell which bot is playing from ten hands about 74% of the time. Here's what's in the box and why identity leaks so fast. dataset kaggle llm-bots benchmark poker https://chipzen.ai/stack/entering-is-publishing https://chipzen.ai/stack/entering-is-publishing Thu, 30 Jul 2026 00:00:00 GMT The leaderboard illusion happened because private variant testing was possible. On a live ladder it isn't: measuring against the field is playing the field, and playing the field is public. Transparency by construction, not by policy. evals benchmarks transparency poker design https://chipzen.ai/stack/ten-llms-one-poker-prompt https://chipzen.ai/stack/ten-llms-one-poker-prompt Fri, 24 Jul 2026 00:00:00 GMT We gave a dozen frontier models the same prompt: write a heads-up poker bot that wins. Every one produced code that runs and plays legal poker. On the live ladder, the best wins about three matches in four and the worst about one in five. Here is what that gap is made of. benchmarks llm poker evaluation results https://chipzen.ai/stack/heads-up-is-the-real-test https://chipzen.ai/stack/heads-up-is-the-real-test Fri, 24 Jul 2026 00:00:00 GMT 6-max No-Limit Hold'em is more popular and more forgiving. Heads-up is harder, less forgiving, and where bot strength actually shows up. Why Chipzen runs HU first. positioning technical poker https://chipzen.ai/stack/beyond-slumbot https://chipzen.ai/stack/beyond-slumbot Mon, 13 Jul 2026 00:00:00 GMT Slumbot is still the standard free benchmark for heads-up bots. How to test against it properly, what a good result means, and what a single fixed opponent can't tell you. slumbot benchmarks technical poker https://chipzen.ai/stack/how-to-build-a-poker-bot-python https://chipzen.ai/stack/how-to-build-a-poker-bot-python Mon, 13 Jul 2026 00:00:00 GMT From a 20-line rule bot to one that computes real equity with Monte Carlo, standard library only. Working code for each step, and where to make it play ranked matches. tutorial technical python poker https://chipzen.ai/stack/benchmark-that-fights-back https://chipzen.ai/stack/benchmark-that-fights-back Mon, 22 Jun 2026 00:00:00 GMT LLM benchmarks are getting saturated and gameable. A competitive arena is a different kind of test: outcome-based, adversarial, and harder to game. We gave five flagship models the same prompt to one-shot a poker bot, then let the results fight it out on a live ladder. benchmarks llm evaluation arena opponent-modeling poker https://chipzen.ai/stack/benchmark-example-gemini https://chipzen.ai/stack/benchmark-example-gemini Mon, 22 Jun 2026 00:00:00 GMT A worked example for The Benchmark That Fights Back: Gemini's full, unedited one-shot response to our poker-bot prompt - its requirements.txt and bot.py. benchmarks llm example poker sdk https://chipzen.ai/stack/the-arena-is-open https://chipzen.ai/stack/the-arena-is-open Tue, 16 Jun 2026 00:00:00 GMT We argued the thing computer poker was missing wasn't another solver - it was a live place to test opponent modeling and exploitation against real, evolving bots. That place is now open. Chipzen is in public beta. beta platform opponent-modeling exploitation poker https://chipzen.ai/stack/poker-isnt-done https://chipzen.ai/stack/poker-isnt-done Mon, 08 Jun 2026 00:00:00 GMT Libratus and Pluribus didn't solve poker - they showed approximate-Nash play at superhuman level. Modeling opponents, exploiting them, and even measuring exploitation are still open research problems. positioning poker opponent-modeling exploitation research-frontier https://chipzen.ai/stack/why-acpc-died https://chipzen.ai/stack/why-acpc-died Mon, 01 Jun 2026 00:00:00 GMT The Annual Computer Poker Competition ran for over a decade, produced the bots that beat the best humans, then quietly stopped in 2018. What happened - and what its disappearance left behind. history research poker acpc https://chipzen.ai/stack/mccfr-explained https://chipzen.ai/stack/mccfr-explained Mon, 25 May 2026 00:00:00 GMT MCCFR - Monte Carlo Counterfactual Regret Minimization - is how every serious poker bot of the last decade learned to play. How the algorithm works, and why it dominates imperfect-information games. technical mccfr cfr poker