what is polymarket-bot-graveyard fr?

This project is a detailed, honest write up of six automated trading bots the creator built and shut down over three months on Polymarket, a prediction market platform. Instead of showing off a bot that made money, this repository documents failure: none of the six bots ever found a real edge, and only a small amount of real money, about $45, was ever risked before the trading account got banned for sending too many orders too quickly. Every version after that ran only in a simulated, paper trading mode. Each bot tried a different strategy: copying other traders, market making, following short term Bitcoin price swings using multiple AI models to vote on a direction, scanning hundreds of slower moving markets, and using an oracle based price model. Every one of them ran into the same wall. Professional trading firms already price these markets efficiently within seconds, so there was no real gap left for a small trader to profit from. AI models were often too slow to act before prices moved. Static pricing models broke down once market conditions shifted. The last version in the series took a different approach: instead of trading right away, it measured whether a signal actually predicted outcomes correctly before letting the bot use real money. It concluded honestly that no edge existed, and refused to trade. The author treats that correct refusal as the actual success of the project, since it stopped further losses. The repository is organized as a step by step journal, one chapter per bot, plus a final analysis document explaining the common mistakes across all six attempts. Code snapshots for each bot are included with secrets removed. It is released under the MIT license for code and Creative Commons for the documentation. Anyone considering building a similar trading bot may find this useful as a warning about what does not work and why.