Recycle Bin
4 items · projects I killed and why. Deleted, not forgotten.
A personal site usually shows the wins. Here are the things that didn't make it — kept because the postmortems taught me more than the successes did.
Historical project notes, clarified on 24 October 2026. They retain the failed outcomes without presenting unverified figures, simulations, or later plans as independently established results.
An autonomous prediction-and-arbitrage agent meant to read odds across books, find edges, and learn from outcomes.
Why I killed it: The earlier postmortem described an incomplete system: advertised coverage exceeded the working implementation, and unused infrastructure accumulated. Those historical counts have not been re-audited for this page.
What it taught me: Don't build infrastructure ahead of a validated need. The part that actually worked — ingesting bookmaker data — was rescued and rebuilt as a focused, standalone service (Siphon).
A soccer prediction system aiming to beat the market on mainstream 1X2 lines.
Why I killed it: Earlier project notes reported failure against a closing-line benchmark. This page does not independently reproduce that evaluation or claim the later design succeeded.
What it taught me: A redesigned target is a new hypothesis, not a demonstrated edge. Freeze the comparison policy and require held-out and forward evidence before acting.
A dozen trading strategies on crypto prediction markets, automated one after another.
Why I killed it: These iterations lost money despite encouraging backtests. The earlier page gave a specific loss total; that figure is not independently verified here.
What it taught me: This is where 'validate before automating' was actually learned — the hard way, in cash. A machine executes a bad idea faster than you ever could by hand.
Market-making and momentum bots across a couple of venues.
Why I killed it: The experiments did not establish a durable live edge worth the operational effort. Backtests and model signals were not sufficient proof.
What it taught me: Keep the evidence collector, retire the bot. Measuring honestly is worth more than a fragile edge.
Note: emptying the Recycle Bin does not undo the lessons. Believe me, I've tried.