I’m closing my prediction bot experiment after 30 days.
Results: basically 50/50.
Here’s what actually happened ↓
When I started this, I never expected it to be “plug a bot and it prints money”. It was always an experiment.
After ~1 month of running it, the reality is pretty simple:
results averaged out around 50/50. Some days the bot could do 3× the bankroll, other days it would give it all back. In the long run it mostly balanced out.
Prediction markets are much harder than they look.
Without extremely fast reaction times, serious infrastructure, or some kind of informational edge, it's very difficult to consistently beat the market.
Most bots that actually make money seem to rely on arbitrage or ultra-short trades — hundreds of transactions per minute capturing tiny spreads.
And many of the “impressive bots” you see online are honestly just fancy screenshots without real long-term results.
I also tested integrating an AI agent into the system, but over time it behaved roughly the same: interesting moments, but no real edge.
So I’m calling the experiment finished.
That said, the main goal was achieved:
I learned a lot about building agents, automation pipelines, and running real experiments in live systems.
And honestly… if it also made money, that would’ve been nice too 😄
For now I’ll move on to other personal projects.
More experiments coming.
English