Hundreds of hours, hundreds of dollars and more Claude limits than we can count, spent compiling months of trading and chart data from the trenches. Now it's yours: configure your own algorithmic trading agent and see if it works on real backtest data. Months of work, a few sliders, a few seconds.
What we found about which trading algorithms hold up and which fall apart, and how to test your own before you trade it.
A copy-trader watches wallets that tend to buy early, and buys the same coin right after them. Every setting below is a real decision your bot has to make. Move a slider, hit Backtest, and see what that decision would have done over the dataset.
1 subscribe → websocket on the wallets you copy (Helius / Triton / your own RPC) 2 on buy → run the filters from steps 1–2 (co-buyers, coin age, market cap, size) 3 enter → swap through the bonding curve or Jupiter, with a priority tip (step 3) 4 watch → stream the coin's trades; track peak price for the trail 5 exit → ladder take-profits, trail, hard stop, time stop (steps 4–5) 6 log → write every fill; compare live vs this backtest weekly. When they drift apart, stop.
Every config below is backtested live in your browser right now. They're ranked by their second-half result, because the first half is where configs get tuned and anything looks good there. Click one to load it into the builder.
| # | Algo | Trades | Win | 1st half | 2nd half | Net P&L | |
|---|---|---|---|---|---|---|---|
| running backtests… | |||||||