On Nov 3 I wrote about paper-testing an AI memecoin trading bot for two days. I kept going. This is the honest result after two weeks, with all the code and raw data public.
Short version: nothing I built made money, and the data shows why.
The setup
Trade Desk scouts brand-new Solana coins every minute, screens each one with my own safety checks (Solana Pulse token_safety: mint and freeze authority, locked liquidity, holder concentration, a test sell), then paper-trades them. No real money. Fills include fees and price impact.
I ran different "lanes" side by side, each with $1,000 of paper money: a careful main desk, a fast lane that sells in 30 minutes, two ladder strategies that sell in steps on the way up, a momentum lane, and one that copies "smart" wallets.
The result
| Paper runs, Nov 1–15 | 16 |
| Closed trades | 2,639 |
| Winning trades | 909 (34%) |
| Total | −$2,665 |
| Runs that ended in profit | 0 |
Some runs looked great for a few hours. The main desk was up 12% after its first 6 hours. Then two different coins, both named AODF, crashed 99% within half an hour of each other, and a third (XTO) followed. Three rugs caused $224 of the first run's $238 loss. A −6% stop-loss means nothing when the price goes from $2.03 to $0.0002 in one sell: the fills came in at −70% to −100%.
It wasn't just the live bots. I replayed 16,071 coins on 1-minute candles with every exit rule I could think of (120 combinations of take-profit, stop and time limit). Every exit lost: the typical trade lost 11.6% to 17.9%. Six brand-new strategy ideas (dip buys, breakouts, second waves, capitulation bounces) tested on 5,705 coins: zero leads.
So who wins on a new coin?
I pulled every trade in the first two hours of 100 new coins: 105,857 transactions, every wallet's profit and loss.
| Who | Wallets | Profitable | Total |
|---|---|---|---|
| Got tokens before trading opened | 2,828 | 95.8% | +1,816 SOL |
| Pool creators | 49 | 59.2% | +124 SOL |
| Bought in the launch slot | 167 | 34.7% | −187 SOL |
| First buy within 10 s | 1,685 | 18.7% | −584 SOL |
| First buy 10–60 s | 8,033 | 4.1% | −6,417 SOL |
| First buy 1–10 min | 18,410 | 1.1% | −1,709 SOL |
| First buy after 10 min | 382 | 0.5% | −31 SOL |
The money goes to insiders. Wallets that held tokens before trading opened made +1,816 SOL. Every group that bought after launch lost, about 8,900 SOL combined. By the time any bot (mine included) sees a coin in a public scan, it's already late.
And "smart wallets"? 51 wallets won on 3+ early coins. On the next coins they won only 39% of the time and lost 12.6 SOL in total. Mostly luck.
The one thing that did work
The safety checks. Of the coins Pulse rated AVOID, 82% were dead (down 90%+ or no longer priced) soon after. Of the 15 rated SAFE, zero died. Telling a rug from a real token is possible. Making money trading new tokens is a different problem, and I couldn't solve it.
What I did with that
- I retired five of the lanes. One (Ladder A) still runs so the record keeps growing in public: 986 more trades since Nov 9, −$581.
- I stopped looking for a memecoin edge and kept building the part that works: free checkers for people (Rug Check for tokens, Drain Check for drained wallets) and the Pulse API for AI agents that sign transactions.
- I published everything, including the parts that make me look bad: trading results page (memecoins, crypto, stocks, forex) and the raw data and scripts.
If someone shows you a bot that turned $280 into $3,412 overnight, ask for the losing runs too.
Top comments (0)