The four best coins made 5.9 cents in week one. Here is what a fresh test showed
We picked the best 4 of 9 coins, then tested them on trades nobody had seen.
Every rule we run gets judged on trades nobody had seen when it was chosen. This is why we insist on that, and it is the most useful thing a backtester can learn cheaply.
In early October our sniper, which buys a late favourite at about 90 cents, had run for a week across nine coins. Four coins stood out: NEAR, SOL, ZEC and DOGE. Together they made +5.9 cents a contract on 115 fills, with only 2 losses. Bitcoin and XRP lost money.
The tempting move is to drop the losers and trade the winners. We wanted to, too. But picking the best 4 of 9 after the fact flatters them, so we wrote a pass mark down and tested those four again, counting only trades from that day on.
What happened next
| Group | First week | Every week after |
|---|---|---|
| The four best coins | 115 fills, +5.94¢, 2 losses | 675 fills, +0.01¢, 52 losses |
| The other five | 280 fills, −0.65¢ | 883 fills, −0.64¢ |
The other five did what they had always done. The four stars fell back to about nothing. ZEC had no losses in its first week, then lost on 9% of its trades in the fresh test.
The separate fresh test on those four coins made +1.27 cents a contract over 687 fills, which is positive and still not enough. It needed a loss rate under 6% and got 6.55%, and it needed a t score of 2 and got 1.34. So it failed and closed on 30 October.
The best coin was not in the top four
After the first week, the strongest coin in the group was HYPE at +2.07 cents. In week one it ranked sixth out of nine. A week of results told us very little about which coins would lead.
What to do with this
- Choosing the winners from a test and then quoting that test is not evidence. Their numbers are part luck, and the luck does not carry forward.
- Decide the bar first, then test on trades that came after the choice.
- Expect the number to shrink. If it survives, you have something.
Two coins from that group still look interesting, and we are not calling them winners, for the same reason. We are still hunting, and this is the filter that keeps us honest while we do. The rule behind our live bot is going through the same thing now: real orders, one contract each, judged at 60 settled trades on a bar written first.
Members get the appendix: all nine coins, the exact rule and the stricter fill recheck.
Want to run rules like this yourself? Try Ping free for 7 days, no card needed. It counts a paper fill only when a real trade proves it could have filled.
Where this comes from: our own paper lanes, 2 to 30 October 2026, one contract, real fees, counting only fills confirmed by a real trade. Results are our own computation. Nothing here is advice. It is the record.