Nobody can say how accurate most rug checkers are, because almost none of them publish it: of the eight Solana tools we compared, only one, Before It Rugs, publishes how the coins it graded actually turned out. That doesn't mean the other checkers are wrong, but it does mean you can't measure them, so it pays to know what a fair accuracy record looks like and what ours shows, including where our grades got it wrong.
We run Before It Rugs, so we have an interest here. Everything we say about other tools comes from their own websites and documentation, checked on 26 October 2026. Every number about our own grades is from our live scorecard at 15:07 UTC that day.
What each checker says about its accuracy
| Tool | What its site says about accuracy |
|---|---|
| RugCheck | Its API docs explain how the score is built and where the Good, Warning and Danger bands sit. We found no accuracy figures. |
| Solsniffer | Its FAQ answers "How accurate is the analysis?" by saying its indicators are tested and inviting feedback. No figures. |
| Token Sniffer | Its API docs say it flags 50 to 75% of newly deployed tokens as scams. That is how often it flags, not how often it is right. |
| GoPlus | Its web check says there is no 100% guarantee and the result is for reference only. No figures. |
| De.Fi Scanner | We found no accuracy figures. |
| Bubblemaps | Its wiki says the score only counts clusters it detects. No accuracy figures. |
| DexScreener | Its audit panel warns that audits may not be fully accurate and that it doesn't validate the auditors' data. |
| Before It Rugs | An open scorecard: how often coins with each first grade collapsed within a day, with sample sizes and the raw numbers at /api/scorecard. |
To be fair to the others, honesty about limits is common. GoPlus and DexScreener both say plainly that their checks can be wrong. What's missing is a measured record you can hold them to.
Why accuracy is hard to show
Measuring a rug checker sounds simple: compare its warnings with what happened. In practice there are three traps.
Most new coins fail anyway. When most new tokens collapse within a day whatever anyone says about them, a checker that flags nearly everything will look right nearly all the time. Token Sniffer flagging half or more of new tokens may well be reasonable, but a high flag rate on its own tells you nothing about accuracy. The useful question is whether coins it flagged failed more often than coins it passed.
"Rugged" needs a definition. A coin that drifts to nothing, one whose creator pulls the pool, and one that halves after a pump are very different events. A checker that doesn't say which one it counts can't be held to anything. We count a coin as collapsed if it fell 90% or more from the highest price we recorded, or its pool was emptied, within about a day of launch.
Rules change. Every checker tweaks its scoring. If old and new grades are mixed together, the record describes neither. We keep grades given under earlier rules separate from current ones.
How to judge a rug checker
You can ask these questions of any tool, ours included:
- Does it publish outcomes at all? A record of what happened to coins after they were graded is the starting point.
- Does it define the outcome and the time window? "Collapsed within 24 hours" can be checked. "Scam" can't.
- Does it show sample sizes? A grade that 34 coins received can swing wildly. Look for the counts, and for small groups being held back.
- Are the grades in order? Coins with worse grades should fail more often than coins with better ones. If they don't, the letters aren't telling you much.
- Does it publish the misses? Clean coins that collapsed and flagged coins that kept trading should both be visible.
- Does it keep rule versions apart? Results under old rules should be labelled as such.
- Can you check the raw numbers? A downloadable table beats a headline percentage.
What our scorecard shows
Our honest answer, at 15:07 UTC on 26 October 2026, is that our earlier grades were out of order, and our current grades don't have results yet.
Earlier rules. We graded 6,754 new coins between 21 October 2026 and the rule change on 26 October. For coins at least a day old, this is how often each first grade collapsed within a day:
| First grade (earlier rules) | Coins a day old | Collapsed within a day |
|---|---|---|
| A | 34 | 71% |
| B | 1,859 | 8% |
| C | 705 | 80% |
| D | 877 | 31% |
| E | 821 | 56% |
| F | 1,144 | 36% |
| E or F together | 1,965 | 45% |
| No warning signs | 1,829 | 8% |
The flags did mean something: coins graded E or F collapsed far more often than coins with no warning signs (45% against 8%). But the letters were out of order. C coins collapsed more often than F coins, and the 34 A coins did worse than B.
Why. The earlier rules ignored a coin's stage. A coin still on a launchpad curve, such as pump.fun's, and a coin already trading on an open pool behave very differently. On the same scorecard, launchpad coins graded B collapsed 3% of the time (1,742 coins), while open-pool coins graded C collapsed 98% of the time (518 coins). A letter mostly told you which stage a coin was in.
Current rules. On 26 October 2026 we changed the rules, and grades given from then on are stamped with that rules version. Each sign is now weighted by stage, and a coin trading on an open pool in its first six hours gets points for that alone. The full reasoning, and the test we ran before switching, are on the methodology page. At 15:07 UTC on 26 October, 79 coins had been graded under the new rules and none was yet a day old. The scorecard shows a result for each grade once 30 coins with it are a day old, so the first results should appear within a day or two.
The figures above will move a little as the last coins graded under the earlier rules reach a day old. The scorecard always has the latest.
What accuracy can't tell you
Even a good record describes groups of coins, not the coin in front of you. A grade with a low collapse rate still includes coins that collapsed, and past rates are not a prediction. Our scorecard is a record of what happened, not advice. An A means we found none of the warning signs we look for, never that a coin is safe. See what our grades mean and can a safe-looking coin still rug?
Is RugCheck accurate?
RugCheck doesn't publish an accuracy figure, so nobody outside it can say. Its report is detailed and its method is applied the same way to every token, and we use its data ourselves. Treat its score as a summary of the risks it found, and read the risks.
Which rug checker is the most accurate?
There is no fair answer, because only one of the eight Solana tools we compared publishes a measured record, and that is ours. Pick tools whose checks you understand, use two that look at different things, and favour any that show their results.
Why were Before It Rugs' grades out of order?
Under the earlier rules, a letter mostly said whether a coin was still on a launchpad or already on an open pool, and open-pool coins collapse far more often. We published the result, changed the rules on 26 October 2026 to weight each sign by stage, and keep the earlier results separate.
What counts as a collapse on your scorecard?
A coin that fell 90% or more from the highest price we recorded, or whose pool was emptied, read once 24 to 30 hours after launch. Coins with no reading because our server was down are left out, and the page says how many.
Can I check your numbers myself?
Yes. The scorecard shows every grade with its sample size, and the raw figures are at /api/scorecard. The rules behind each grade are on the methodology page. If you spot a mistake, our corrections policy explains what happens next.
Related
- Solana rug checkers compared
- RugCheck alternatives: when to get a second opinion
- What our A to F grades mean
- How we grade coins