Why any call without a timestamp is worth nothing

The question was: how do I judge whether a crypto caller is any good? The uncomfortable answer is that most callers make the judgement impossible on purpose — not by lying about individual tokens, but by leaving out the one detail that turns a claim into evidence: when the call happened, and what the token was worth at that exact second. Without that, every screenshot is just a picture of a chart.

A call is a claim about a moment, not a token

"Token X did a hundred times" is not a statement about Token X. It is a statement about a starting point and an ending point, and if you move either one, the number changes. Move the start earlier and the multiplier grows. Move it later and it shrinks.

The starting point that matters is the market capitalisation — the total value the market is putting on all circulating tokens — at the moment the call was published. Call it the entry. If someone posts about a token after it has already run, the entry is inflated, and the multiple they later claim was never available to anyone reading them.

Three things have to be present before a call can be evaluated at all:

1. A timestamp. The moment of publication, precise enough to identify a price.
2. An entry recorded at that moment. Not a round number, not "around here", but the market cap when the words went out.
3. Peak measured only after that moment. A high that happened before the call is not the call's high. It is someone else's.

Miss any one of the three and the claim is unfalsifiable. Unfalsifiable claims are not evidence, and a caller who never provides them is not being modest. They are keeping their options open.

Survivorship bias, shown with arithmetic

Survivorship bias is judging a set by the members that survived. The classic example is a fund that closes its bad portfolios; what remains looks excellent. In crypto it works the same way, one deleted tweet at a time.

Here is our own logged Solana data, forward-measured from the moment each token was recorded. Across 36,165 logged calls, 7,705 doubled — 21.3%. Almost twice as many, 18,963 of them, went flat or down, which is 52.4%. The remainder sat between flat and a double: up, but not by much.

Now watch what one editing decision does. Drop the 18,963 flat-or-down tokens out of the denominator. The same 7,705 successes now sit on a base of roughly seventeen thousand two hundred instead of thirty-six thousand. The apparent rate almost doubles. Nothing happened in the market. Nothing happened to the tokens. Only the denominator changed.

That is the whole trick. A caller who mentions only the tokens that worked is not reporting a track record; they are reporting a denominator they chose after the fact. And because the losing half is the boring half, nobody screenshots it.

The tail is loud and small

The headline number is always the extreme. On the same Solana set: 6.4% reached five times the entry, 2.6% reached ten times, and 0.3% reached fifty times. Those are the stories that get retold, and they are the smallest buckets in the data.

This is not an argument that large multiples never happen. It is an argument about which number a caller shows you. If someone's track record is a list of the fifty-times tokens, ask what the base was. The base is where the 52.4% lives.

Categories move differently, so compare like with like

One more trap: comparing a caller from one kind of token to a caller from another. On Robinhood Chain, our logged meme tokens doubled 5.6% of the time across 8,150 tokens. Unclassified tokens doubled 5.0% of the time across 3,955. Stock-paired tokens doubled 16.0% of the time across 868. Utility tokens doubled 27.5% of the time across 639.

Two things follow. First, a caller working in one category cannot be judged against the average for a different category — the underlying odds differ before anyone opens their mouth. Second, look at the size of the bucket. Six hundred and thirty-nine tokens is a small population, and in a small population a percentage