Paste a prediction. When its deadline passes we check it against exchange prices, stamp the verdict, and add it to the author's public record.
Tested on 170 real posts from X (25 October 2026): all 81 calls it scored on its own were read correctly. How verdicts work
MarkMyCallReceipt no. call_3_19
$SOL down to $84.8 within 35 days. This is happening.
Ranked by a hit rate that accounts for luck: an account needs 10 resolved calls to rank, and 3 for 3 does not beat 70 for 100. “Says vs hits” is how much surer they sounded than they turned out to be.
| # | Account | Resolved | Hit rate | Says vs hits |
|---|---|---|---|---|
| 1 | @example_quietquant | 34 | 74% | −4 pts |
| 2 | @example_macromaven | 22 | 55% | +20 pts |
| 3 | @example_chartwizard | 27 | 48% | +39 pts |
| 4 | @example_moonboi | 41 | 29% | +63 pts |
| – | @example_newcaller provisional | 6 | 100% | −22 pts |
A record is half the story. The other half is how sure they sounded. A stated percentage wins; otherwise we read it from their own words. Every account then shows the gap: sounding 95% sure and landing 30% of calls is +65 points of says vs hits.
mark my words,
screenshot this,
guaranteed95%
will,
definitely,
for sure85%
I think,
likely,
should70%
could,
might,
maybe60%
A model pulls the claim, the target and the deadline from the post. A second, rule-based pass re-reads the author's own words for the deadline and checks the price at the time. If the two disagree, a person decides.
Touch calls (“hits $10k by Q1”) are checked hourly and can win early. Everything else waits for the deadline.
Coinbase minute, hour and day candles decide; Bitstamp must agree. Disagreements and near misses within 1% go to a person, not the scoreboard.