How Often Are Financial News Sources Right?
We check the tone of every headline against the later price move — and compare each source with chance.
All sources, from the first tradable price
31,244 checks of the 10 sources that have a rate at every horizon: did the market move the way the article pointed, measured from the first price a reader could have traded? Sources with fewer checks are left out here so that no withheld rate can be worked out from these totals.
| Horizon | Checked reports | Days | Hit rate | Chance | Vs. chance (95 % interval) | Detectable from | Verdict |
|---|---|---|---|---|---|---|---|
| 1 d | 10,684 | 254 | 37.4% | 34.0% | +3.5 pts(+1.8 to +5.2) | ±2.4 pts | Better than chance |
| 3 d | 10,451 | 251 | 36.8% | 34.4% | +2.4 pts(−0.2 to +5.1) | ±3.7 pts | Not distinguishable from chance |
| 7 d | 10,109 | 248 | 36.6% | 34.3% | +2.3 pts(+0.4 to +4.3) | ±2.8 pts | Better than chance |
Of 17 sources with at least 300 checks, 1 beat chance after correction for testing many sources.
Chance: how often the same mix of up, down and sideways calls would be right if shuffled against the same price moves.
The interval is the range the edge would most likely fall in if the measurement were repeated (95 %). If it contains zero, the source cannot be told apart from chance.
By source (all horizons)
| # | Source | Hit rate | Chance | Vs. chance (95 % interval) | Verdict | Checked reports |
|---|---|---|---|---|---|---|
| 1 | CNBC Technology | 44.0% | 38.8% | +5.2 pts(−1.2 to +11.5) | Not distinguishable from chance | 507 |
| 2 | U.Today | 44.6% | 41.4% | +3.2 pts(−0.9 to +7.3) | Not distinguishable from chance | 1,570 |
| 3 | FXStreet | 36.4% | 33.2% | +3.2 pts(+1.6 to +4.8) | Better than chance | 10,669 |
| 4 | CNBC Top News | 38.3% | 35.4% | +2.9 pts(−0.3 to +6.1) | Not distinguishable from chance | 3,618 |
| 5 | Investors Business Daily | 35.0% | 33.6% | +1.3 pts(−2.8 to +5.5) | Not distinguishable from chance | 412 |
| 6 | Borsen DK | 38.3% | 37.0% | +1.3 pts(−3.4 to +6.0) | Not distinguishable from chance | 392 |
| 7 | Dagens Industri | 36.0% | 35.1% | +0.9 pts(−2.2 to +4.0) | Not distinguishable from chance | 2,579 |
| 8 | Handelsblatt | 34.3% | 33.9% | +0.4 pts(−2.2 to +3.0) | Not distinguishable from chance | 3,032 |
| 9 | BeInCrypto | 38.2% | 37.9% | +0.3 pts(−2.6 to +3.2) | Not distinguishable from chance | 2,590 |
| 10 | Tagesschau | 32.7% | 33.9% | −1.3 pts(−5.3 to +2.8) | Not distinguishable from chance | 866 |
| 11 | The Block | 36.8% | 38.2% | −1.4 pts(−5.9 to +3.0) | Not distinguishable from chance | 944 |
| 12 | CoinDesk | 36.9% | 38.8% | −1.9 pts(−5.5 to +1.8) | Not distinguishable from chance | 1,367 |
| 13 | wallstreet-online Nachrichten | 34.4% | 36.3% | −2.0 pts(−6.3 to +2.4) | Not distinguishable from chance | 646 |
| 14 | CryptoPotato | 34.4% | 36.4% | −2.1 pts(−6.0 to +1.9) | Not distinguishable from chance | 1,813 |
| 15 | Financial Times | 37.3% | 39.4% | −2.1 pts(−3.7 to −0.5) | Not distinguishable from chance | 3,062 |
| 16 | Decrypt | 35.6% | 37.9% | −2.3 pts(−7.5 to +2.9) | Not distinguishable from chance | 826 |
| 17 | Bitcoin Magazine | 35.8% | 39.8% | −4.0 pts(−11.3 to +3.3) | Not distinguishable from chance | 349 |
13 more sources have fewer than 300 checks; no rate is published for them yet.
By kind of publisher and by language
The same numbers for groups of sources, so a pattern that no single source shows clearly can still be seen.
| Horizon | Checked reports | Days | Hit rate | Chance | Vs. chance (95 % interval) | Detectable from | Verdict |
|---|---|---|---|---|---|---|---|
| Crypto editorial (5 sources), 1 d | 2,780 | 227 | 38.2% | 38.3% | −0.1 pts(−3.3 to +3.2) | ±4.6 pts | Not distinguishable from chance |
| Crypto editorial (5 sources), 3 d | 2,767 | 224 | 38.3% | 38.8% | −0.4 pts(−4.4 to +3.6) | ±5.7 pts | Not distinguishable from chance |
| Crypto editorial (5 sources), 7 d | 2,737 | 222 | 38.1% | 38.6% | −0.6 pts(−4.5 to +3.4) | ±5.6 pts | Not distinguishable from chance |
| Editorial news (4 sources), 1 d | 4,205 | 246 | 37.9% | 35.0% | +2.9 pts(+0.8 to +5.0) | ±3.0 pts | Not distinguishable from chance |
| Editorial news (4 sources), 3 d | 4,071 | 244 | 35.4% | 36.0% | −0.6 pts(−3.1 to +1.9) | ±3.6 pts | Not distinguishable from chance |
| Editorial news (4 sources), 7 d | 4,015 | 238 | 36.5% | 36.0% | +0.5 pts(−2.1 to +3.2) | ±3.8 pts | Not distinguishable from chance |
| Specialist press (1 sources), 1 d | 3,699 | 151 | 36.4% | 33.6% | +2.8 pts(+0.9 to +4.7) | ±2.7 pts | Not distinguishable from chance |
| Specialist press (1 sources), 3 d | 3,613 | 151 | 37.2% | 33.2% | +4.1 pts(+0.4 to +7.7) | ±5.2 pts | Not distinguishable from chance |
| Specialist press (1 sources), 7 d | 3,357 | 149 | 35.6% | 32.8% | +2.8 pts(+0.8 to +4.7) | ±2.8 pts | Not distinguishable from chance |
| Deutsch (1 sources), 1 d | 1,042 | 210 | 36.9% | 33.9% | +2.9 pts(−0.5 to +6.3) | ±4.9 pts | Not distinguishable from chance |
| Deutsch (1 sources), 3 d | 1,003 | 207 | 32.7% | 33.8% | −1.1 pts(−4.1 to +2.0) | ±4.4 pts | Not distinguishable from chance |
| Deutsch (1 sources), 7 d | 987 | 202 | 33.2% | 33.9% | −0.7 pts(−5.1 to +3.8) | ±6.3 pts | Not distinguishable from chance |
| English (9 sources), 1 d | 9,642 | 252 | 37.5% | 34.0% | +3.5 pts(+1.8 to +5.3) | ±2.5 pts | Better than chance |
| English (9 sources), 3 d | 9,448 | 249 | 37.2% | 34.4% | +2.8 pts(+0.1 to +5.6) | ±3.9 pts | Not distinguishable from chance |
| English (9 sources), 7 d | 9,122 | 246 | 37.0% | 34.3% | +2.7 pts(+0.8 to +4.6) | ±2.8 pts | Not distinguishable from chance |
Our Methodology
AI classifies each article's tone as up, down or sideways for the market it covers. From the first price a reader could trade after publication we check the move after 1, 3 and 7 days. A source's hit rate is compared with chance for the same mix of calls and outcomes. Each difference comes with a 95 % interval; checks close in time are resampled together in blocks of 7 or 14 days, because they share market moves. A verdict needs at least 300 checks and must hold after correcting for the number of sources tested (Holm). This answers "did the market move that way?": a hit rate, not a return. AIamond makes no forecasts of its own.
Method source-record-v1, registered on November 3, 2026 before the numbers were published; this run was computed Nov 3, 2026, 05:44 PM UTC. Read the method
SHA-256 868f91c926ab2f151262048a9e9fb97d929a6b8e9882e3dfab20ffafbb234832
Are you a named source and think a number is wrong, or want to reply? Write to [email protected]; corrections are listed in the change log.
Changes to how these numbers are measured, and mistakes we found in the data, are listed in the change log.