The Clutch crowd was right on 76.4% of 10,692 questions
Every question resolved on Clutch since May 14, 2026, scored against what the crowd predicted just before the answer was known. Updated hourly from live data.
Data as of November 22, 2026
Based on 2,650,869 predictions by real players.
Cite this data
Free to reuse under CC BY 4.0. Please credit Clutch and link to this page.
Citation
According to Clutch, the crowd’s final call was right on 76.4% of 10,692 settled prediction questions, as of November 22, 2026 (clutch.vip/accuracy). https://clutch.vip/accuracy
Last updated November 22, 2026MethodologyMedia guidelines
Is the crowd calibrated?
Calibration asks a simple question: when the crowd says 70%, does it happen about 70% of the time? Each point groups questions by their final probability. The closer the points sit to the dashed line, the better calibrated the crowd.
Bands with no question are skipped. Questions no real player bet on are left out.
| Probability | Questions | Avg. predicted | Happened |
|---|---|---|---|
| 0–10% | 384 | 6.7% | 2.6% |
| 10–20% | 786 | 15% | 7.4% |
| 20–30% | 954 | 25% | 16.1% |
| 30–40% | 976 | 34.7% | 24.8% |
| 40–50% | 933 | 45% | 33.4% |
| 50–60% | 1,130 | 54.9% | 49.1% |
| 60–70% | 1,244 | 65% | 59.7% |
| 70–80% | 1,283 | 75.1% | 71.8% |
| 80–90% | 1,484 | 85.2% | 84% |
| 90–100% | 1,556 | 94.3% | 94% |
Accuracy by category
How often the crowd picked the right side in each category. Only categories with at least 20 scored questions are shown.
| Category | Questions | Right | Brier |
|---|---|---|---|
| ⚽ Sports | 3,719 | 76.3% | 0.162 |
| 🏛️ Politics | 1,550 | 78.1% | 0.157 |
| 🎬 Entertainment | 1,471 | 80.5% | 0.145 |
| 🔬 Science | 1,156 | 74.6% | 0.174 |
| 🎭 Culture | 994 | 71.1% | 0.194 |
| 💻 Tech | 740 | 77.9% | 0.157 |
| 👤 People | 596 | 77.1% | 0.152 |
| 🌀 Other | 466 | 72.1% | 0.176 |
| 📈 Economy | 38 | 69.4% | 0.248 |
Are bigger crowds more accurate?
The same score, grouped by how many real players bet on each question. The chart shows what the data says, whichever way it points.
| Players | Questions | Right | Brier |
|---|---|---|---|
| 1–4 players | 1,260 | 68.6% | 0.215 |
| 5–19 players | 2,190 | 72% | 0.190 |
| 20–99 players | 2,394 | 78.4% | 0.154 |
| 100+ players | 4,886 | 79.3% | 0.142 |
Methodology
- What counts: every Clutch question that resolved YES or NO. Voided questions are left out.
- What is excluded: questions about Clutch itself (bets placed, shop and contest-ticket counts, leaderboard champions) and mechanical readings (temperatures, train delays). They track a number players can watch tick up rather than a forecast, so they would flatter the score.
- Final probability: the YES price real players set on the question when it closed, the last thing the crowd said before the answer was known. It is rebuilt from real players' bets only, so the automated accounts that keep quiet questions active have no say in it.
- Crowd right: the final probability was above 50% and the answer was YES, or below 50% and the answer was NO. Questions at exactly 50% have no pick and are left out of this figure.
- Brier score: the average of (final probability − outcome)², with YES = 1 and NO = 0. 0 is perfect, and always answering 50% scores 0.25.
- Calibration: questions are grouped into ten bands of final probability (0–10%, 10–20%, …). Each band compares the average prediction with how often YES actually happened.
- Crowd size: the number of real players who bet on a question. The crowd is real players only: bets from automated accounts are left out of every figure on this page. Questions no real player bet on count as resolved but are not scored.
- Data window: questions resolved from May 14, 2026 to November 22, 2026.
- Updated hourly from live data. Every number on this page comes straight from the Clutch database, with no hand-picked examples.
Data licensed CC BY 4.0, cite Clutch (clutch.vip).
How to read this
New to these measures? These guides explain them in plain words:
Think you can beat the crowd?
Clutch is free to play. Make your own calls on the news and see how your record compares.