Wenn eine Plattform 70 % sagt, wie oft tritt das dann ein? Zuverlässigkeitskurven und erwarteter Kalibrierungsfehler je Plattform, berechnet aus aufgelösten Märkten. Transparenzbericht der Plattformen →
Bewertet die Wahrscheinlichkeit, die jeder Markt rund 24 Stunden vor seiner Auflösung angesetzt hat, über Events, deren vollständiges Ergebnisbuch erfasst wurde. Ein niedrigerer ECE ist besser.
| Vorhergesagt | Ø Vorhersage | Tatsächliche Rate | Paare | |
|---|---|---|---|---|
| 0-10% | 2.8% | 3.0% | 1,355 | |
| 10-20% | 14.7% | 13.0% | 203 | |
| 20-30% | 25.3% | 24.2% | 520 | |
| 30-40% | 35.4% | 30.6% | 222 | |
| 40-50% | 45.8% | 41.9% | 394 | |
| 50-60% | 53.6% | 55.0% | 340 | |
| 60-70% | 63.8% | 69.4% | 140 | |
| 70-80% | 74.4% | 70.8% | 69 | |
| 80-90% | 84.6% | 81.0% | 46 | |
| 90-100% | 96.1% | 96.1% | 108 |
| Vorhergesagt | Ø Vorhersage | Tatsächliche Rate | Paare | |
|---|---|---|---|---|
| 0-10% | 3.1% | 0.7% | 243 | |
| 10-20% | 14.8% | 16.4% | 83 | |
| 20-30% | 24.3% | 13.4% | 76 | |
| 30-40% | 34.6% | 39.0% | 31 | |
| 40-50% | 46.0% | 50.0% | 40 | |
| 50-60% | 52.7% | 51.4% | 53 | |
| 60-70% | 64.9% | 67.2% | 40 | |
| 70-80% | 73.9% | 93.6% | 32 | |
| 80-90% | 84.4% | 79.6% | 44 | |
| 90-100% | 97.1% | 99.7% | 101 |
| Vorhergesagt | Ø Vorhersage | Tatsächliche Rate | Paare | |
|---|---|---|---|---|
| 0-10% | 0.9% | 6.9% | 22,108 | |
| 10-20% | 15.0% | 9.7% | 1,158 | |
| 20-30% | 25.0% | 22.5% | 927 | |
| 30-40% | 35.2% | 33.6% | 919 | |
| 40-50% | 45.2% | 43.3% | 1,535 | |
| 50-60% | 54.1% | 52.3% | 1,493 | |
| 60-70% | 64.1% | 59.9% | 691 | |
| 70-80% | 74.7% | 61.4% | 655 | |
| 80-90% | 84.6% | 66.9% | 647 | |
| 90-100% | 94.7% | 58.1% | 1,060 |
| Vorhergesagt | Ø Vorhersage | Tatsächliche Rate | Paare | |
|---|---|---|---|---|
| 20-30% | 26.5% | 66.7% | 6 | |
| 30-40% | 36.8% | 0.0% | 5 | |
| 40-50% | 46.4% | 53.3% | 75 | |
| 50-60% | 53.1% | 46.4% | 84 | |
| 60-70% | 62.7% | 100.0% | 6 | |
| 70-80% | 73.5% | 33.3% | 6 |
| Vorhergesagt | Ø Vorhersage | Tatsächliche Rate | Paare | |
|---|---|---|---|---|
| 0-10% | 5.0% | 0.0% | 8 | |
| 10-20% | 14.4% | 0.0% | 6 | |
| 20-30% | 25.5% | 0.0% | 6 | |
| 30-40% | 35.6% | 58.4% | 10 | |
| 40-50% | 43.2% | 54.5% | 19 | |
| 50-60% | 56.7% | 42.7% | 21 | |
| 60-70% | 64.6% | 42.4% | 12 | |
| 70-80% | 74.7% | 84.1% | 5 | |
| 80-90% | 83.7% | 48.9% | 3 | |
| 90-100% | 94.3% | 75.7% | 13 |
Noch nicht bewertbar: ForecastEx (no events in the provider-confirmed cohort with recorded t24h input) · Futuur (only 2 scorable events (need 30)) · Metaculus (no events in the provider-confirmed cohort with recorded t24h input) · Myriad (no events in the provider-confirmed cohort with recorded t24h input) · PredictIt (no events in the provider-confirmed cohort with recorded t24h input) · Robinhood (Rothera) (no events in the provider-confirmed cohort with recorded t24h input) · Smarkets (no events in the provider-confirmed cohort with recorded t24h input)
Calibration compares the probability the market assigned to each outcome ~24h before resolution against the realised result, over resolved markets with provider-confirmed outcomes and ONE captured price snapshot in the inclusive 20-28h window before resolution. Select the retained timeline capture nearest 24h; equal-distance ties use the earlier capture. Validate that selected book without falling back to another capture if it is invalid or partial. The separate winner-only t24hProbability is a cohort gate, not the scored book. Downsampling can remove an otherwise available window capture, so historical coverage is limited to retained observations. That is a single pre-resolution observation, NOT a claim of continuous 24h history: a market with no snapshot that far back (most same-day markets) is excluded entirely rather than scored on a nearer capture. Reliability bins group (outcome, probability) pairs; a well-calibrated venue's realised win rate tracks its predicted probability. calibrationError is the event-weighted mean gap (ECE, lower is better): each resolved event contributes one unit regardless of outcome count, preventing large outcome catalogs from dominating venue comparisons. Plausibly complete books are normalized to remove overround; independent ladders remain raw. An event is scored only when the t-24h snapshot captured the market's COMPLETE outcome book: a partial snapshot would make inclusion depend on whether the eventual winner happened to be captured, which biases the result. Markets with no capture in that window and partial-book snapshots are excluded. Invalid books, including missing or nonnumeric prices, are excluded; a genuine numeric zero remains valid. The per-venue counts are published in `excluded` on scored and pending rows; the exclusion rate differs sharply by venue, so compare sampleSize alongside calibrationError.
Bewertet den von der Plattform gemeldeten letzten Kurs vor der Auflösung über aufgelöste Märkte mit zwei Ergebnissen. Der Vorlauf variiert, daher ist diese Serie nicht mit der 24-Stunden-Serie oben vergleichbar.
Basis: der von der Plattform gemeldete letzte Kurs vor der Auflösung.
| Jahr | Bewertete Events | Kalibrierungsfehler (ECE) | |
|---|---|---|---|
| 2017 | 10,732 | 8.72pp | |
| 2018 | 22,445 | 9.19pp | |
| 2019 | 21,480 | 2.05pp | |
| 2020 | 8,384 | 0.52pp | |
| 2021 | 14,113 | 1.03pp | |
| 2022 | 14,396 | 2.42pp | |
| 2023 | 14,160 | 1.19pp | |
| 2024 | 12,857 | 1.03pp | |
| 2025 | 14,460 | 1.44pp | |
| 2026 | 10,236 | 1.66pp |
| Vorhergesagt | Ø Vorhersage | Tatsächliche Rate | Paare | |
|---|---|---|---|---|
| 0-10% | 5.7% | 4.4% | 8,952 | |
| 10-20% | 14.7% | 11.7% | 15,717 | |
| 20-30% | 25.0% | 21.9% | 21,851 | |
| 30-40% | 35.1% | 30.7% | 37,478 | |
| 40-50% | 44.4% | 41.5% | 49,513 | |
| 50-60% | 53.5% | 55.6% | 64,066 | |
| 60-70% | 63.9% | 68.1% | 39,599 | |
| 70-80% | 74.0% | 77.4% | 22,962 | |
| 80-90% | 84.3% | 87.2% | 15,907 | |
| 90-100% | 93.7% | 95.1% | 10,487 |
The final-price series scores the probability the venue itself reported as the market's last pre-resolution price (persisted at ingest) against the realised result, over resolved two-outcome markets whose winner is one of the outcomes and whose both outcome prices are present. Lead time before resolution is unknown and not uniform, so these figures are NOT comparable with the t-24h series above and are reported as a separate basis. Plausibly complete books (price sum 95-125) are normalized to remove overround; others are read raw. Currently published only for venues whose persisted resolved prices were verified to be genuine trading prices rather than settled 0/100 values (Futuur, verified 2026-08-18 across a monotone 2016-2026 reliability curve). Per-year rows below the minimum sample are withheld.
Bewertet die letzte Wahrscheinlichkeit, die CoinRithm vor der Auflösung eines Marktes erfasst hat, gegen das tatsächlich eingetretene Ergebnis. Bei einem medianen Vorlauf von Minuten ist dies eine plattformübergreifende Integritätsprüfung der Kurse kurz vor der Auflösung, kein Maß für Prognosefähigkeit.
Basis: unsere eigene letzte Erfassung vor der Auflösung. Der Vorlauf wird je Plattform ausgewiesen; nicht vergleichbar mit der 24-Stunden-Serie.
| Jahr | Bewertete Events | Kalibrierungsfehler (ECE) | |
|---|---|---|---|
| 2026 | 14,467 | 0.48pp |
| Vorhergesagt | Ø Vorhersage | Tatsächliche Rate | Paare | |
|---|---|---|---|---|
| 0-10% | 1.0% | 0.7% | 12,092 | |
| 10-20% | 14.8% | 13.3% | 1,218 | |
| 20-30% | 25.0% | 25.7% | 1,040 | |
| 30-40% | 35.1% | 35.7% | 1,098 | |
| 40-50% | 45.6% | 46.3% | 1,441 | |
| 50-60% | 54.1% | 53.3% | 1,552 | |
| 60-70% | 64.7% | 64.0% | 1,082 | |
| 70-80% | 74.8% | 73.7% | 1,034 | |
| 80-90% | 84.9% | 86.6% | 1,093 | |
| 90-100% | 99.0% | 99.3% | 9,781 |
Basis: unsere eigene letzte Erfassung vor der Auflösung. Der Vorlauf wird je Plattform ausgewiesen; nicht vergleichbar mit der 24-Stunden-Serie.
| Jahr | Bewertete Events | Kalibrierungsfehler (ECE) | |
|---|---|---|---|
| 2026 | 1,272 | 1.62pp |
| Vorhergesagt | Ø Vorhersage | Tatsächliche Rate | Paare | |
|---|---|---|---|---|
| 0-10% | 1.8% | 0.9% | 1,752 | |
| 10-20% | 14.1% | 7.3% | 174 | |
| 20-30% | 24.6% | 25.5% | 93 | |
| 30-40% | 34.4% | 33.0% | 76 | |
| 40-50% | 46.1% | 44.6% | 95 | |
| 50-60% | 53.4% | 54.1% | 99 | |
| 60-70% | 65.1% | 68.7% | 55 | |
| 70-80% | 74.8% | 75.7% | 68 | |
| 80-90% | 85.5% | 92.9% | 110 | |
| 90-100% | 98.0% | 99.2% | 912 |
Basis: unsere eigene letzte Erfassung vor der Auflösung. Der Vorlauf wird je Plattform ausgewiesen; nicht vergleichbar mit der 24-Stunden-Serie.
| Jahr | Bewertete Events | Kalibrierungsfehler (ECE) | |
|---|---|---|---|
| 2026 | 70,374 | 1.89pp |
| Vorhergesagt | Ø Vorhersage | Tatsächliche Rate | Paare | |
|---|---|---|---|---|
| 0-10% | 2.3% | 1.1% | 146,782 | |
| 10-20% | 14.4% | 9.9% | 11,410 | |
| 20-30% | 24.6% | 19.8% | 8,863 | |
| 30-40% | 34.7% | 31.5% | 8,966 | |
| 40-50% | 44.6% | 43.8% | 9,368 | |
| 50-60% | 54.5% | 54.8% | 9,274 | |
| 60-70% | 64.6% | 67.3% | 8,767 | |
| 70-80% | 74.7% | 78.6% | 8,121 | |
| 80-90% | 84.9% | 88.8% | 9,559 | |
| 90-100% | 97.4% | 96.8% | 38,769 |
Basis: unsere eigene letzte Erfassung vor der Auflösung. Der Vorlauf wird je Plattform ausgewiesen; nicht vergleichbar mit der 24-Stunden-Serie.
| Jahr | Bewertete Events | Kalibrierungsfehler (ECE) | |
|---|---|---|---|
| 2026 | 534 | 5.67pp |
| Vorhergesagt | Ø Vorhersage | Tatsächliche Rate | Paare | |
|---|---|---|---|---|
| 0-10% | 3.9% | 0.5% | 259 | |
| 10-20% | 14.3% | 9.9% | 141 | |
| 20-30% | 24.7% | 13.9% | 109 | |
| 30-40% | 35.4% | 25.3% | 100 | |
| 40-50% | 44.8% | 32.4% | 93 | |
| 50-60% | 54.9% | 54.0% | 107 | |
| 60-70% | 64.5% | 73.4% | 90 | |
| 70-80% | 75.4% | 81.9% | 92 | |
| 80-90% | 85.5% | 81.2% | 127 | |
| 90-100% | 96.0% | 92.4% | 267 |
Basis: unsere eigene letzte Erfassung vor der Auflösung. Der Vorlauf wird je Plattform ausgewiesen; nicht vergleichbar mit der 24-Stunden-Serie.
| Jahr | Bewertete Events | Kalibrierungsfehler (ECE) | |
|---|---|---|---|
| 2026 | 777 | 14.60pp |
| Vorhergesagt | Ø Vorhersage | Tatsächliche Rate | Paare | |
|---|---|---|---|---|
| 0-10% | 3.0% | 0.9% | 338 | |
| 10-20% | 15.1% | 10.3% | 78 | |
| 20-30% | 25.1% | 19.1% | 47 | |
| 30-40% | 34.1% | 35.3% | 34 | |
| 40-50% | 49.0% | 8.9% | 248 | |
| 50-60% | 50.8% | 82.7% | 312 | |
| 60-70% | 65.8% | 66.7% | 33 | |
| 70-80% | 74.4% | 77.8% | 45 | |
| 80-90% | 84.6% | 89.9% | 79 | |
| 90-100% | 96.9% | 99.1% | 340 |
The own-capture series scores the LAST probability CoinRithm itself captured strictly before the provider-reported resolution time (from the durable frozen timeline) against the realised result. Lead time varies and is typically SHORT — per-venue median and p90 lead seconds are published — so with a median lead of minutes this is a near-final-price integrity check over our own observations, NOT a forecast-skill measure, and is not comparable with the t-24h series. An event is scored only when the captured point holds the market's complete outcome book (the same complete-book rule as the t-24h series: a partial capture makes inclusion depend on whether the winner was captured). Plausibly complete books (price sum 95-125) are normalized to remove overround; others are read raw. Events are unit-weighted regardless of outcome count. Coverage grows with forward capture and spans venues the t-24h series undersamples. Per-year rows below the minimum sample are withheld.