Kalshi odds today › Accuracy study

How accurate is Kalshi?

Sample collected September 25, 2026 · re-collected weekly · 1,766 settled markets · method and free data below

Across 1,766 settled Kalshi markets from 336 events that closed between September 28, 2025 and September 24, 2026, the gap between the price 7 days before close and the share of markets that resolved Yes was within normal sampling noise in 6 of 10 price bands (outside it, 0 to 10%: 3.8% resolved Yes against an average price of 2.6%; 60 to 70%: 51.3% against 64.3%; 70 to 80%: 60.4% against 74.0%; 80 to 90%: 69.1% against 84.6%). Markets priced at 50% or more a week out resolved Yes 68% of the time (379 favorites). Of 1,100 long shots priced at 20% or less, 67 resolved Yes (6.1%, against an average price of 4.8%). The Brier score is 0.109 overall and 0.223 on the 503 markets priced between 20% and 80%; always saying 50% would score 0.250, and lower is better.

Favorites

68% resolved Yes

379 markets priced at 50% or more a week out

Long shots

67 of 1,100 resolved Yes

markets priced at 20% or less a week out

Brier score

0.109 overall

0.223 on markets priced 20% to 80%; always saying 50% scores 0.250

Calibration: the price a week before close against what happened

0%0%25%25%50%50%75%75%100%100%0 to 10%: 893 markets, average price 2.6%, resolved Yes 3.8%10 to 20%: 196 markets, average price 14.1%, resolved Yes 16.3%20 to 30%: 111 markets, average price 24.1%, resolved Yes 20.7%30 to 40%: 107 markets, average price 34.3%, resolved Yes 30.8%40 to 50%: 80 markets, average price 43.9%, resolved Yes 40.0%50 to 60%: 69 markets, average price 54.8%, resolved Yes 46.4%60 to 70%: 76 markets, average price 64.3%, resolved Yes 51.3%70 to 80%: 53 markets, average price 74.0%, resolved Yes 60.4%80 to 90%: 68 markets, average price 84.6%, resolved Yes 69.1%90 to 100%: 113 markets, average price 95.8%, resolved Yes 93.8%Price 7 days before closeShare resolved Yes

Each dot is a 10-point price band, placed at its average price and the share of its markets that resolved Yes; dots on the dashed diagonal are perfectly calibrated. Dot size grows with the number of markets; amber dots are bands where the gap is larger than normal sampling noise.

Price 7 days before closeMarketsAverage priceResolved YesDifferenceSampling noise
0 to 10%8932.6%3.8%+1.2 ptsoutside
10 to 20%19614.1%16.3%+2.2 ptswithin
20 to 30%11124.1%20.7%-3.4 ptswithin
30 to 40%10734.3%30.8%-3.4 ptswithin
40 to 50%8043.9%40.0%-3.8 ptswithin
50 to 60%6954.8%46.4%-8.4 ptswithin
60 to 70%7664.3%51.3%-13.0 ptsoutside
70 to 80%5374.0%60.4%-13.6 ptsoutside
80 to 90%6884.6%69.1%-15.5 ptsoutside
90 to 100%11395.8%93.8%-2.0 ptswithin

Sampling noise: the gap between a band's average price and its Yes share is "within" when it is at most two standard errors for a band of that size and price.

A month, a week and a day before close

On the 1,094 markets with a price 30 days, 7 days and 1 day before close, the Brier score fell steadily as the close approached: 0.120, 0.088 and 0.063 (lower is better). Favorites resolved Yes 63%, 72% and 79% of the time at those three points.

Price takenMarketsBrier scoreBrier, priced 20% to 80% (markets)Favorites resolved Yes
30 days before close1,0940.1200.229 (358)63% of 231
7 days before close1,0940.0880.211 (290)72% of 247
1 day before close1,0940.0630.206 (240)79% of 259

Accuracy by Kalshi category

Among Kalshi categories with at least 20 markets priced between 20% and 80% a week out, Sports had the lowest Brier score on those markets (0.207 on 239) and Entertainment the highest (0.297 on 60). With a few dozen markets per category, differences of a few hundredths are within noise.

CategoryMarketsBrier scoreBrier, priced 20% to 80% (markets)Favorites resolved YesLong shots that resolved Yes
Sports8120.1080.207 (239)68%43 of 561
Economics2120.1080.210 (64)77%2 of 83
Entertainment1890.1790.297 (60)31%8 of 108
Elections1420.0300.150 (12)91%2 of 112
Politics1220.1190.240 (45)68%4 of 62
Crypto1050.1050.234 (27)88%5 of 76
Mentions650.2050.228 (40)54%2 of 19
Science and Technology630.0180.069 (3)94%0 of 46

Categories with at least 40 markets in the sample, as Kalshi labels the event. Categories full of long shots score better overall, so the 20% to 80% column is the fairer comparison.

Method and caveats

Free data

Every market in the sample, with its prices 30, 7 and 1 day before close and its result: JSON · CSV (CC BY 4.0: credit “Kalshi View” with a link to this page). Current prices on the same markets' successors: Kalshi odds today and the biggest movers. Score your own forecasts the same way: Brier score and calibration calculator. Costs and returns of a position: Kalshi tools (fee, payout and APY calculators). A narrower check of eight 2024 election contracts: Kalshi 2024 election accuracy.

Cite this data

APA

How accurate is Kalshi? Calibration study. (2026, September 25). Kalshi View. Retrieved September 25, 2026 at 07:31 UTC, from https://kalshiview.com/markets/accuracy/

BibTeX

@misc{kalshiview2026marketsaccuracy, title = {How accurate is Kalshi? Calibration study}, year = {2026}, month = {Sep}, note = {Snapshot 2026-09-25 07:31 UTC}, url = {https://kalshiview.com/markets/accuracy/}}

HTML credit

<a href="https://kalshiview.com/markets/accuracy/">Odds data: Kalshi View</a>

CC BY 4.0: keep a visible link to this page. More formats: how to cite and reuse Kalshi View.

Frequently asked questions

How accurate is Kalshi?

Across 1,766 settled Kalshi markets from 336 events that closed between September 28, 2025 and September 24, 2026, the gap between the price 7 days before close and the share of markets that resolved Yes was within normal sampling noise in 6 of 10 price bands (outside it, 0 to 10%: 3.8% resolved Yes against an average price of 2.6%; 60 to 70%: 51.3% against 64.3%; 70 to 80%: 60.4% against 74.0%; 80 to 90%: 69.1% against 84.6%). Markets priced at 50% or more a week out resolved Yes 68% of the time (379 favorites). Of 1,100 long shots priced at 20% or less, 67 resolved Yes (6.1%, against an average price of 4.8%). The Brier score is 0.109 overall and 0.223 on the 503 markets priced between 20% and 80%; always saying 50% would score 0.250, and lower is better.

Are Kalshi prices more accurate closer to the close?

On the 1,094 markets with a price 30 days, 7 days and 1 day before close, the Brier score fell steadily as the close approached: 0.120, 0.088 and 0.063 (lower is better). Favorites resolved Yes 63%, 72% and 79% of the time at those three points.

Which Kalshi categories are the most accurate?

Among Kalshi categories with at least 20 markets priced between 20% and 80% a week out, Sports had the lowest Brier score on those markets (0.207 on 239) and Entertainment the highest (0.297 on 60). With a few dozen markets per category, differences of a few hundredths are within noise.

What does calibration mean for a prediction market?

A calibrated market's 70% prices come true about 70% of the time. The table groups settled markets by their price a week before close and shows how often each group resolved Yes.

How was this study done?

We scanned Kalshi's most traded series for markets that settled Yes or No in the 365 days before the collection date, traded at least 1,000 contracts and were open for at least 8 days, kept up to 15 of the most traded markets per event and the 2,000 most traded of those, and read each market's closing trade price for the last day with a trade in the week up to 7 days before its close from Kalshi's public daily price history. Single games, multi-leg parlays and hourly or daily series are left out.

Can I use this data?

Yes. Every market in the sample is in the free JSON and CSV files under CC BY 4.0: credit Kalshi View with a link to this page.