Election Market Review Finds Kalshi Favorites Won 96.4% of Scored Races

Prediction market traders got most election winners right, but a closer look at the underlying contracts shows why that conclusion needs more detail than a single accuracy figure.

A DeFi Rate analysis of an election-market spreadsheet found that Kalshi’s pre-election favorites won 489 of 507 scored events, a 96.4% hit rate. Polymarket US favorites won 408 of 417 scored events, or 97.8%. The comparison uses prices observed at or before midnight Eastern entering Tuesday and reported election outcomes through Thursday evening.

Those figures support the broad conclusion that traders correctly anticipated most winners. They also identify notable blind spots:

• Half of Kalshi’s 18 misses came in Texas.

• Both exchanges favored Darline Graham over eventual South Carolina Senate winner Annie Andrews.

The results differ from Kalshi’s own post-election recap released Wednesday, which reported 479 winning favorites across 487 resolved races, a 98% hit rate. The figures are not directly interchangeable: DeFi Rate’s analysis uses an explicit pre-election cutoff, a broader event inventory, and eligibility checks that exclude incomplete price histories. Kalshi’s recap uses an earlier results snapshot and does not clearly identify a single price cutoff for its headline calculation.

What the spreadsheet actually measures

DeFi Rate’s workbook contains 1,256 exchange-specific events: 716 on Kalshi and 540 on Polymarket US. It also includes 2,688 individual outcome-contract records.

The distinction matters. A race with several candidate contracts should contribute one observation to a winner-selection score, not one for every losing candidate correctly assigned a low price.

For each eligible event, the review selects the highest-priced listed outcome using observations available at or before Tuesday at 12 a.m. Eastern. It scores that favorite only when every listed outcome has an available price and the reported winner matches. Results are based on media calls, not official certification.

Calculated from the workbook’s event-level records. The exchanges do not have identical coverage, so this is not a like-for-like platform ranking.

The Polymarket figures concern its U.S. exchange, not its international platform. A race listed on both exchanges contributes one observation to each, so the combined total is not a count of unique elections.

Kalshi anticipated congressional control and the major Senate flips

The spreadsheet records Democratic control of the House as Kalshi’s highest-priced outcome at 93.5%, and Democratic Senate control at 76%. Both matched the reported results.

Its individual Senate markets also performed strongly. Of 33 Senate events eligible for scoring, 32 favorites won. Successful selections included several competitive Republican-held seats:

The findings also show that traders didn't favor Democrats everywhere. Kalshi correctly favored Republican Ashley Moody in Florida at 85% and Pete Ricketts in Nebraska at 67%. 1343

South Carolina was the scored Senate exception. Graham was the favorite at 67% on Kalshi and 68% on Polymarket US. Andrews won.

Texas accounted for half of Kalshi’s misses

Kalshi’s 18 losing favorites included nine Texas events, spanning statewide offices, a judicial seat and House districts.

That concentration supports the argument that traders saw the Democratic direction without fully anticipating its reach.

They favored Talarico for Senate while still backing Republicans in several other Texas contests:

The other Texas misses were agriculture commissioner, land commissioner and two House races. All are recorded as media-called outcomes, not certified results.

The Brett Busby and Don Huffines contracts stand out because their prices expressed considerably more confidence than a coin flip. Elsewhere, some misses were much narrower judgments: Kalshi favored Rhett Marques at 52% in an Alabama House race ultimately won by Shomari Figures.

Both count as a losing favorite in this score. They do not represent the same degree of forecasting error.

The exchanges shared several blind spots

All nine of Polymarket US’s losing-favorite events also appear among Kalshi’s misses. They include South Carolina Senate, Texas governor and seven House races.

The prices were often close across venues. Both favored Abbott at 60%. Kalshi priced Victoria Spartz at 79%, while Polymarket US had her at 78%; J.D. Ford was the reported winner. Both also favored Republican candidates who lost House races in Missouri, Nevada, Ohio, Alabama and Texas.

That overlap suggests the mistakes were not confined to one platform’s pricing. It does not establish why traders reached those judgments or prove the exchanges were responding to identical information.

Nor does Polymarket US’s higher overall hit rate establish that it forecast better on the same slate. Kalshi’s scored sample includes additional state and local offices, and the workbook does not provide a matched-sample platform ranking.

Why these figures differ from Kalshi’s 98%

Kalshi’s recap counted six categories: House, Senate, governor, attorney general, lieutenant governor and secretary of state. It reported 479 winning favorites out of 487 races resolved as of 6 a.m. Wednesday.

DeFi Rate’s analysis covers a wider range of offices and uses reported outcomes through Thursday. It also excludes events without enough historical pricing to reliably identify a favorite. Those methodological differences mean 96.4% is a separate measurement, not, on its own, a correction proving the company’s 98% claim false.

The company also highlights markets reaching 99% on House control at 8:10 p.m. and Senate control at 9 p.m. Election Night. That measures how markets incorporated incoming results, a different test from identifying winners before voting ended.

Hundreds of events remain outside the score

The workbook does not treat every listed event as a completed forecast.

For Kalshi, 29 events lack complete outcome prices, 14 have no forecast price, and 166 are not called or not independently matched to a result. For Polymarket US, those categories contain 52, six and 65 events, respectively.

Excluding incomplete price coverage is particularly important. If only one candidate’s historical price is available, you cannot safely label that candidate the favorite without knowing the alternatives.

The underlying observations have limitations, too. Kalshi prices are last-trade observations from the final positive-volume candle available before the cutoff; some are older than others. They are not synchronized executable quotes.

A strong hit rate is not the same as calibrated probabilities

The review establishes winner-selection performance. It does not establish that every displayed probability was accurate.

A 53% favorite and a 97% favorite both receive one correct mark when they win. A 52% favorite and an 83% favorite both receive one incorrect mark when they lose. That discards information about confidence.

Evaluating calibration requires asking whether outcomes assigned similar probabilities occurred at roughly those rates across a defined sample. DeFi Rate’s headline score does not answer that question.

The defensible conclusion is still substantial: pre-election favorites won more than 96% of eligible events on both exchanges. Traders correctly anticipated congressional control and most major Senate outcomes. But Texas, South Carolina and shared House misses show why that record should be presented as a transparent, provisional score, not proof that markets had the entire election figured out