Pricing Gaps in Midterm Election Prediction Markets: One Question, Four Answers, and a 9.7-Point Spread

Author: Coin Metrics Research Team

Translated by: TechFlow

TechFlow Editor's Note:As the 2026 U.S. midterm elections approach, cumulative trading volume in prediction markets has crossed the $1.4 billion mark, becoming the sharpest barometer of win probabilities for global capital outside traditional polls. Yet most observers focus only on the single, spotlighted "win probability," overlooking the complex micro-market structure beneath the surface.

In this research report, crypto quant and on-chain data giant Coin Metrics uses Senate control events on the regulated prediction platform Kalshi as a case study, tracking 38 independent contracts linked to the same election outcome. Remarkably, the objectively binary political question "Can Democrats take control of the Senate?" is priced into four mutually divergent implied probabilities across different contract families on the platform, with the maximum intraday spread once surging to 9.7 percentage points.

By constructing a "Nine-Leg" package that buys control and shorts the granular seat-count ladder, one can theoretically lock in a constant payoff under any election outcome and capture arbitrage opportunities. However, cumbersome market-making friction, extremely polarized order-book liquidity, leg execution risk, and fee deductions of up to 5.2 cents vividly reveal the brutal gap prediction markets must cross from "theoretical arbitrage" to "real returns".

Key Takeaways

For the same core question, "Will Democrats control the Senate?", Kalshi prices the outcome through four different contract families.

Since August, their implied win probabilities have differed by an average of 3.3 percentage points, and the maximum spread once widened to 9.7 percentage points. As of October 25, probability valuations across the different contracts still ranged from 62.5% to 65.3%.

The theoretical "Nine-Leg" arbitrage package—buying the Democratic Senate control contract while selling all standalone seat-count contracts for Democrats winning 51 seats or more—would pay a fixed $8 principal-and-interest payout under any final election outcome if the two contract families settle on the same facts.

Based on end-of-day closing quotes, the package's total buy cost was below $8 on 19 of 51 trading days, creating a theoretical arbitrage opportunity (before fees on October 24, the peak arbitrage margin per $8 face-value package reached 8.1 cents).

Actual execution assumptions directly drive final profitability. After incorporating estimated trading fees, if the main leg is assumed to fill passively via resting orders, simulated excess returns remained positive on 7 trading days (up to +4.7 cents); but if all nine legs aggressively crossed the bid-ask spread as takers, the number of positive-return days dropped to just 2.

Liquidity distribution within the platform is severely imbalanced. Since August, the position-concentrated flagship Democratic control contract has recorded $3.3 million in trading volume, while some granular seat-count contracts needed to price the package have accumulated only $2,000 in total volume.

The Senate case is only a microcosm of the broader prediction-market ecosystem. Kalshi has more than 880,000 active contracts captured by Talos CM prediction-market data, and this highly fragmented market architecture routinely creates similar pricing gaps across different events and even different platforms. Capturing this alpha depends on systematic event-mapping identification and sophisticated execution algorithms.

Introduction

With only one week remaining until the midterm elections, cumulative trading volume on Kalshi and Polymarket for contracts linked to the 2026 midterm elections has surpassed the $1.4 billion mark.

Most media coverage and market commentary typically focus on a single headline number: Kalshi currently shows a probability of about 63% that Democrats will win the Senate.

However, this widely cited single number is only the tip of a much larger iceberg. Around the Senate control issue, Kalshi lists as many as 38 independent contracts under five different event categories, each with its own order book and trading community.

In this article, we will compare in depth the differences in the true probabilities implied by these derivative contracts, estimate when these spreads converge into a simulated arbitrage opportunity with positive expected value, and examine the microstructure that keeps these pricing gaps from being closed over time.

The Same Question, Four Very Different Answers

On Kalshi, investors can express bets on Senate control through four different routes:

The first is a binary contract directly tied to the control outcome (whether Democrats control the Senate); the second is a ladder contract on the exact number of seats each party wins in the Senate (covering 51 seats through 57 seats and above); the third is a balance-of-power combination contract; and the fourth is a reverse contract on Republican control.

Figure: Four contract families and event structures for Senate control on Kalshi

If every derivative contract in the prediction market were perfectly price efficient, the four curves reflecting the implied probabilities of the four contract families should overlap exactly.

Figure: Comparison of implied win probabilities for different Senate-control event contracts on Kalshi

Broadly, the four contract types do follow the same macro trend: the probability of a Democratic win rose amid volatility from about 46% in early August to the low range of 60% to 65% in October. However, they rarely reached absolute agreement at specific points.

On as many as two-thirds of trading days, the implied win probability derived from the Democratic seat-count ladder was significantly higher than that of the headline control contract. On September 20, valuations among the four answers diverged sharply between 55.7% (implied by Republican seat counts) and 65.4% (implied by Democratic seat counts).

Figure: Trajectory of the maximum and minimum implied-probability spreads across different contract families

Statistics show that the average spread between the highest and lowest implied probabilities is 3.3 percentage points. This gap was widest in September (averaging 4.5 percentage points, compared with 3.2 percentage points in August); in October, as volume continued to expand, the average spread remained 2.5 percentage points.

This indicates that pricing gaps do not automatically close as liquidity is added: on October 21, the maximum valuation gap among the four contract types was still as high as 4.0 percentage points.

Assembling a Nine-Leg Risk-Free Arbitrage Package

When two contracts that should be equivalent disagree on the same election fact, in theory, buying the cheaper side of each at the same time can convert the spread into a certain risk-free profit. Here, the most intuitive arbitrage structure is precisely assembled from nine legs (Nine Legs):

Leg 1: Buy the YES contract for Democrats winning Senate control. If Democrats win, the contract pays $1 at expiration.

The remaining eight legs: Buy NO contracts for each Democratic seat-count tier from 51 seats upward (51 seats, 52 seats, 53 seats, through 57 seats and above). If the final election result does not land exactly on that tier, the corresponding contract pays $1.

Figure: Simulated payoff matrix for buying Democratic control and selling each seat-count contract

As long as the two contract families settle at expiration based on the same factual logic, the arbitrage package can reliably recover $8 in expiration payout under any election outcome.

In other words, as long as the total cost of buying across these nine dimensions is below $8, the entire package locks in a positive arbitrage opportunity in the mathematical model.

Figure: Historical simulated returns of the nine-leg arbitrage package before and after trading fees

Backtest analysis based on daily closing prices since August (all estimates are theoretical simulations):

In an ideal state excluding trading fees, if each leg is bought by taking the best ask (Ask), the package's total cost was below $8 on 19 of the 51 trading days.

However, after deducting estimated fees (about 5.2 cents per arbitrage package on average), only 2 trading days maintained positive returns under aggressive ask-taking.

Using a real order-book depth snapshot from October 25 as an example, opening 10 arbitrage packages costs $79.65 in total (including estimated fees), while the fixed expiration payout is $80, leaving a net profit of only $0.35.

Most positive-return opportunities were concentrated in the final two weeks, mainly driven by pricing divergences catalyzed by the approaching election and a sharp rise in market trading activity.

Why the Spread Persists: Liquidity Disparity and Trading Frictions

If the strategy's final cash flow is locked, why did rational arbitrage capital not quickly eliminate the spread entirely? The core answer lies in liquidity.

Chart: Severely Imbalanced Volume Distribution Across Kalshi Senate Contract Types

Since August, the cumulative notional trading volume across all Kalshi Senate contracts has been approximately $14.3 million, but more than 90% of that capital was absorbed by the single flagship Senate control contract.

The ultimate success of the arbitrage portfolio depends entirely on its least liquid leg. The Democratic control contract had $3.3 million in cumulative executed volume, while the hedge leg, 'Democrats win 54 seats,' had only $3,000 in cumulative volume, and 'Democrats win 56 seats' had just $2,000.

Beyond extremely thin liquidity, live trading also faces many real-world friction costs that must be addressed:

The first is legging risk. Traders can fully lock in theoretical profits only after all nine orders are successfully filled; if slippage occurs or an order on a thin leg fails to fill, the remaining orders are highly exposed to sharp one-sided price moves.

The second is high trading fees. Kalshi's fee schedule is parabolic, peaking near the 50-cent price and declining toward both 0 cents and 100 cents. An average fee outlay of about 5.2 cents per package can instantly consume most of the thin spread alpha.

The third is settlement basis and timing mismatch. The strategy assumes that 'Senate control' and 'exact seat count' settle on the same facts. However, control contracts settle once the election outcome is clear, while seat-count contracts settle based on the formal congressional composition on February 1, 2027. If resignations, party switches, or recounts in disputed districts occur in the interim, the settlement bases may diverge, causing the final payout to deviate from $8.

The fourth is capital lockup and opportunity cost. Most of the initial capital will be forced to remain locked in the seat-count contracts until final settlement on February 1, 2027; the multi-month horizon creates non-negligible financing costs.

Profound Implications for Prediction Market Participants

The structural cracks exposed by the Senate episode are not isolated; they are a common structural feature of the rapid expansion of prediction markets. Monthly trading volume across prediction markets has grown explosively from under $5 billion in September 2025 to about $24 billion in April 2026, and this scale surge has been accompanied by unprecedented market fragmentation:

A single topic spawns a massive number of contracts. According to Talos CM prediction market reference data, Kalshi alone currently has more than 880,000 active event contracts. From election politics, Federal Reserve decisions, and macroeconomic data to sports events, each macro theme is sliced into win/loss, range, threshold, and various permutation and combination contracts.

Multiple trading venues coexist in a fragmented manner. Platforms such as Kalshi and Polymarket often list competing contracts on exactly the same objective events, but subtle differences exist in contract text definitions, trading fees, and settlement rules.

Liquidity depth is extremely uneven. As the Senate data show, headline mainstream contracts absorb the vast majority of market trading capital, while liquidity in related derivative contracts is often only a tiny fraction of that. Such thin order books with shallow depth are fertile ground for significant price divergence, but also killers of arbitrage profits as execution slippage rapidly erodes them.

For quantitative market makers and hedge funds, systematically capturing such pricing cracks depends on building capabilities in two dimensions:

The first is precise event identification. Across different platforms and contract matrices, quickly find contracts tied to the same objective outcome, convert them to a unified basis, and calculate truly executable prices after fees and depth. This requires building standardized cross-platform event maps, thoroughly reviewing settlement terms to prevent factual conflicts, and accessing high-frequency order books and fee models in real time.

Second is precise trade-capture capability. A theoretical opportunity on the screen is never the same as real money in the account. In practice, this requires the algorithm to patiently place passive limit orders on the most liquid primary leg and wait for fills, then quickly sweep eight shallow order books to complete the hedge before price drift occurs, while precisely controlling slippage and taker-fee drag.

This rigorous analytical framework also applies to any major prediction event involving multiple contracts or cross-platform listings.

Conclusion

In public perception, prediction markets are often simplistically viewed as black boxes that output a single probability for each question. However, Kalshi's Senate contracts reveal that prediction markets are essentially a set of interconnected yet independent order books.

These order books broadly follow the same macro narrative, but local 2-3 percentage-point frictions in market perception persist. Notably, even amid explosive volume growth, these gaps have not been fully closed. On multiple trading days this month, the accumulation of these small price differences evolved into positive-expected-value mathematical arbitrage opportunities.

The 2026 midterm elections are only one representative sample of this wave. As prediction markets continue to expand and the ecosystem becomes increasingly fragmented, similar opportunity windows will continue to emerge across sectors. To succeed in this space, institutional participants need two core foundations: underlying data assets that provide a comprehensive view of the market to identify opportunities, and a calm, rational framework for live execution costs to place bets prudently.

Methodology:This study is based on daily candlestick data for 38 Senate-related contracts obtained from Kalshi's public API from August 1, 2026 to October 25, 2026, with trading days defined in Eastern Time. All contracts included in the portfolio were formally listed on Kalshi before August 2026. Implied probabilities were calculated from the closing mid-price between the best bid and best ask. Seat-count and power-balance probabilities were normalized using staircase sums to remove excess premiums caused by bid-ask spreads.

The cost of the combination package uses the best ask for the YES leg and 1 minus the best bid for the NO leg. Notional volume equals the number of executed contracts multiplied by price. The contract universe comes from the Talos CM prediction market reference database. All derived results are theoretical simulations and do not include additional slippage beyond order-book quotes; they do not constitute investment advice.

Reference Notes

(1) Data comes from the Talos CM API endpoint /reference-data/markets (type=event_contract, exchange=kalshi), capturing the total number of active Kalshi contracts as of October 25, 2026 (882762 contracts). This figure changes dynamically each day as contracts expire and new listings go live.

(2) CNN reported, "Cumulative trading volume in prediction markets for the 2026 midterm elections has surpassed USD 1.4 billion," published on October 23, 2026.

(3) Pew Research Center research report, "Prediction market trading volume has surged explosively in recent months," published on May 27, 2026.

Source: 深潮 TechFlow