Research Paper: Hold-to-Expiry Edge in CFTC-Regulated Crypto Binary Options

Published: March 2026 | OddsReference Research

This paper presents a comprehensive analysis of pricing efficiency in CFTC-regulated crypto binary options on Kalshi. Using three distinct pricing models tested against 877,606 settled contracts, we evaluate whether theoretical edge translates to tradeable profits and document three critical pricing bugs discovered during the research.

Download

Download Full Paper (PDF)

Abstract

We evaluate the pricing efficiency of CFTC-regulated crypto binary options on the Kalshi exchange using 877,606 settled BTC and ETH contracts spanning January 2024 through March 2026. Three pricing models are tested: TWAP-adjusted Black-Scholes (the baseline), Kou double-exponential jump-diffusion (capturing crypto kurtosis), and isotonic recalibration (a nonparametric correction layer). The TWAP-adjusted model produces a hold-to-expiry edge of +1.2 cents per signal, statistically significant but insufficient to overcome execution costs at taker fee levels. Kou jump-diffusion reduces mean absolute error by 8.2% but adds no tradeable improvement. Isotonic recalibration appeared to produce a 12.3c edge but was discovered to contain a data leakage bug that inflated apparent performance by 340%. Across 72.1 million trades, market makers earned +1.12% while takers lost -1.12%, confirming that the structural advantage belongs to liquidity providers. We conclude that Kalshi crypto binary options are approximately efficient, with residual edge below the minimum fee floor for profitable exploitation.

Key Findings

Paper Structure

1. Introduction — Research questions and market context
2. Market Structure — 5-minute contract mechanics, fee structure, $60M daily volume
3. Data — 877,606 contracts, 72.1M trades, OHLCV price data
4. Model 1: TWAP-Adjusted Black-Scholes — Volatility estimation, bracket probability computation
5. Model 2: Kou Jump-Diffusion — Asymmetric jump process, kurtosis calibration
6. Bug Discovery — Three critical implementation errors and their impact
7. Model 3: Isotonic Recalibration — Data leakage diagnosis, the calibration trap
8. Hold-to-Expiry Analysis — Edge decomposition, fee sensitivity, execution constraints
9. Market Microstructure — Maker-taker dynamics, order flow, favorite-longshot bias
10. Discussion — Efficiency interpretation, cross-domain comparison, implications

Related Research

This paper is part of a two-paper series on prediction market efficiency:

• Companion paper: Weather market efficiency (NWS data, 40,032 contracts)
• This paper: Crypto market efficiency (877,606 contracts, three pricing models)

Source Articles

The underlying analysis is documented in four detailed articles:

• We Backtested Three Pricing Models on 877,000 Crypto Contracts — Full backtest methodology and results
• Inside Kalshi’s Crypto Binary Options — Market structure and participant analysis
• We Built a Weather Forecasting Model to Trade Kalshi — Weather model for cross-domain comparison
• 800,000 Trades Exposed: How Kalshi Weather Markets Converge — Convergence mechanics comparison

For real-time prediction market data across all platforms, visit our live dashboard.

Frequently Asked Questions

What is this crypto binary options research paper about?

This paper tests three pricing models against 877,606 settled Kalshi crypto binary option contracts. We find that a +1.2 cent per-signal edge exists using TWAP-adjusted Black-Scholes, but it cannot survive execution costs at taker fee levels. The structural advantage belongs to market makers, who earned +1.12% across 72.1 million trades while takers lost -1.12%.

What pricing bugs were discovered during the research?

Three critical bugs: (1) B-value interpretation — the contract center value was misread as the floor, shifting probability mass by 5-15 cents on 40% of contracts. (2) Width encoding — ETH brackets were coded as $20 when the actual width was $40, doubling the implied volatility for ETH. (3) API field changes — Kalshi migrated from integer cents to string dollar fields mid-dataset, causing silent price misreads.

How does crypto market efficiency compare to weather markets?

Both markets show approximately efficient pricing, but the structural details differ. Crypto markets have a larger favorite-longshot bias (2-4 pp vs 0.4 pp in weather), higher emotional stakes, and deeper liquidity ($60M daily volume vs $2K-$5K per weather event). The common finding: public information is incorporated within minutes, and no model using only public data can consistently beat fees.

Key Takeaways

• A +1.2 cent per-signal edge exists in crypto binary options but requires maker-fee execution and cannot survive taker costs
• Three pricing models tested: TWAP-adjusted Black-Scholes (best), Kou jump-diffusion (marginal improvement), isotonic recalibration (data leakage trap)
• Market makers earn +1.12% while takers lose -1.12% across 72.1 million trades — the structural advantage is liquidity provision, not signal quality
• Three implementation bugs discovered that affected 40%+ of contracts, highlighting the practical challenges of pricing model deployment
• Cross-domain comparison with weather markets confirms the same efficiency conclusion: public prediction markets aggregate information faster than individual models