METHODOLOGY
How the benchmark is constructed
FairOdds compares stored Polymarket CLOB YES prices with an option-implied benchmark price (OIP) for the same BTC or ETH payoff and horizon. This page explains how that OIP is built, maturity-aligned, and checked before it appears on the dashboard.
From market data to a comparison
The two lines on FairOdds originate in different markets. The CLOB line records what Polymarket participants paid for YES. The OIP line estimates the value of the same binary payoff under an options-implied, risk-neutral distribution.
- 01Discover
Find supported BTC and ETH markets and parse their settlement rules.
- 02Observe
Store hourly CLOB prices and current order-book information.
- 03Normalize
Turn volatility-surface inputs into comparable option prices.
- 04Calibrate
Fit a Kou distribution for each asset and standard tenor.
- 05Compare
Price the payoff, align its maturity, and calculate the wedge.
THE DISPLAYED WEDGE
WedgetCLOB YEStaligned OIPt
A positive wedge means the stored Polymarket price is above the OIP at that timestamp; a negative wedge means it is below. Direction and size identify a difference to investigate, not a guaranteed mispricing or trade.
Data inputs and timing
Every published comparison is assembled from timestamped inputs with distinct roles. FairOdds keeps acquisition, calibration, payoff state, and presentation separate so an unavailable source does not silently become a model estimate.
| Source | Input | How FairOdds uses it |
|---|---|---|
| Polymarket | Gamma market metadata, public CLOB books, and price history | Identifies supported contracts, parses payoff rules, records current liquidity, and supplies the stored hourly YES-price series. |
| Block Scholes | Hourly composite BTC/ETH delta-volatility surfaces and spot indices | Provides the standardized option surface and underlying level used to fit each risk-neutral Kou distribution. |
| U.S. Treasury | Public short-rate observations | Supplies the rate used when converting delta/volatility observations into strikes and normalized option prices. |
| Binance | BTC and ETH candle histories | Checks whether Reach or Dip barriers have already been hit and verifies terminal settlement states when required. |
TIMING CONVENTION
Polymarket history is requested at 60-minute fidelity and the production pipeline refreshes hourly. Current order books are retained for live bid, ask, midpoint, spread, and liquidity context; the charted comparison uses the stored CLOB history point and the OIP calculated for that same historical timestamp.
Building the option-implied distribution
FairOdds uses the Kou double-exponential jump-diffusion model. It combines continuous price variation with upward and downward jumps, allowing the options-implied distribution to express asymmetric tails without requiring a large parameter set.
- σ
- Diffusion volatility
- λ
- Jump intensity
- p
- Probability that a jump is upward
- η1
- Positive jump-tail decay
- η2
- Negative jump-tail decay
CALIBRATION FIT
FairOdds fits the model to normalized option prices using bounded, spot-scaled pricing errors, with greater weight on observations nearer the money.
SURFACE NORMALIZATION
Comparable option prices
For every complete hourly delta grid, FairOdds combines implied volatility, spot, maturity, and the Treasury rate to recover strikes and Black forward option prices.
PRODUCTION FIT
Four synchronized tenors
BTC and ETH are fitted separately at 7, 14, 30, and 60 days. Each fit uses three consecutive hourly surface snapshots, and the production calibration runs every hour.
CONTINUITY
Warm-started estimation
The previous successful parameters initialize the next bounded optimization. Only successful, compatible calibrations can enter a published synchronized tenor set.
PRICING CONVENTION
Rates serve different roles
Treasury rates enter surface normalization. The binary benchmark shown as OIP is then evaluated under FairOdds’ zero-rate digital-pricing convention.
Pricing the four contract types
The settlement rule determines the valuation method. FairOdds supports four precisely parsed payoff families and refuses to price contracts whose rules cannot be normalized without ambiguity.
| Type | YES pays when | OIP method | Additional state |
|---|---|---|---|
| Above | The terminal reference price is above the stated level. | Kou terminal probability from the fitted return distribution. | Settlement candle after expiry. |
| Range | The terminal reference price lies inside the stated interval. | Difference between two Kou terminal cumulative probabilities. | Settlement candle after expiry. |
| Reach | The reference price touches or exceeds a barrier during the observation window. | Kou upward first-passage probability. | Binance highs from the observation start. |
| Dip | The reference price touches or falls below a barrier during the observation window. | Kou downward first-passage probability. | Binance lows from the observation start. |
WHY BARRIER CONTRACTS DIFFER
Above and Range depend on the price at one terminal time. Reach and Dip depend on the entire path: a barrier hit remains decisive even if the price later reverses. Their OIPs therefore require both observed candle coverage and an explicit first-passage model, making them more model-dependent.
Maturity alignment and no-lookahead history
A Polymarket contract rarely has exactly 7, 14, 30, or 60 days remaining. FairOdds prices the required tenor components at the contract’s actual remaining maturity, then combines their probabilities using calendar-time weights.
INTERPOLATED OIP
OIP(T) = wlower · OIPlower(T) + wupper · OIPupper(T)
The weights sum to one and are determined by the contract’s calendar distance between the two surrounding standard tenors.
Use the matching 7-, 14-, 30-, or 60-day calibration.
Linearly interpolate the two component probabilities.
Use the 7-day calibration as the short endpoint.
Use the 60-day calibration as the long endpoint.
No-lookahead selection
- 1
Start from the timestamp of the stored Polymarket CLOB observation.
- 2
Select the newest successful same-timestamp calibration set that does not occur after that observation.
- 3
Require compatible numerical settings and a common spot across every available tenor.
- 4
Mark a calibration more than six hours older than the CLOB observation as stale instead of publishing an OIP.
WHEN NO OIP IS SHOWN
FairOdds suppresses the benchmark when a required calibration is missing or stale, barrier history is incomplete, a delayed observation window has not started, or the numerical result fails validation. Known completed payoff states may instead be shown deterministically as 0% or 100%.
Reading the comparison
OIP is a second valuation lens, not a prediction of the physical probability that an event will occur. It reflects option prices, the fitted risk-neutral distribution, the contract mapping, and FairOdds’ stated numerical conventions.
WHAT OIP IS
A comparable benchmark price
A model-derived value on the same 0–1 scale as a $1 binary YES contract, matched to the asset, payoff definition, observation state, and remaining horizon.
WHAT OIP IS NOT
A guaranteed trading value
It is not a quoted executable option price, a real-world forecast, proof of arbitrage, or a recommendation to enter a prediction-market position.
What can create or distort a wedge
- Different participants, constraints, liquidity, and settlement mechanisms across markets.
- Bid-ask spreads, execution costs, price-history sampling, and asynchronous source updates.
- Model specification, surface quality, tenor interpolation, and numerical approximation.
- Path dependence and incomplete candle coverage for Reach and Dip contracts.
- Contract wording or reference-venue details that do not map cleanly to the supported payoff families.
Research and references
FairOdds adapts the paper’s cross-market pricing framework for a continuously refreshed website. The production system retains the Kou model and payoff logic while changing the data-delivery, cadence, and maturity-alignment machinery needed for live operation.
RESEARCH METHOD AND LIVE IMPLEMENTATION
Research paper
Deribit surfaces obtained through Amberdata. Kou parameters are estimated daily from rolling seven-day windows of 7-day constant-maturity surfaces, and previous-day parameters are used for hourly contract pricing.
FairOdds website
Block Scholes-delivered hourly composite surfaces at 7, 14, 30, and 60 days. Separate three-hour rolling fits run hourly, then the newest synchronized no-lookahead set is aligned to each contract’s maturity.
The website therefore operationalizes the research framework; it does not reproduce the paper’s historical estimator unchanged.
| Working paper | Cross-Market Pricing in Prediction Markets: A Comparison with Derivatives M. Fabi, R. Marfè, V. Ruffo, and L. Schönleber View the paper on SSRN (opens in a new tab) |
|---|---|
| Core method | Kou (2002) double-exponential jump-diffusion pricing for terminal digital and first-passage barrier claims. |
| Implementation | The current FairOdds code and its validated data contracts are authoritative for values displayed by the live website. |