Pricing Model Overview

This page explains how prediction.click computes fair values. You don't need to understand the math to use the app — but if you're curious, here's how it works.

The Big Picture

We're answering one question: "What is the true probability that this crypto asset finishes above its starting price?"

For example, if BTC was $73,500 at the start of a 5-minute window and is now $73,600 with 3 minutes left — what's the probability it stays above $73,500 when the window closes?

The Formula

Fair Value UP = 1 - Student_t_CDF(z, v=7)

where z = ln(K / S) / (σ x sqrt(T))

In plain English:

Symbol
Meaning
Where it comes from

S

Current spot price

Binance real-time trades (~100ms delay)

K

Starting price (strike)

Price at the beginning of the market window

σ

Volatility

Calculated from recent 1-minute candles

T

Time remaining

Countdown to market expiry

v=7

Fat-tail parameter

Fixed — accounts for crypto's extreme moves

Why Student-t (Not Normal)?

Crypto prices make extreme moves more often than a standard bell curve (Normal distribution) predicts. The Student-t distribution with v=7 has fat tails, meaning it assigns higher probability to big price swings.

Distribution
Chance of a 3-sigma move

Normal (bell curve)

0.27%

Student-t (v=7)

~1.0%

This matters because if you underestimate extreme moves, your fair values will be wrong at exactly the moments when they matter most.

How Volatility is Measured

We use EWMA (Exponentially Weighted Moving Average) volatility from 1-minute candle data.

  • Recent candles get more weight than older ones (decay factor λ = 0.94)

  • Half-life is about 11 minutes — the model adapts quickly to changing conditions

  • Volatility is annualized for the formula, then scaled by the time remaining

What this means for you:

  • After a big price move, volatility spikes → fair values move closer to 50% (uncertain)

  • During calm periods, volatility drops → fair values become more extreme (confident)

Intuitive Effects

What happens
Effect on Fair Value UP

Price goes up (S > K)

Fair Value increases (more likely to finish up)

Price goes down (S < K)

Fair Value decreases

Time running out (T → 0)

Fair Value becomes extreme (0% or 100%)

High volatility (σ up)

Fair Value moves toward 50% (anything can happen)

Low volatility (σ down)

Fair Value becomes more extreme (current trend likely continues)

Update Frequency

The pricing engine recomputes fair values every 1 second for all 24 combinations (6 assets x 4 timeframes). The spot price from Binance updates even faster (~100ms).

Limitations

  • The model assumes continuous price movement — it doesn't predict sudden news events

  • Volatility is backward-looking — a sudden regime change takes a few minutes to fully reflect

  • The model doesn't account for Polymarket-specific factors like orderbook depth or whale activity

  • No model is perfect — a 53% fair value doesn't guarantee a win

Want More Detail?

Check out the Backtest Results to see how this model performed on thousands of historical markets.

Last updated