BTC Implied Volatility Prediction Competition

This competition challenges participants to predict the 30-day implied volatility (DVOL) of Bitcoin using the Deribit volatility index.

About

The BTC DVOL Competition tests your ability to forecast Bitcoin's 30-day implied volatility – a forward-looking measure derived from options prices that reflects market expectations of future price movement. Unlike price prediction, this competition focuses on volatility forecasting, a critical component of options pricing and risk management.

This is NOT a price prediction competition. The target is Deribit's DVOL index, which represents the market's consensus view of Bitcoin's volatility over the next 30 days.

The competition uses real-time data from Deribit's public API to evaluate predictions against actual implied volatility measurements.

Partnership

This competition is developed in partnership with Autonity, a decentralized protocol for forecasting and prediction markets. Autonity hosts the Forecastathon, a series of forecasting and trading competitions that bring together data scientists and traders to predict real-world events and market movements.

Quick Start

Install the package and run the starter notebook:

pip install btcvol
jupyter notebook ../notebooks/Getting_Started.ipynb

Reward Scheme

The BTC DVOL Competition is distributing $5,000 USDC to top performers.

# Top 5 Reward Distribution
rewards = {
    1: 2000,  # 1st place: $2,000 USDC
    2: 1000,  # 2nd place: $1,000 USDC  
    3: 1000,  # 3rd place: $1,000 USDC
    4: 500,   # 4th place: $500 USDC
    5: 500,   # 5th place: $500 USDC
}

Cumulative Reward Distribution

Top 1 get 40% of total rewards
Top 1-2 get 60% of total rewards
Top 1-3 get 80% of total rewards
Top 1-5 get 100% of total rewards

Dataset

Data Sources

Historical Deribit BTC DVOL data
- notebooks/deribit_btc_dvol_15min_ohlc.csv
- notebooks/deribit_btc_dvol_daily_ohlc.csv
Deribit Volatility Index API
- Target DVOL (30-day implied volatility)
- Endpoint: /public/get_volatility_index_data
CrunchDAO Price API (optional)
- Historical and real-time Bitcoin price data
- Alternative sources allowed (e.g. Pyth Network)

Data Structure

asset: "BTC"
horizon: 3600s (1h) or 86400s (24h)
step: 900s (15 minutes)
predicted_value: implied volatility in 0–1 range

Target

The target is the Deribit DVOL, representing:

Expected 30-day BTC volatility
Derived from at-the-money options
Updated in real time
Expressed as a decimal (e.g. 0.40 = 40%)

Tournament Structure

Timeline

Training: February 16 – March 13, 2026
Scoring: March 16 – April 18, 2026
Competition End: April 18, 2026

Submission Cut-off

Submit your model anytime during the training period. Once submitted, it runs continuously and generates predictions every 15 minutes.

Scoring and Evaluation

Participants are evaluated using Mean Squared Error (MSE).

error = predicted_dvol - actual_dvol
mse = error ** 2

Scoring Windows

30D Final Score (official)
7D Live Score
24H Short-term Score
1H Recent Score

Prediction Frequency

Predictions every 15 minutes for:

1 hour horizon
24 hour horizon

Key Implementation Notes

Return a list of floats
Volatility in 0–1 range
Exactly horizon // step values
No exceptions allowed

Penalty Score Policy: Failed predictions receive a score equal to 10× the worst model. If all fail, penalty is set to 1.0.

Useful Links

Resources

Package

btcvol Python package
Install: pip install btcvol

Community

Competition Host: CrunchDAO
Prize Pool: $5,000 USDC
Duration: 30 days
Target: Deribit 30-day BTC Implied Volatility (DVOL)