Machine learning

Crypto Options Prediction Bot — Inside Our Next-Gen AI Trading Engine

| 19 seen

At Terramatris, we’ve spent years exploring the intersection of quantitative finance, machine learning, and blockchain markets.
Our latest internal project - Crypto Options Prediction Bot - represents a major leap in how AI can analyze and rank crypto options across Deribit in real time.

Unlike retail “signal” bots, our bot doesn’t guess. it learns, measures, and scores every BTC and ETH options contract based on statistical probabilities, expected returns, and volatility dynamics.

  • Fetches live Deribit options data for BTC and ETH every week.
  • Filters all contracts with Friday expiries — matching standard options cycles.
  • Uses machine learning models to estimate:
    • The probability an option expires out-of-the-money (P(OTM))
    • Its expected return (%)
    • Liquidity and volatility conditions
    • Market regime indicators like implied-volatility rank

These inputs are combined into a composite scoring system that ranks the most statistically favorable opportunities — for both option sellers and buyers.

How It Works

The bot connects to the Deribit API, fetching full BTC and ETH options chains up to 365 days ahead. Raw data is processed and normalized - calculating spreads, deltas, implied volatility, and distance from the money. A custom scoring algorithm blends risk, reward, and liquidity to highlight the most promising contracts. Weekly results are stored locally building a historical dataset for future model retraining.

Our algo trading bot currently runs on our internal machines, air-gapped from external systems. It is not connected to exchanges or wallets, and does not execute trades -it simply produces data-driven insights for internal analysis.

Each Friday at 08:00 UTC, the bot automatically:

  • Fetches new Deribit data
  • Scores all contracts
  • Saves the top results to an internal archive

The system helps our research team identify weekly patterns and optimize strategies — especially for covered calls and short puts.

Over the next few months, we will continue gathering and analyzing weekly data. Once sufficient history is built, we’ll begin internal testing for on-chain execution and closed-loop learning.

Public rollout isn’t planned before late 2026, as our focus remains accuracy, stability, and compliance.

We’re open to partnerships, research collaborations, and institutional pilot discussions. If you’re building in the crypto options or quantitative analytics space, let’s talk.

Subscribe to Machine learning