🤖Trading Bot
Execution platforms: Hyperliquid (More platforms to come). Chains: hyperliquid (more chains to add)
The Xtreamly bot is comprised of multiple AI-trained (tree-based) long and short biased models combined as a meta model with a focus on having consistent P&L (both on shorts and long) and improving Sharpe ratio. The model tries to predict the direction of market (long/short) for a specific amount of time ahead. The strategy based on this model is developed following the betting concept (based onthe framework by Dr.De Prado at https://www.amazon.co.uk/Advances-Financial-Machine-Learning-Marcos/dp/1119482089) in which each new position is considered a bet with its own stop loss, take profit, and expiry time. The superposition of all the bets reflects the overall bot performance. In developing the model we used both market data from central exchanges as well as leveraging on-chain data, including lending protocols (e.g., AAVE), DeXs (e.g,. Uniswap), and token flows, among others. The frequency of trades can be controlled by adjusting the minimum confidence level required in all the internal models and also the meta model itself, both of which are tree-based.
ENGINEERING THE YIELD
The system is an engineering tool, not a trading bot.
Yield Focus: The goal is verifiable, positive Net PnL (yield), not speculative gains.
Smartness: A specialized A.I. is trained to maximize risk-adjusted returns (Sharpe). It is programmed to avoid losses, enforcing superior discipline over any human.
Safety: Non-custodial smart contracts protect capital. Leverage is used only for capital efficiency, governed by hard-coded, immutable limits.
Perpetual Action: When no high-conviction trade exists, capital is automatically rotated to a safe, on-chain base yield, ensuring continuous asset work.
Last updated