Revettr Trade Agent

Revettr Alpha is an autonomous AI trading agent that executes mean-reversion strategies on crypto perpetual futures, guided by a proprietary Markov regime detection layer. The core insight: most trading systems fail not because their signals are wrong, but because they trade in the wrong market conditions. Mean reversion works exceptionally well in range-bound, high-volatility environments — but gets systematically crushed during strong trending regimes. Revettr Alpha solves this by classifying the market into one of nine discrete states across a volatility × trend Markov chain, computing rolling transition probabilities, and only entering trades when the probability of transitioning to a mean-reversion-favorable regime is rising. The signal engine scores every tradeable asset across four independent dimensions: trend structure, mean reversion quality, volume confirmation, and regime context. Only the single highest-conviction setup above a calibrated confluence threshold trades each day. Position sizing is ATR-calibrated at approximately 20% of Kelly criterion — conservative enough to survive worst-case drawdowns while compounding returns meaningfully. Stop losses are set at 2x ATR from entry, with take-profit targets at 1.25x the stop distance. Risk per trade is fixed at 5% of equity with a hard 10x leverage cap. Backtested on 5.5 years of Hyperliquid perpetual futures data (October 2020 to March 2026), spanning the 2021 bull run, 2022 bear market, and 2023-2026 recovery. The system includes a nested machine learning layer reserved for future activation once live trading data provides a richer training distribution. All features are lagged by one bar to eliminate look-ahead bias, and the backtester uses conservative same-bar resolution. Built for the Kraken CLI hackathon track — executing autonomously via Kraken's paper trading sandbox with live PnL tracking on Surge