Technical profile and market coverage
Bender
uses Bybit
on 15m timeframe.
Sample capital: $1000.
Recommended minimum: $200.
Technical notes
Gen 4 is a full redesign: the trading policy is trained offline with reinforcement learning and only EXECUTED live — no learning happens at trade time. Model inputs combine multi-timeframe trend/momentum/volatility structure with validated quant-research signals (per-asset realized-performance memory, event risk, order-book depth, funding). Leverage selection is part of the model's own action, and execution runs inside a structural risk box: hard brackets, a daily trade cap, and an automatic halt after a losing day. Every model version must clear several independent statistical tests (walk-forward validation on unseen data and strict overfitting checks) and beat the incumbent before deployment; versions are controlled and traceable.
Covered pairs (4)
SOL/USDTAVAX/USDTBTC/USDTBNB/USDT
Full bot details
Bender is designed to be predictable: at most 5 trades a day, each with a predefined margin-based take-profit/stop-loss plus a profit-locking mechanism, and leverage the model itself picks between 3x and 10x based on its confidence — scaled down by an external risk layer in risky conditions. After a bad day the bot stops taking new entries for 24h. The model retrains nightly, but a new version replaces the running one only if it is profitable on out-of-sample tests AND beats the incumbent; performance is also reviewed weekly under a fixed engineering framework.
Reinforcement-learning futures bot; pairs are picked dynamically by the quant-research platform. Gen 4: training is fully separated from live execution, and a new model version deploys only after passing several independent statistical tests on unseen data; every trade carries a predefined take-profit and stop-loss from entry.
Supported leverage range: 2x to 8x.
The model warm-up uses 1000 initial candles.
Features
Generation-3 reinforcement-learning engine with continual training on fresh market data.Release gate: a new version goes live only by outperforming the current one on unseen data.Structural risk box: margin-based take-profit and stop-loss, identical in training and live.Steady cadence of 1-5 trades per day with two-sided (long/short) coverage.Adaptive 2-8x leverage matched to market conditions.Limited scale-in (DCA) entries, only with renewed model confirmation.