Agent & quant supervised
Gadget does not decide alone
Gadget sits under the same quant-research and fleet-agent layer; its key numbers come from its own realized trades rather than a fixed setting.
-
No version ships untested
Every new version must beat the running one on unseen data, or it never reaches your signal.
-
Risk adjusts itself
The bot's record is scored regularly: weaken and its capital share shrinks, recover and it is released.
-
It steps back before risky conditions
Unusual market conditions cut the capital share before entry — not after the loss.
-
The agent reads the record daily
Losses do not go unanswered: the agent isolates the cause and puts the fixes that matter first.
This layer guarantees nothing and does not replace your own risk management; its job is to keep decisions sitting on fresh evidence.
See the whole loop
Technical profile and market coverage
Gadget
uses Bybit
on 15m timeframe.
Sample capital: $1000.
Recommended minimum: $200.
Technical notes
Gadget's reward is aligned with realized profit, so behaviour that does not actually earn money is not reinforced. Leverage and capital share move inversely with volatility: calmer markets mean more risk taken, volatile ones mean less. Exits are composite, and a backstop stop-loss always sits behind the trade.
Covered pairs (3)
BNB/USDTORDI/USDTJTO/USDT
Full bot details
Gadget weighs market quality before every entry and stays out when conditions are ambiguous. Its difference is that no fixed number decides the exit — the engine has learned when to take profit.
A risk-first reinforcement-learning engine on Bybit futures that learns for itself when to exit.
Supported leverage range: 3x to 7x.
The model warm-up uses 320 initial candles.