BigWay: the result is verified, not fabricated
AI & Machine Learning
GuardLabs internal quantum derivatives laboratory: 50 days on Binance Futures, 1,809 closed trades, accumulated total +20,293 conditional dollars with a maximum drawdown of 5.4% and a profitable trades ratio of 68.7%.
Eight configurations of the same signal stream at the same costs resulted in a total variance of 42 thousand conditional dollars — therefore, each hypothesis undergoes a rigorous verification conveyor: clustered block-bootstrap CI95, walk-forward on a window that the model has not seen, survival control of instruments (delisted coins remain in the sample) and a significance threshold of 200 trades.
Separately, the laboratory hosts Cloud Diver — with an open verifier for trades based on raw data.
Honestly: this is not a backtest — the laboratory operates live, capturing the Binance market in real-time and maintaining an account continuously for 50 days. Each trade was opened at the moment the signal appeared, and its result was calculated later, based on data that did not yet exist at the input. The execution is virtual: orders were not sent to the exchange, real money was not traded on these lines, commission, funding, and slippage were modeled on actual quotes. The position is fixed at $1,000 without reinvestment — the total is the sum of individual trades, not the growth of the deposit. We do not sell signals and do not promise profitability.
Live case with complete methodology: guardlabs.online/studio/cases/bigway-quant-lab/ #python #quant #trading #binance #dataanalysis
Eight configurations of the same signal stream at the same costs resulted in a total variance of 42 thousand conditional dollars — therefore, each hypothesis undergoes a rigorous verification conveyor: clustered block-bootstrap CI95, walk-forward on a window that the model has not seen, survival control of instruments (delisted coins remain in the sample) and a significance threshold of 200 trades.
Separately, the laboratory hosts Cloud Diver — with an open verifier for trades based on raw data.
Honestly: this is not a backtest — the laboratory operates live, capturing the Binance market in real-time and maintaining an account continuously for 50 days. Each trade was opened at the moment the signal appeared, and its result was calculated later, based on data that did not yet exist at the input. The execution is virtual: orders were not sent to the exchange, real money was not traded on these lines, commission, funding, and slippage were modeled on actual quotes. The position is fixed at $1,000 without reinvestment — the total is the sum of individual trades, not the growth of the deposit. We do not sell signals and do not promise profitability.
Live case with complete methodology: guardlabs.online/studio/cases/bigway-quant-lab/ #python #quant #trading #binance #dataanalysis