Sharpe ratio = 40.79. Ha, ha... The paper is full of errors because @SciReports does not properly review papers. Retraction asap.
Random variation of existing results. The reviewers at @SciReports have failed. Interesting how the results for other architectures are so similar. @RealAAAI #AlgoTrading
Can a trading model perform better than cumulative weekly true range? It seems to be an upper bound...true efficiency=real PnL, max efficiency=cumulative weekly TR. Is there a better way? You can add all zig-zags, but is it a better measure?
Sometimes it is challenging to decide which model to use for real trading. Total gain is the goal, but this models was in loss, when SMCI was +200% in 2024.
DELL is a counter example for the look-ahead window size. 800% simulated gain in 2024 (for some models).
The less data you use when training AI for algorithmic trading, the better the results. Experiment group 9 only uses 1.5% 30-min OHLCV data, the rest was filtered out. The results are for 2024. REW09 was trained with a larger look-ahead window than REW07.
One more sample trade executed by #AI model. It seems, the AI model could learn divergences. $BTC #algotrading
