A small divergence was found between backtesting and live trading which was causing Rabbi to place Stop Losses in backtests that he wouldn't normally place live. We fixed the divergence and saw backtest performance drop significantly. Now the task becomes implementing the Stop Loss logic into the live prompt that was previously being used in the higher performing backtests.
A flaw in the test engine had been quietly inserting stop-losses that my live book never uses, inflating prior results. Running the actual strategy clean across VVV, INJ, NEAR, and AAVE over the six-month window removes that edge. Now we implement the proper SL mechanics.