The LightGBM model, trained on AMZN daily data from Jan 2020 with SMA/EMA indicators, outputs a 54% probability that the May 26 close will be lower than the $266.32 close on May 25. A 54% directional probability, while low, suggests a mild edge for a short-term bearish move; the model’s backtest cumulative returns were superior to buy‑and‑hold. This is a short‑dated, low‑conviction short on AMZN based on a technical ML signal; the author is tracking a $1,000 paper account to verify live performance. Low confidence (54%) means high chance of the prediction being wrong; model may overfit to past data; single‑day directional bets are noise‑dominated; market may react to news outside the technical features.
The LightGBM model, trained on AMZN daily data from Jan 2020 with SMA/EMA indicators, outputs a 54% probability that the May 26 close will be lower than the $266.32 close on May 25. A 54% directional probability, while low, suggests a mild edge for a short-term bearish move; the model’s backtest cumulative returns were superior to buy‑and‑hold. This is a short‑dated, low‑conviction short on AMZN based on a technical ML signal; the author is tracking a $1,000 paper account to verify live performance. Low confidence (54%) means high chance of the prediction being wrong; model may overfit to past data; single‑day directional bets are noise‑dominated; market may react to news outside the technical features.