In this study, a stock-outperformance prediction framework was developed using a comprehensive dataset of 150 stocks from the NASDAQ exchange equities covering the period from 1962 to 2020. This diverse set of technical, price momentum-oriented, dispersion-based and market-relative features was engineered to capture trend-following behavior and systemic risk behaviors in financial time series. Four machine learning models were applied, the models that were considered were-logistic regression, random forest, gradient boosting and XGBoost. These models were evaluated using accuracy, F1-score and ROC-AUC metrics to assess their predictive capability. Experimental results indicate an economic significance assessment that the price momentum-oriented, dispersion-based indicators were the strongest predictors of short-term stock outperformance. On the other side, the market-relative features further improved the model's discriminative performance. Among all the evaluated models, XGBoost achieved the highest overall accuracy, F1-score and ROC-AUC which indicated its effectiveness in modeling non-linear interactions and managing correlated variables within high-dimensional financial data. The strategy produced positive risk-adjusted returns which also highlighted that the model captures signals with original financial value rather than purely statistical patterns. Overall, the findings illustrate the strong potential of machine learning techniques, particularly gradient-boosted decision trees, because of their enhanced short-term stock outperformance prediction.