The paper explores ECG screening for Chagas, a parasitic disease found primarily in South America, as part of The George B. Moody PhysioNet Challenge 2025. Our approach is twofold: identify and suppress artifacts that are not indicative of Chagas and develop detection models based on power spectra and hand-engineered features. Across three datasets, SaMi-Trop (BR), CODE-15% (BR), and PTB-XL (DE), we show that 50 and 60Hz frequencies indicative of location can classify patients as having or not having Chagas at an above chance rate using a single-feature logistic regression classifier. Additionally, we show that such features are picked up by more sophisticated machine learning methods. Since such features are not truly representative of Chagas, we suppress them for the remainder of our analysis. Decision tree models are trained and evaluated on single features alone. These single-feature models evaluate samples from positive cases across the SaMi-Trop and CODE-15% datasets, as well as negative cases across the PTB-XL and CODE-15% datasets. Model error is evaluated with a Kolmogorov-Smirnov test, and features which reject the null hypothesis that both samples are drawn from the same set are rejected. A decision tree model is trained on remaining features, and rejected features are examined. As team Blue and Gold we achieved an official score of 0.091 with a decision tree classifier using spectral features (0-170Hz). A second decision tree classifier had an official score of 0.150 using both spectral and engineered features.