Throw darts at a target while adjusting bias and variance. Watch how they trade off and understand why the ideal model is both accurate and consistent.
Dart Settings
Quick Presets
Dart Board
How to read this: each dart is one model trained on a different random sample of data, and the bullseye is the true value it should predict. Bias = how far the centroid (average model) sits from the bullseye. Variance = how scattered the darts are around their own centroid.
Current Regime
Low Bias Low Variance
Accurate & consistent
← IdealLow Bias High Variance
Accurate but inconsistent
High Bias Low Variance
Consistent but wrong
High Bias High Variance
Worst of both worlds
The bias-variance tradeoff means reducing one often increases the other. Complex models have low bias but high variance; simple models the opposite.
MSE = Bias² + Variance + Irreducible Noise
High Bias (Underfitting)
Model is too simple. Misses patterns in training data. High training AND test error.
High Variance (Overfitting)
Model memorizes training data. Works perfectly on training but fails on new data.
Sweet Spot
Regularization, cross-validation, and the right model complexity minimize total error.