Expected Stats vs What Happened

An actual statistic records outcomes. An expected statistic asks: given the quality of the process — how hard the ball was hit and at what angle, how open the shot was, how far the kick was — what would typically result?

Neither is more real than the other. The outcome genuinely happened. But the expected version is usually the better predictor of the next stretch, because it strips out the part that will not repeat.

• Actual well above expected — outcomes have run in the player's favour. Expect the number to come down, not because he will play worse, but because he was not playing that well.
• Actual well below expected — the process has been better than the results. This is the more actionable direction, because the market is often slower to notice.
• Both moving together — the change is real. This is the case where something has actually changed.

The gap is a flag for a conversation, not a conclusion on its own.

They are models, and models have blind spots:

• They typically ignore who did it. A fast runner beats out grounders that an expected model treats as outs.
• They assume average context — average defence, average park, average conditions.
• They can be systematically wrong for unusual player types, precisely the players most often discussed.

Use them to ask why a gap exists. A player who beats his expected numbers every single season is not lucky every season; something about him is not in the model.