A decade of
drift & ascent.
Strength = attack + defense, rescored on the same 0–100 axis used elsewhere in this dossier. Each point is a complete refit of the model at quarter-end, using every match played up to that date with a six-year exponential half-life. Movement at the top of the table is genuinely small — a few points a year is a lot for an established team.
Strength · 0–100 scale
Higher = stronger · y-axis auto-fitThe y-axis auto-fits to the selected teams' range. Most established sides move only a few points per year, so expect bands of 5–20 points rather than a full sweep.
Exponential recency weighting means a match from 2020 contributes about half its evidence today; 2014 is closer to a quarter. Matches before 2010 are nearly cosmetic.
Each fit is independent — no temporal smoothing between quarters. Bumps you see are real Bayesian regret. A tournament hot streak doesn't appear until the next quarter rolls over.