opportunity to prove their merit based on the current season's performance. Starting with rankings influenced by prior-season records or recruiting ratings is equivalent to awarding bonus or penalty points to teams before the game even begins, effectively skewing the competition in favor of historically stronger teams.

Let me repeat that. Mathematically, the effect on the final rankings explained by the factors for previous season performance, recruiting ratings, etc. is equivalent to adding some number of points (positive or negative) to each team's score for every game in the season. This undermines the idea of a level playing field and the principle that rankings should reflect current-season performance alone.

As for the error concern, it's true that starting everyone equally may lead to less predictive accuracy early in the season, but predictive accuracy isn't the same as fairness. Imagine that race/ethnicity was predictive of claims severity/frequency. Would you suggest that a fair claims model should include it?

By mid-season, as more current-season data is incorporated, the model's accuracy improves naturally, while maintaining fairness throughout. The tradeoff is worth it to ensure rankings reflect on-field merit rather than historical advantage.