Methodology About The Math League Parameters

Calculation Methods

About BigFourElo

BigFourElo is an automated, data-driven platform that tracks the historical and current strength of professional sports franchises using the objective Elo rating system. By removing media bias and voter subjectivity, our model evaluates teams purely on their results, the quality of their opponents, and the context of each game.

The Mathematics Behind the Ratings

Every team starts their historical journey with a baseline rating of 1500. After every game, points are transferred from the loser to the winner based on the formula below.

1. Expected Outcome ($E$)

Before a game is played, the model calculates the probability of winning for both teams based on their current Elo difference, factoring in the Home Field Advantage ($HFA$):

$$E = \frac{1}{10^{-\Delta Elo / 400} + 1}$$

Where $\Delta Elo = (Elo_{home} + HFA) - Elo_{visitor}$.

2. Rating Update ($Elo_{new}$)

Once the game concludes, the ratings are updated using the actual outcome ($S$, where 1 = win, 0.5 = tie/OT-split, 0 = loss):

$$Elo_{new} = Elo_{old} + K \times (S - E) \times M$$

Where $K$ is the weight of the game, and $M$ is an optional Margin of Victory (MoV) multiplier that dampens or amplifies the shift based on the final score blowout.

League-Specific Parameters

Because each sport has a different season length, scoring pacing, and home-field impact, our core algorithm is finely optimized per league:

League K-Factor Home Adv. Inter-Season Regression MoV Constant ($MOV\_C$) / Rule
MLB 20 +30 Elo 1/3 (33.3%) 0.0 (Win/Loss only, score margin excluded)
NBA 20 +90 Elo 1/3 (33.3%) 0.10 (Points margin factored)
NFL 20 +55 Elo * 1/3 (33.3%) 0.60 (Points margin heavily factored)
NHL 20 +30 Elo 1/3 (33.3%) OT/SO win counts as a 0.5 coin-flip split
* Note on NFL: Home Field Advantage drops to 0 on neutral sites (such as the Super Bowl or International Series games).

What is Inter-Season Regression?

Between seasons, teams naturally change through the draft, free agency, and trades. To prevent historical ratings from locking teams into permanent baselines, every franchise is regressed 1/3 (33.3%) back towards the league average of 1500 before the start of a new season:

$$Elo_{start} = (Elo_{end} \times \frac{2}{3}) + (1500 \times \frac{1}{3})$$