The Math Behind Your Rating: How Nerd Leagues Calculates ELO
Every rating change on Nerd Leagues comes from one formula, applied once for every winner–loser pairing in a game. This article walks through it with real numbers. The plain-English version is MTG ELO Explained.
The expected score
Each pairing starts with an expected score: how likely the math thinks one player is to beat the other, based on the gap between their ratings.
expected = 1 ÷ (1 + 10^((opponent's rating − your rating) ÷ 400))
Two players at the same rating each have an expected score of 0.5. A 200-point gap puts the favorite at about 0.76 and the underdog at about 0.24. Every 400 points of gap multiplies the odds by ten.
Points for one match-up
The winner of a pairing gains K × (1 − expected), where the expected score is the winner's. The loser drops by exactly the same amount, so points move between the two players and none are created or lost.
In a 1v1 game, K is 20:
- Two players at 1000: the winner gains 10.0 and the loser drops 10.0.
- A 1100 beats a 900: the favorite gains 4.8 and the underdog drops 4.8.
- The 900 beats the 1100 instead: the underdog gains 15.2 and the favorite drops 15.2.
Upsets move ratings further because the result was less expected. Ratings are kept to a fraction of a point and shown rounded to whole points.
Commander pods: every winner against every loser
In a multiplayer game, the winner is paired with each loser separately, and each pairing uses the formula above. Losers are never paired with each other, so the order the losers were knocked out in doesn't affect ELO.
K drops by one for each player beyond two:
- 2 players: K = 20
- 3 players: K = 19
- 4 players: K = 18
- 5 players: K = 17
- 6 players: K = 16
In a four-player pod with everyone at 1000, the winner takes 9 points from each of three opponents for +27, and each loser drops 9.
Every pairing uses the ratings players had when the game started. A mixed four-player pod, with the winner at 1000:
- Against a 1050: +10.3
- Against a 1000: +9.0
- Against a 950: +7.7
The winner gains 27.0 in total. The three losers drop 10.3, 9.0 and 7.7: the highest-rated loser gives up the most, because they were expected to do best.
Two-Headed Giant and other team games
When a game has more than one winner, the players are split into teams. The number of teams is the number of players divided by the number of winners, rounded up, and K becomes:
K = (20 − (teams − 2)) ÷ (teams − 2, minimum 1) ÷ number of winners
Each winner is still paired with every player on the losing side. In Two-Headed Giant with all four players at 1000, K works out to 10 per pairing: each winner gains 10 across their two pairings and each loser drops 10. That is the same swing as a 1v1 game between two equal players. Teammates are rated separately, from their own numbers.
Same formula, separate scoreboards
Every rating on Nerd Leagues uses this calculation: your all-time rating, your yearly rating, your rating in each league and each season, and every deck's rating. Each scoreboard keeps its own numbers, so one game can move them by different amounts. A deck's change is worked out from the deck ratings at the table, not the players' ratings. Player ELO vs Deck ELO compares the two, Yearly ELO covers the annual reset, and MTG League ELO covers league standings.
When a game is archived
Archiving a game doesn't subtract its points. Ratings are rebuilt from 1000 by replaying every remaining game in the order it was played, so every later game is recalculated against the corrected numbers.