Essay 3 · interactive figure mockups · not for publication
Five figures for the Big Ten and SEC essay
Figure 1 now lives on its own page (figure1/). Each figure opens with one sentence that says what to notice; tap or hover anything for details. A dashed note under each figure says what is still to come.
Figure 1 · the bridge mapserves: the links · Notre Dame · Elo · leave-one-out
Every 2026 game, and the paths that compare two leagues
Still to comeIn the essay, scrolling steps through the four states in place of the buttons. On phones the map opens as ten league dots that expand into teams when tapped. The page uses every 2026 game played through week 10. Drag a dot to pull the network around.
Figure 2 · the game cardserves: cold open · Auburn's upset · rivalry appetizer
Georgia is favored to beat Georgia Tech by about 30 points this year. A forecast is three parts added together, and one of them is simply which league each team plays in.
Georgia against Georgia TechGeorgia against AuburnTexas against Ohio State
Left card: the 2026 game, or the last meeting. Right card: any past year; click a bar in the strip below.
Figure 2. Three parts of a forecast. A forecast is three parts added together: how much better one team's league is than the other's, where each team stands inside its own league, and home field. Rivals from the same league, like Georgia and Auburn, have a league part of zero; Georgia's 11.8-point edge over Georgia Tech comes from the SEC against the ACC. In the strip, each bar is the real result and each dot is the forecast; red bars are years the forecast picked the wrong winner.
Still to comeIn the essay the Tech card opens the piece alone, and the Auburn card arrives at the upset beat. Every chance uses the same conversion as the 2026 cards, sized to each game's own rating uncertainty.
Figure 3 · the gap, two waysserves: verdict · how sure · Elo · rank by rank
In 2026 the average SEC team is about 5 points better than the average Big Ten team (somewhere between 1 and 9). The two best teams are about even; the gap is in the teams just below them.
2026
Still to comeClick a season on the first view to open it rank by rank. In the essay, the Elo lines draw in at the Elo beat.
Figure 4 · scoreboard of methodsserves: outside checks · Elo · the Machine
Our rating predicts games better than five versions of Elo, the most common rating system. Only Vegas does better.
Figure 4. Prediction scores, lower is better. Each column is one test. Each dot is a rating, placed by how far off its win chances were on games it had not yet seen; further left is better. Red is our network rating, the grey dots are five Elo ratings, and the white ring is Vegas. Hover a dot for its score and its gap behind ours.
Still to comeA details toggle that opens the full methods table, one line per model we ran.
Figure 5 · every matchupserves: the ending
Pick any two leagues to see how often a typical team from one would beat a typical team from the other.
2026
Figure 8. Every pairing of leagues. Each cell is the chance that the row league's team beats the column league's team on a neutral field. Red means the row league is favored; dark grey means it is the underdog. The small number in a cell's corner is how many times the two leagues met that season. Click a cell to see the pair rank by rank.
Still to comeA rank-by-rank chip that matches the leagues' #1 teams, #2 teams and so on; the side panel already does this for one pair.