2026 March Madness Elite Eight Picks & Predictions

Saturday, March 28

Two games separate four teams from Indianapolis. For the Elite Eight, we’re running both of our primary frameworks side by side — the efficiency-based model that has been our foundation throughout the tournament, and a power ratings and adjusted offensive/defensive power ratings (AOPR) model that generates three separate predictions based on full season data, the last four games, and the last seven games. Where the two frameworks agree, the signal is strong. Where they diverge, the picture gets complicated. Saturday delivers one of each.

Game 1: Illinois vs. Iowa

Illinois -7 | Total: 138.5

Efficiency Model

Illinois 77.8, Iowa 73.2

The efficiency model has Illinois winning by 4.6 — a clear Illini win, but nowhere near covering the 7-point spread. Illinois owns the highest offensive efficiency (1.232) of any team remaining in the tournament, and Iowa’s defensive efficiency (1.017) has been adequate without being dominant. But the Hawkeyes have been a nightmare to put away all tournament long. Their ability to control pace — they project at 64.8 possessions in a vacuum, helping drag this game down toward the 65-possession range — is exactly the profile that turns a comfortable efficiency edge into a single-digit final margin. Illinois wins, but the efficiency model says Iowa and the points.

The efficiency model’s projected combined score of 151 against a 138.5 total is a 12.5-point over gap — one of the larger over signals we’ve tracked all tournament.

Power Ratings / AOPR Model

This is where it gets genuinely interesting. The full-season power ratings model essentially matches the posted line — Illinois wins by 8, covers the 7. That’s consistent with what we saw in the Sweet 16, where full-season data aligned almost perfectly with the market. But the moment you shift to recent form, the picture flips entirely. Both the last-seven and last-four windows project Iowa winning outright — by a single point in each case, but winning nonetheless. Two independent recent-form timeframes pointing to an Iowa upset, while the efficiency model says Illinois wins but doesn’t cover.

The convergence of signals here is notable: not one of the four model outputs — efficiency model, last-4, last-7, or full-season — projects Illinois covering -7. The full-season power model is the only framework that even has Illinois winning by enough to approach the spread.

The read: Illinois is the better team on paper, and probably wins this game. But -7 appears to be a full-season number against an Iowa team that has been playing its best basketball of the tournament. The model consensus points strongly toward Iowa and the 7 points, with genuine upset potential if the Hawkeyes control pace and make it a half-court game in the final ten minutes. The over at 138.5 also has support from the efficiency model, though the power ratings total projections (ranging from 137 to 147) are more mixed.

Game 2: Arizona vs. Purdue

Arizona -6 | Total: 153

Efficiency Model

Arizona 80.3, Purdue 75.8

Arizona wins by 4.5 in the efficiency model — another case where the favorite wins but doesn’t cover. Purdue’s offensive efficiency (1.224) is the highest of any team left in the bracket, and Braden Smith remains one of the most impactful players in the country. Arizona’s defensive efficiency (0.941) is legitimate, but containing Smith while also managing Purdue’s frontcourt depth is a serious challenge. The efficiency model respects both teams, projects a quality game, and doesn’t see Arizona pulling away for a 6-point win. The projected total of 156.1 is a mild 3.1-point over lean against the 153 line — present but not a major signal.

Power Ratings / AOPR Model

The power ratings model tells a three-part story here. Full season: Arizona wins, but only by 2 — well short of -6. Last seven games: a literal tie at 82-82. Last four games: Arizona covers comfortably, winning by 9. The three timeframes point in completely different directions on the margin, though they all agree Arizona wins or ties. What’s striking is that the last-four window is the only one with enough Arizona-side conviction to justify the -6 line — and even there, the question is whether the last four games represent Arizona’s genuine peak or a hot streak that may not hold at this level of competition. The totals across power ratings windows range from 162 to 179 — all above the 153 line, reinforcing the mild over signal from the efficiency model.

The read: Arizona is the right side directionally — every model framework has the Wildcats winning or tying. But covering -6 requires a consistent level of separation that only one of the four model outputs supports. The efficiency model and two of three power rating windows project a 2-to-4-point Arizona win. Purdue and the 6 points is the consistent cover signal here, with the Boilermakers’ offensive firepower keeping them within striking distance regardless of how Arizona’s backcourt performs. If Braden Smith has a big game and Purdue controls their half-court possessions, this is a 2-possession game late. The over also has modest support across models, with most projections landing in the 156-165 range against a 153 total.

Saturday Summary

Both games project the favorite winning but not covering. That’s not a coincidence — Elite Eight teams are separated by incredibly thin margins, and the betting market prices these games with full-season data that the efficiency model largely confirms. The recent-form model is where the divergences show up, and on Saturday both divergences point toward the underdog covering.

Iowa covering Illinois and Purdue covering Arizona would be the outcomes most consistent with the overall weight of the model evidence heading into Saturday. Both games project as closer than the spread in multiple frameworks, and in Illinois/Iowa, two independent recent-form windows are calling an outright Iowa upset.

The Final Four picture will come into focus Saturday night. Enjoy the games.