2026 March Madness Second Round Preview

Saturday, March 21

The first round delivered. High Point pulled off one of the best 12-over-5 upsets of recent memory, knocking off Wisconsin, while VCU validated what our model projected by eliminating North Carolina. Both were among our model’s upset picks heading in, and both cashed — which makes Saturday’s slate considerably more interesting, because the same framework is projecting a few more results the betting market isn’t pricing correctly.

A quick reminder on methodology: these efficiency-based projections perform better across larger regular-season samples than in the one-and-done tournament environment. Use them as analytical context, not gospel. That said, the model has earned some credibility this weekend.

Also worth noting on the totals front: our under plays in the first round went 2-1, continuing a multi-year trend of hitting when there’s a significant gap between the model’s projected combined score and the posted total. There’s one qualifying under on the board for Saturday, and we’ll get to it.

---

Saturday Under Play

Michigan -12.5 vs. Saint Louis | Total: 161.5 ⬇️ UNDER

The model projects Michigan 79, Saint Louis 76 — a combined 155.5 against a 161.5 line. That’s a 6-point gap, which clears our historical threshold and matches the type of under signal that has hit consistently in first-weekend tournament action. Saint Louis came into this tournament 28-5 and knocked off Georgia in the first round playing controlled, deliberate basketball. Michigan’s projected pace of 72.2 and Saint Louis’s 73.5 suggest neither team is going to run. The model actually sees the Billikens keeping it surprisingly close — which, if true, only reinforces the under: tight games kill scoring volume. Michigan wins, but this total is too high.

---

The Games

Michigan State vs. Louisville | Total: 151 | Spread: MSU -4 Model: Louisville 76.7, Michigan State 75.3

The model calls the mild upset here — Louisville over Michigan State by a point and a half. The Spartans have been one of the most consistent programs in the country, but Louisville came into this tournament as a 6-seed with legitimate offensive efficiency numbers (1.171 OE) and has been undervalued most of the season. The projected pace of 69.2 keeps this in the low 150s, and the model total of 152 is essentially right on the posted line. The spread is the story: laying 4 with Michigan State when the model leans the other way is a hard sell.

---

Michigan -12.5 vs. Saint Louis | Total: 161.5 | Spread: Michigan -12.5 Model: Michigan 79.0, Saint Louis 76.5 — UNDER 161.5

Covered above. Michigan wins — but not by 12.5, and not at 161.5 points total if the model is right. The 6-point under gap is the best totals signal on Saturday’s board.

---

Duke -11.5 vs. TCU | Total: 139 | Spread: Duke -11.5 Model: Duke 78.7, TCU 65.0

Duke is the most dominant model output of the day at a projected 13-point win, which lands right in line with the spread. The efficiency numbers are stark: Duke’s 1.201 offensive efficiency and 0.915 defensive efficiency are elite, while TCU (1.083 OE) has struggled to score against quality competition. The projected pace of 69.1 keeps the total modest, and the model’s combined score of 143.7 suggests a mild lean toward the over at 139 — a 4.7-point gap, just under our formal trigger. Duke is the most trustworthy cover on the board if you’re looking for favorites to win big.

---

Illinois -10.5 vs. VCU | Total: 151.5 | Spread: Illinois -10.5 Model: Illinois 81.0, VCU 76.1

VCU earned this by beating North Carolina outright in Round 1, and the Rams won’t go quietly. But Illinois is built differently — their offensive efficiency of 1.232 is the highest mark among all eight teams on Saturday’s slate. The model projects a 5-point Illinois win, which means VCU covers even if they lose. The projected total of 157.1 vs. a 151.5 line is a 5.6-point over gap, which technically clears our threshold — though this is an over rather than an under, and over plays haven’t carried the same historical hit rate in our first-weekend data. Still worth noting if you’re playing totals on this game.

---

Nebraska vs. Vanderbilt -2.5 | Total: 147.5 | Spread: Vanderbilt -2.5 Model: Nebraska 75.4, Vanderbilt 75.3

Essentially a coin flip. The model projects a 0.07-point Nebraska edge — so close it’s functionally a pick ’em — while the market has Vanderbilt as a 2.5-point favorite. Nebraska has never won an NCAA Tournament game, which adds an interesting psychological layer to a matchup the model sees as dead even. Vanderbilt enters playing their best basketball of the season after the SEC tournament run, but their defensive efficiency (1.019) is a concern against any team with scoring ability. This projected pace of 69.8 and model total of 150.75 lands near the 147.5 line — no strong total signal.

---

Houston -10 vs. Texas A&M | Total: 142 | Spread: Houston -10 Model: Houston 79.9, Texas A&M 71.9

Houston wins by 8 in the model, which means Texas A&M covers the 10-point spread. The more interesting number here is the total: model projects 151.7 against a posted 142 — a 9.7-point over gap, the largest on Saturday’s board. This would normally flag as a strong over signal, though the same caveat applies as with Illinois/VCU — over signals haven’t been as clean in our historical first-weekend data as the under plays have been. Houston’s defensive efficiency (0.927) and slower 67.2 pace suggests they’ll keep the lid on somewhat, but the market may be dramatically undervaluing both teams’ scoring output here.

---

Arkansas -11.5 vs. High Point | Total: 168.5 | Spread: Arkansas -11.5 Model: High Point 86.3, Arkansas 81.5

Let’s just say it clearly: the model likes High Point to pull off a second consecutive upset. The Panthers beat Wisconsin in the first round and the same efficiency framework that projected that upset is projecting them to eliminate Arkansas as well. High Point’s offensive profile (1.181 OE) is legitimate — they averaged a 19.7-point margin all season — and Arkansas’s defensive efficiency (1.065) is a significant liability. The projected pace of 72.6 makes this a high-scoring affair, and the model total of 167.7 is nearly a perfect match to the 168.5 line. There’s no totals edge here — this is purely a result call, and it’s the most aggressive one on the board. Whether you believe in it is up to you, but this model called Wisconsin in Round 1.

---

Gonzaga -5.5 vs. Texas | Total: 147.5 | Spread: Gonzaga -5.5 Model: Gonzaga 84.5, Texas 71.8

Gonzaga covers comfortably in the model at a projected 12.7-point win, nearly double the spread. Texas (1.168 OE) has offensive capability but a defensive efficiency of 1.085 that has been exploited by quality offenses all year — and Gonzaga (1.179 OE, 0.914 DE) is exactly the type of team that exploits it. The projected combined score of 156.3 vs. a 147.5 total represents an 8.8-point over gap, another large over signal. At a projected pace of 70.1, neither team is especially slow, and if Gonzaga gets rolling early, this could become a track meet. The over at 147.5 appears underpriced if the model is anywhere close.

---

Summary

Designated under play: Michigan/Saint Louis under 161.5 (model: 155.5, gap: -6.0)

The High Point call is the headline, but the Saint Louis under is the play with the clearest historical backing. Enjoy Saturday.