Western Kentucky at Indiana. Lean: Western Kentucky +44.5. Lean: Under 60.

Spread.Take WKU +45.5 at FanDuel, displayed at -115 on the opened odds board. EL makes Indiana -30.2, but that 15.3-point difference is not a demonstrated edge. I place substantially more weight on the market: WKU's offensive struggles and defensive mismatch justify a huge spread. Still, Georgia's 70 points included defensive and kickoff-return touchdowns plus short fields; treating that score as repeatable offensive production overstates the mismatch. Indiana also emptied its bench against Howard, and Friday's conference opener creates a plausible incentive to shorten another comfortable win. That is an inference, not a promised substitution plan. The counterargument is substantial: Indiana's reserves scored touchdowns Saturday, Marsh participated despite the injury listing, and Cignetti expects Sharpe back. I cannot confirm WKU's starting quarterback for this game. My position is a small price-sensitive lean on late scoring and reduced possessions, not confidence that WKU can compete.

Total.Take under 60.5, available at BetMGM -105 on the opened board. The over case is real: Indiana has scored 52 and 55, WKU's defense has collapsed, warm weather presents little passing obstruction, and backup possessions can still produce touchdowns. The captured rise from 59.5 to 60.5 supports that concern. My slight under preference rests on WKU's inefficient offense and the possibility Indiana shortens the game after establishing control. Georgia's 90-point game depended partly on returns and short fields, making it a poor baseline. EL's 59.01 forecast supplies negligible support, and its wind deduction should not be treated as persuasive in light winds. A late WKU touchdown or another return score could easily defeat this modest lean.

Where our number looks wrong.Indiana -30.2 looks too low. The prior-heavy ratings understate Indiana's demonstrated scoring depth and WKU's current offensive dysfunction. Returning-production percentages cannot establish functional quarterback continuity or protection quality. Conversely, final-margin learning cannot separate Georgia's offensive performance from return scores and short fields. The apparent 15-point bargain should shrink substantially; it does not justify high confidence.

Spread.At 44.5 you are not betting on Western Kentucky, you are betting on how Indiana plays the last twenty minutes. Cignetti already pulled the plug early in a 55-0 win over FCS Howard, and the local preview framing this week is explicitly that he may again limit starters' meaningful snaps in the second half, partly to get corners Jamari Sharpe and A.J. Harris (seven combined snaps on the year) real work before Big Ten play rather than run up a score. Indiana needs roughly 52-7 to win this bet. The Hilltoppers are genuinely bad, but they are not shut-out bad: they put 20 on Georgia and 14 at Nevada, and Brock Glenn ran for two scores in Athens once Georgia's twos were in. Our model's -30.2 is too kind to WKU and I discount most of the -14.3 gap, because the rating is still mostly a preseason prior that has WKU at +0.4 and has never shown an ATS edge. What is left is the structural point: the number is the biggest on the board, the backdoor window is 15-plus minutes long, and I want the points.

Total.Our totals model is at 59.0 against a 60.1 consensus, but it applied a -1.57 wind drag off a captured 8 mph reading, and the NWS Bloomington forecast for Saturday is sunny, near 90, southwest 3 to 7 mph. Correct that one input and our own number is about 60.5, at or a hair above the market rather than under it. The path to the over is the same path that gets Western Kentucky the cover: Indiana has scored 52 and 55 against a defense that allows 59.5 a game, the Hilltoppers have scored 14 and 20 and did their Georgia scoring against backups, and a running-clock second half with Indiana's twos on the field tends to produce points at both ends, not neither. Risk is the Howard script - 55-0, one team scoring - plus a questionable kicker.

Where our number looks wrong.The -30.2 is 14 points off the market and I think the market is closer to right. The in-season rating still carries WKU's preseason prior of +0.4 (7.72 expected wins, 83.7% returning production), which two games have now falsified - losing by 35 at Nevada is the disqualifying result, not the Georgia game. The rating also cannot see that Indiana replaced its quarterback with Josh Hoover and is scoring 53.5 a game. So I am taking the same side our model takes, but for a different reason: not because Indiana is worth only 30, but because Cignetti stops trying in the fourth quarter. If Indiana keeps its starters in - and Cignetti has an AP No. 4 team with playoff seeding incentives - this loses by three scores.

Spread.Our model leans away at a 'high' tier (our number -30.2 vs market -44.5/-45.5), but I don't buy the lean here. Indiana was 16-0 last year and is outscoring opponents 53.5-8 through two 2026 games; Western Kentucky is 0-2 and being outscored 59.5-17, having just lost 70-20 to Georgia and 49-14 to Nevada. A 44.5-45.5 spread against a team conceding 59.5 ppg to a program averaging a 45.5-point per-game margin itself is not obviously stretched, it's close to arithmetic. Indiana's minor injury notes (WR Marsh, K Radicic both questionable per team reporting 9/12-9/14; CBs Sharpe/Harris day-to-day but backups getting praised from the coach) are depth-chart noise, not margin-relevant. WKU's own beat coverage shows their QB competition (Tisdale over Glenn) hasn't produced a offense that can hang with anyone real. The market has already priced the mismatch correctly in my view; our model's calibration is what looks stale against a team this dominant. Risk: IU pulled starters in both blowouts already, so garbage time could shave the final margin under the number even if the game is never competitive.

Total.Our totals model (59.0) and the market (59.5-60.5) are within a point of each other, well inside the model's own no-edge zone, so there's no real model signal to lean on. My lean to the under is about game state: Indiana has already pulled starters in both of its 2026 wins once the score got out of hand (55-0 over Howard, 52-16 over North Texas), and with a 44.5+ point spread implied here, expect the same script, a fast, high-scoring first half followed by backups and a running clock that caps the total below what a full-effort 60 minutes would produce. WKU's offense has managed only 17 ppg against real competition (Nevada, Georgia) and has shown nothing suggesting it can add garbage-time points at scale against a good Big Ten defense. Warm weather (upper-80s) and moderate wind are neutral factors. This is a soft lean, not a conviction position, since our own totals model shows no historical edge at this gap size.

Where our number looks wrong.The rating is fit on final margins and calibrated broadly across FBS; it has no mechanism for recognizing a team playing at a genuinely elite tier (Indiana unbeaten in its last 18 games, outscoring opponents by 45+ ppg) versus a team that is bottom-tier even by G5 standards (WKU outscored 129-34 by an average and a good team). The model's own documentation admits it has no proven ATS edge and is least reliable in extreme-mismatch games, which is exactly this spot, I'd treat its away lean as an artifact of calibration compression on large raw margins, not a real signal.