Week 0 Advanced Stats 2026: Virginia Dominated, NC State Collapsed, and TCU Was Flattered by the Score

A final score is one draw from a distribution. Two teams play, a few bounces go one way, and the number on the scoreboard becomes the thing everyone argues about for a week. So for all eight Week 0 games we did something different: we took what each team actually earned play by play, held that fixed, and asked what the result *should* have been.

Three of the eight winners played better than their final score says. Two of the losers played better than theirs. And the single most impressive performance of the weekend came from a team that was, on paper, the less talented side.

Virginia beat a more talented team by four touchdowns

Every team carries a recruiting-talent rating, and it is the cleanest available answer to *who should be better*. NC State's roster rates **757**. Virginia's rates **700**. NC State was the more talented team by a comfortable margin, on the road at a program that won three games two years ago.

Virginia won 34–8 and played like they won by 28.

**NC State is the most alarming result of the weekend** and it is not close. They had the better roster, they lost by 26, and the play-by-play says they deserved to lose by 28 — there is no bad-luck story available to them. Everything that happened, they earned.

Memphis is the quieter version of the same thing in reverse. UNLV had the more talented roster; Memphis won by six and played like nine.

Stanford was the best team nobody will mention

Stanford beat Hawai'i 37–27, which reads as a mildly comfortable win over a bad team. The play-by-play says they were **21 points better** — the largest gap between performance and result of any team in Week 0.

Where did the eleven points go? Some of it is ordinary variance. But **3.65 of it was their own fourth-down decisions**, the single most expensive set of calls anyone made all weekend. The worst was a punt from the Hawai'i 26 in the third quarter with an 18-point lead, after a pre-snap penalty turned a 38-yard field-goal attempt into a 43-yarder. Eighty-six percent of coaches kick there. Stanford punted from inside the opponent's 30.

Three coaches whose fourth downs cost more than you would guess

TCU's was the loudest. Trailing by two in the second quarter, they went for it on fourth-and-9 from their own 31 — where 85% of coaches punt — and were tackled in their own end zone for a safety. Two points to North Carolina, in a game TCU lost by five.

North Carolina's went the other way and nobody noticed: leading 3–0 in the first quarter, fourth-and-2 at the TCU **2**, and they kicked. It was good. In a five-point game, that decision cost 1.44 expected points — and a made 19-yard field goal is the last thing anybody blames.

Which winners would win it again?

Hold each team's actual per-play performance fixed, re-draw everything that does not repeat — fumble bounces, kicking variance, one-score luck — and replay each game a thousand times.

This is the least dramatic finding here and worth stating anyway: **there were no flukes in Week 0.** Every winner survives at least five replays in six. The upsets — North Carolina over TCU, Memphis over UNLV — were earned, not stumbled into.

What to actually take into Week 1

• **Virginia** is the buy. They beat a more talented NC State by four touchdowns and the play-by-play says they deserved every point of it.
• **NC State** is the sell, and there is no excuse available — they were outplayed as badly as they were outscored.
• **Stanford** won by ten and played like twenty-one. If you only read the score you got the wrong idea about that team.
• **TCU** lost by five and played like ten. The score was kinder to them than the game was.
• **Hawai'i** got eleven points of help from the scoreboard, the most of anyone.

One caveat worth stating plainly: this is one game each. Per-play performance stabilises faster than win-loss records do, which is why it is worth looking at in Week 0 at all — but a single afternoon is still a single afternoon, and the honest read on Virginia is 'that was real and we will know much more in three weeks.'