Stats Model Retrospective: 1993 Sugar Bowl

NT17 brought this up, and with the big game today, I thought I'd share my findings. I realize, of course, that it's VERY EASY to retroactively make something look better than it is. That was NOT my intent here. My intent was to use the stats model I'm trying to polish a bit and see what the results were.


What a lot of you don't know about that time was that Alabama was being ridiculed as not worthy of being in the game because "everybody knows" that Florida State and Miami "are the best two teams." Alabama - not that much different than Michigan St a few years back - got into the national title game basically by virtue of being an undefeated conference champion (I realize, of course, the bowls complicated matters much more back then).

And what my findings show is that this was ALWAYS going to be a closer game than the so-called analysts realized. No, it did not give me the 34-13 final, but come along with me as I explain where things went awry.


The model ONLY predicts "expected performance based on average." It gives a range, and I tend to choose a number in the middle of the range. And this won't work prior to December because you don't have enough data points, although perhaps it could be developed as the season progressed. The "obvious" in retrospective SHOULD have been obvious in 1992, but I can explain why it wasn't. Before I do that, though, let me explain the data. I only removed the Florida A/M game from Miami's list, and I used the La Tech game for some Tide stats but not others - generally to see how much it affected things, which wasn't much (after all, the Dogs scored 0).


Miami's Offenive PPG was 31.8 while their defensive PPG surrendered was 12.7. Those are very balanced and good numbers overall.


But perhaps the biggest "secret" was Alabama's offense that was MUCH BETTER than was recognized at the time.

Alabama's offensive PPG was 27.6 while the defense surrendered an average of 9.1 PPG.

Miami's average for the season was that their offense tended to score 9.9 points ABOVE the normal opposition average. This meant Miami on offense should be expected to score 19 points.

Alabama's defense on average held teams 9.2 points BELOW their average. This meant Alabama's defense vs Miami's offense would yield 22.6 points to Miami. This meant Miami's scoring range was between 19 and 23 points.


Alabama, by contrast, averaged 8.4 ppg above what the opposition allowed. This means (8.4 plus Miami's average of 12.7 means Alabama could be expected to score 21.1 points.

Miami's defense held opponents to 12.5 points below their normal average, so Alabama's second point expectation (27.6 - 12.5 = 15.1), so Alabama's scoring range with an AVERAGE performance would be 15-21 points.


SCORING RANGE
Miami 19-23
Alabama 15-21

Keep in mind that at this point I've made zero compensation for the relative conference strengths of the SEC or Big East. Let me explain the details of this in the next post. Note: Alabama was an 8-point underdog and THAT was considered generous. Oh yeah, and 23-15 is........

So the game was, in essence, either a toss-up or perhaps Miami as a slight favorite. Why did the pundits miss it? Well, there are a lot of reasons, some okay and some the product of laziness.


1) Miami had the big name.

There's no doubt this is true. Miami had won 3 national titles in the preceding five years, lost the title game on the last play in another, lost another title thanks to one controversial loss to Notre Dame......and even in 1990, the writers thought Miami would have rolled either Colorado or Georgia Tech, who split the title.

Miami had the name. And because they had that name, the writers who were too busy mocking Corky Simpson for saying Alabama was better never stopped to actually LOOK at the data. If they had, they would have seen some points I make below. But it is true that Miami was a good team and had the powerhouse name.

2) They allowed themselves to be dazzled - once again - by offensive powerhouse numbers.

They fell once again for eye-popping numbers on offense, not so much due to Miami as because of Florida State. Miami's national titles from 1987 onward were viewed through the prism of FSU's offense. Their logic went like this: "FSU has a high octane offense, Miami beat FSU, therefore, Miami has a powerhouse offense." This level of analysis is why Gino Torretta became one of the most undeserving Heisman winners in history.

But the Canes didn't "really" have this in 1992. Yes, they averaged 31 ppg, but look closer at those numbers. In a four-game stretch in mid-season, Miami scored 171 points in four lopsided wins and added 63 more against San Diego State. Those five opponents had a combined record of 15-35. Subtract those games against inferior competition, and Miami only averaged 16 ppg against good teams and - in fact - their performance average on offense drops from 8 points above to 0.2 points below.

But let's set that aside for the moment. What basically happened was the media narrative was, "Florida State is playing the best ball" and they'd look at the Noles scoring 139 points in two games in November and then punching 45 in against Florida and conclude, "Miami has a good offense - after all, they outscored FSU."


Coach Stallings was the one guy I heard say it clear - "Miami's strength wasn't their passing game, it was their DEFENSE." He was right.

3) They disparaged the SEC somewhat unfairly as a run only conference and missed Alabama's offense.

I'll bet most of you were shocked to find out that Alabama's offense was almost as good as Miami's. And then consider this: Alabama played two defenses - Ole Miss (13 ppg allowed) and MSU (14.6) - that were about as good as Miami's. You have to remember that while the Cane D was good, it obtained some of those numbers with games against those also rans mentioned above. Consider Alabama's performance against those two defenses - the Tide's supposedly poor offense punched in 31 and 30 points (though one TD in the MSU game was a blocked punt). So it would not have been unrealistic to expect the Tide's offense to score 24 points on Miami. The Tide OFFENSE, in fact, scored 27 points, their season average on Miami.

4) The George Teague play is the anomaly that reduces Miami's score from the expected range.


With Miami's range falling between 19 and 23 points, I would have opted for 21. The Canes, of course, got 13. They also lost 7 when a defender did not give up on the play but instead ran him down and ripped the ball away from him. A Miami TD there gets them a final score of 20 points.

Of course, I realize you can't exactly do it that way, but there's no question that play saved some Miami points. If he only tackles Thomas and the defense holds, Dane Prewitt was a decent enough kicker that they get three (most likely).

5) The Difference Between Conferences Was Enough To Justify A Huge Variable

Miami's schedule was not bad in 1992. No doubt scheduled years before, they played two historically decent Big Ten teams (Iowa and Penn St), Florida St, and their Big East schedule.

The problem is that the Big East simply wasn't very good. They had three decent teams: Miami, Syracuse, and BC. BC got rolled by Tennessee in the bowl game. Syracuse did beat Colorado, but the Big 8 was even worse than the Big East. When you have only 3 bowl teams and only KANSAS wins a bowl game....your conference isn't very good.

By contrast, the SEC of 1992 was, in fact, a juggernaut. Six teams made bowl games, five won (three in routs). In all sincerity, if you had put Alabama, Ole Miss, or MSU defense in place of Miami's with that schedule, it is very likely all 3 would go unbeaten or lose no more than one game.In context of the defenses they faced, Alabama's offense in 1992 was actually "better" than Miami's was.

Folks just didn't notice it because they got caught up in glitz and flash.

6) The look back showed me things I didn't know.

Although they started horribly, Auburn's 1992 defense was NOT a bad defense. In fact, the Tigers held Alabama just below their average points surrendered for the year. Of course, the resignation of Coach Dye the night before no doubt lent an emotional element to the game for them. But they were a decent defense with no offense. Five of Alabama's SEC opponents in 1992 had substantially better defenses than they had offenses and another (MSU) basically had about the same on each side of the ball.

The Tide's BEST performance by points allowed in 1992 (pre-Miami) was the Tennessee game. The Vols were the best offense in the SEC that year, and Alabama held Heath Shuler to 10 points. (Technically, the game against USM was the best overall defensive performance for the year, but the game was close thanks to a turnover). The Tide's WORST defensive performance of the year will surprise nobody - Miss State. The Dogs got six more points in that game than they averaged the rest of the season (cupcake excluded).

MSU was the ONLY team the entire year to exceed their offensive PPG in their game with us. Not even Florida managed that.

78Alum said:

What does the stats model predict for our game tonight?

Click to expand...

54-35, us.

As a reminder - there are things it CANNOT predict or compensate
- injuries to key players
- turnovers
- penalties that negate scores
- bad officiating calls that negate/allow scores
- insanely stupid decisions by coaches/players
- an EXTREMELY good performance by one team concurrent with an extremely bad performance by the opponent


The 1993 Sugar Bowl is remembered as an Alabama blowout. Not only was the game stats-wise closer than any of the big-time pundits at the time suggested, but if you take away TWO plays by George Teague (the pick six and the strip), the score is 27-21, Alabama. Which would IN FACT have been within expected reason based on the overall performance of both teams in 1992. What inflated Miami's stats were the four-game stretch after FSU and prior to Syracuse.



Alabama's OFFENSE, most specifically the running game - exceeded what would have been expected.

Assessing it by the previously obtained numbers, the defense (which was given due credit) played precisely as they had all year regardless of the opponent - they didn't "really" do anything more spectacular against Miami than they had all season, it just wasn't expected and was viewed by millions on TV. The Alabama offense exceeded expectations and forced the Miami defense to play its game.

(Note: that last sentence is Oklahoma's hope for this game. Don't think that because the model puts a wide range score that it means OU cannot win. They can, but they will have to score a bunch and get some stops - which is what every pundit is saying is their chance).

NationalTitles17 said:

Very interesting retrospective all around.

Some things in football can't be accounted for in stats: the intangibles. Turnovers change a game. Vegas lines aren't a reflection of relative strength, but of perceived relative strength (per the masses).

That Alabama defense was loaded with elite talent. Miami had some stand-outs as well, but that was an all-time defense.

The compensation for relative SoS or conference differences would make the analysis even more interesting.

I didn't expect this today and appreciate it. My question to you on this was one of curiosity on how this prospective model would work with known results (or in a retrospective manner). That's always an interesting endeavor (well, not always) and can make or break a statistical model.

Thanks again for taking the time and for posting.

Click to expand...

Any time man. Learning to do it quicker as well. Have discovered a few ways to expedite it after doing it longhand.


Another intangible in this case was that it was essentially an Alabama home game. Per Keith Jackson, it was about 7/8 Alabama crowd. And it sounded like it on TV, too. A home field usually gets you 3 point advantage.

Another one is that I'd have to run the turnover numbers. That's how I figured that we were going to be plus 2 against Clemson two years ago. On average, they turned the ball over a lot and we GOT a lot of turnovers. The key play of the entire game - more than anything - was Clemson stopping Ryan Anderson's fumble recovery from becoming a touchdown.

The point, though, is that the 93 Sugar Bowl was always going to be a closer game than was suspected, even if Miami won.

cbi1972 said:

Putting 11 men on the line of scrimmage daring Torretta to throw was unexpected AND spectacular.

Click to expand...

That's in terms of strategy.

My comment was in terms of total performance.

RTR91 said:

Yep. Once I have the formulas down, I just copy the spreadsheet and change the numbers for the respective team. Much easier that way.


Sent from my iPhone using Tapatalk

Click to expand...

Yeah, I just figured it out, ha ha....