3 in 4 chance of undefeated regular season?

You can make numbers and statistics say whatever you want. 23.4% of people know THAT!

Pass_The_Rum said:

So the average percentage must be lower than .9, right? No, because then the number of 15-1 and 14-2 teams becomes a statistical abberation of its own. The probability for a team to go undefeated is too complicated to figure by such simple approximations.

Click to expand...

15-1 is exceedingly rare. It's only happened 4 times in the 30 years since the league went to 16 games 30 years ago. We can assume that this means there have been about 4 teams with about a .93 win probability per game. Such a team would have had about a 30 percent chance of being undefeated through the regular season. So we should expect that roughly one in three such teams would end at 16-0, all else being equal. Is it surprising that so far it's zero out of four? No. It's a very small sample size. While you would expect at least one occurrence out of four, it's certainly not an aberration that there isn't.

14-2 isn't quite as rare. It's happened 16 times in 30 years. Let's thus suppose this means that there have been 16 teams in the last 30 years with a .875 win probability per game. You'd expect about 11 percent of such teams to go undefeated. Again, the fact that there have been 16 trials with an 11 percent chance, but none has gone undefeated is mildly surprising but not abberant.

Even combining these 20 teams, I still think you're well within the zone of reasonableness here. Stats would suggest that about 1.7 of these teams should have gone undefeated. Let's call it 1.9, to take into account the 10, 11, 12, and 13 win teams who all had some non negligible chance to go undefeated. I think the most you can say it's a bit surprising, but not abberant.

To the extent it appears like an abberation, I would agree with the other posters who suggest that most 14-2 or 15-1 teams, as the season winds down, have less incentive to win.

Ahh come on guys. You don't need stats. All you need to do is watch the games. We will go 16-0 because we are simply unbeatable. If Indy couldn't beat us with the refs on their side while keeping our offense to only 24 points, with a 10 point lead in the 4th, no one will. Of course there is the possibility that we lose a game or two, but its not going to happen. Just watch the games. I can't find a single reason other than horrible HORRIBLE, HORRIBLE (how can I emphasis this...) luck that we would lose a game.

PatsFaninAZ said:

15-1 is exceedingly rare. It's only happened 4 times in the 30 years since the league went to 16 games 30 years ago. We can assume that this means there have been about 4 teams with about a .93 win probability per game. Such a team would have had about a 30 percent chance of being undefeated through the regular season. So we should expect that roughly one in three such teams would end at 16-0, all else being equal. Is it surprising that so far it's zero out of four? No. It's a very small sample size. While you would expect at least one occurrence out of four, it's certainly not an aberration that there isn't.

14-2 isn't quite as rare. It's happened 16 times in 30 years. Let's thus suppose this means that there have been 16 teams in the last 30 years with a .875 win probability per game. You'd expect about 11 percent of such teams to go undefeated. Again, the fact that there have been 16 trials with an 11 percent chance, but none has gone undefeated is mildly surprising but not abberant.

Even combining these 20 teams, I still think you're well within the zone of reasonableness here. Stats would suggest that about 1.7 of these teams should have gone undefeated. Let's call it 1.9, to take into account the 10, 11, 12, and 13 win teams who all had some non negligible chance to go undefeated. I think the most you can say it's a bit surprising, but not abberant.

To the extent it appears like an abberation, I would agree with the other posters who suggest that most 14-2 or 15-1 teams, as the season winds down, have less incentive to win.

Click to expand...

Actually, given 16 instances of a team with a .875 winning percentage (11.8% chance of winning them all), the odds against none going undefeated are 88.2% raised to the 16th power, or 13.4%. You would expect two undefeated teams. That's not even counting the 15-1's.

It is an abberation, by which I mean that actual football departs from what we'd expect just crunching numbers. Real-life factors make the outcome of games interdependent, such that the crude analysis above falls short.

...I'm tempted to chart NFL wins by season and see if they fall on a nice Gaussian curve. **fires up spreadsheet**

yeah...that Gauss has a NASTY hook...Sawx should sign him

There are many problems with this guy's approach.


A much more sophisticated approach can be found at Football Outsiders.

They give us a 22.8% chance of going undefeated and a 32.5% chance of winning the Superbowl. (They haven't published odds for doing both.)

I think these odds are low. I attribute this to an inadequate discounting of our performance in games we have already won (Especially the Cleveland and Miami games).


Accuscore gives us a 54% chance of going 16-0. I think this is about right, but I have much more serious problems with their methodology that I do with FO.

Actually, I think the odds for your team to go undefeated are about one in three, maybe a bit better.

Incidentally, I went ahead and sorted out the number of wins in the league, using the years 1997-2006. It doesn't really follow a normal statistical distribution. While it tapers off nicely on the left like you'd expect, there's an almost linear decline from 11 wins to the 0 mark at 16 wins. The middle is much flatter than I'd expect, with a noticeable crater in the graph at one point. Which leads me to the following conclusions:

1) Really crappy teams win by pretty much by luck alone.
4-loss and worse teams are hopeless. This is the only part of the graph that resembles a random-chance distribution, as if mindless automatons were playing the game. Mindless automatons usually is a good description for these teams, their coaching staff, or both.

2) Teams really, REALLY try hard to not be a 7-9 team.
There's a noticeable crater in the graph at 7 wins, and a corresponding jump at 8 wins. Teams managing to eke out the extra win to not have a losing record is the only departure from a flat line running from 6 wins to 10 wins.

3) 11 wins seperates the men from the boys.
There's a noticeable drop-off from the 6-10 win plateau to 11 wins, after which there's a linear decline with increasing wins. That linear decline seems to be a function of coaches calling off the dogs and resting starters at various points depending on HFA and clinched byes.

Pass_The_Rum said:

Actually, I think the odds for your team to go undefeated are about one in three, maybe a bit better.

Incidentally, I went ahead and sorted out the number of wins in the league, using the years 1997-2006. It doesn't really follow a normal statistical distribution. While it tapers off nicely on the left like you'd expect, there's an almost linear decline from 11 wins to the 0 mark at 16 wins. The middle is much flatter than I'd expect, with a noticeable crater in the graph at one point. Which leads me to the following conclusions:

1) Really crappy teams win by pretty much by luck alone.
4-loss and worse teams are hopeless. This is the only part of the graph that resembles a random-chance distribution, as if mindless automatons were playing the game. Mindless automatons usually is a good description for these teams, their coaching staff, or both.

2) Teams really, REALLY try hard to not be a 7-9 team.
There's a noticeable crater in the graph at 7 wins, and a corresponding jump at 8 wins. Teams managing to eke out the extra win to not have a losing record is the only departure from a flat line running from 6 wins to 10 wins.

3) 11 wins seperates the men from the boys.
There's a noticeable drop-off from the 6-10 win plateau to 11 wins, after which there's a linear decline with increasing wins. That linear decline seems to be a function of coaches calling off the dogs and resting starters at various points depending on HFA and clinched byes.

Click to expand...

Really interesting stuff. Thanks!

Pass_The_Rum said:

Actually, given 16 instances of a team with a .875 winning percentage (11.8% chance of winning them all), the odds against none going undefeated are 88.2% raised to the 16th power, or 13.4%. You would expect two undefeated teams. That's not even counting the 15-1's.

Click to expand...

Yes, that sounds about right. Mildly surprising, but not abberrant, though. Unless by "abberant," you're meaning something other than 2 standard deviations.

I think the thing that strikes me as most 'wrong' about the proposed statistical math are just the idea of 90% chance of victory. Honestly, I would be tempted to not give those odds even when we play the jets or miami. I think if you figure 80% you're being VERY generous. On average, I'd say we have maybe a 75% chance of winning each game going forward, which gives us about a 7% chance of winning 7 in a row. (.75^7) -- in my opinion that's a remarkably high number and probably overly generous.

Still, whatever number is laid out there might as well be based on the number of berries in the rectum of a virgin goat or augured from the entrails of a fatted quail. It just doesn't have any basis in reality at this point.

PatsFaninAZ said:

Yes, that sounds about right. Mildly surprising, but not abberrant, though. Unless by "abberant," you're meaning something other than 2 standard deviations.

Click to expand...

I don't think you look at a teams final winning percentage as their percent change of winning each game. More likely, one looks at the Vegas odds on a week to week basis and use those. (at least looking at a historic context, obviously this isn't possible for projections)

Winning percentage reflects actual performance not the chances that something could have happened. Teams under and over perform all the time. It is conceivable that there have been 13-3 teams that should have been undefeated and 15-1 teams taht should have been 12-4 statistically speaking.

RodRustsGhost said:

IWinning percentage reflects actual performance not the chances that something could have happened. Teams under and over perform all the time. It is conceivable that there have been 13-3 teams that should have been undefeated and 15-1 teams taht should have been 12-4 statistically speaking.

Click to expand...

That's true, but I think this method takes that into account. There have been 16 teams in 30 years that have gone 14-2. What I said (and I tried to be careful about it) was that this tends to suggest that there were about 16 teams that had about an .8725 win percentage.

You're correct that they might not have been the SAME teams. That is, a couple of these teams may have overperformed and it might have been 13-3 teams that same year that actually had the .8725 win percentage. Or that a 15-1 team actually overachieved and really had a less high win percentage but got lucky one or two games.

Still, I think that 16 is the right number to start with. View it as an average.

The point about using actual performance to work backwards and predict win percentage is also an ok one, but I disagree. I think it's a really good proxy.

Think of this way, forgetting about the stats. In 30 years, there have been 16 teams good enough to get to 14 wins and 4 teams good enough to get to 15 wins. One would think any team good enough to get 14 or 15 wins was also good enough to have at least a shot at 16 wins.

I think that's right. In fact, experience backs it up. The 1985 Bears could have done it. Some of those San Francisco teams could have done it.

The fact that none have, is interesting. Zero for 20. One might think somewhere in there one of those teams should have gotten lucky and nailed that last game or two. Statistics are used to ask whether this is just surprising or whether there is "something going on here" that makes 16-0 categorically different from 15-1 or 14-2. The answer to that seems to be, at this point, "unclear." Two standard deviations is where we would start to say, "hmmm, this really is unusual and either our methodology is faulty or there is something categorically different about that last win that makes 16-0 significantly more difficult, not just incrementally more difficult, than 15-1 in a way that confounds us."

What might be a better way to look at it would be as a winning streak. How many times in NFL history has any team won 16 in a row? 15-1 could be 7-1-8 or any combination of 2 winning streaks. The longest was the pats at 21 or whatever. The dolphins were around 19 or something, right? how many times in the history of NFL have there been 16 game winning streaks? Now the odds of that happening starting with the first game and ending with the last applied, it probably makes it a statistically VERY rare event. (which seems to be supported by reality.

RodRustsGhost said:

What might be a better way to look at it would be as a winning streak. How many times in NFL history has any team won 16 in a row? 15-1 could be 7-1-8 or any combination of 2 winning streaks. The longest was the pats at 21 or whatever. The dolphins were around 19 or something, right? how many times in the history of NFL have there been 16 game winning streaks? Now the odds of that happening starting with the first game and ending with the last applied, it probably makes it a statistically VERY rare event. (which seems to be supported by reality.

Click to expand...

Way too many variables here. Teams personnel changes year over year -- not just the subject team but its opponents as well. Schedules are weighted at times against teams that have performed well the year before. If you're including playoffs in these streaks you're talking about games where the opponents are specifically selected precisely because they are good and thus have a high chance of beating the opponent.

I'm also not really sure what the "in a row" adds. I think the proposition being advanced is that a team that wins 15 games out of 16 was one that had at least a decent chance of winning the 16th as well. The fact that there have been 4 of these and none has done it is an interesting fact. What does it mean? My only point is that statistics would probably tell us, at this point, that we don't know the answer. If we get to a point where we have had a three or four more 15 win teams but no 16 win teams, that's where statistics would start to say that something needs to be investigated.

Actually, I based my math off already completed, on the books seasons and used those figures as representative of dominant teams. Those games actually happened, so... not sure what your point is, unless you're only criticizing the original poster.