Interesting Stat, That Shows Football Stats Arent That Meaningful

[Ring 6]

Per profootball reference

The Patriots played 1024 snaps on defense this year, which are all combined together to give total defensive stats and rankings.
They played 572 of them (56%) when they had above an 80% probability of winning. They played only 109 (11%) with less than a 40% probability of winning.

I'm not sure if that means their stats are overrated or underrated, because you could make an argument either way, but they are not apples to apples with other teams.

To compare, the Eagles played 38% with a 80+ probability and 18% with under 40%.

Jacksonville played 36% at 80+ and 30% at 40-.

The Bills played 20% at 80+ and 49% at 40-.

Finally for comic relief I will add the Browns.

The Browns played 1026 snaps of defense. On 57% of them they have less than a 20% probability to win, on 88% they had less than 40, and on a whopping 23 snaps, or 2% they have an 80% or better probability of winning.

 

[Chris Stevenson]

numbers are numbers when studying anything...they only tell part of a story.

anyone who relies only on numbers, does not believe in critical thinking or they “why” as to how those numbers are the way they are.

 

[Nikolai]

numbers are numbers when studying anything...they only tell part of a story.

anyone who relies only on numbers, does not believe in critical thinking or they “why” as to how those numbers are the way they are.

Stats on their own are no good without a check on the results. Some would call it a common sense check, but sometimes it's the opposite that's needed.

In my little "****ing with Stats"/prediction project, I'm learning a lot more about assumptions and the right ones to make or the wrong ones. For example, I had two possibilities for Bortles' performance; 197 yards or 264 yards, and I went with the "common sense" angle of leaning toward the lower number. Whoops.

I have to remember that it's not the end of the world if I challenge conventional wisdom, but also important not to get too cute. There's a balance in there and I'll get better, but it's been an interesting hobby so far.

 

[Chris Stevenson]

Stats on their own are no good without a check on the results. Some would call it a common sense check, but sometimes it's the opposite that's needed.

In my little "****ing with Stats"/prediction project, I'm learning a lot more about assumptions and the right ones to make or the wrong ones. For example, I had two possibilities for Bortles' performance; 197 yards or 264 yards, and I went with the "common sense" angle of leaning toward the lower number. Whoops.

I have to remember that it's not the end of the world if I challenge conventional wisdom, but also important not to get too cute. There's a balance in there and I'll get better, but it's been an interesting hobby so far.

basically, i use the eye test in conjunction with the numbers. if the numbers don’t really align with what actually happens during then game, the yes, i still lean towards the eye test over the numbers.

it’s general motors telling us “all is well” from 1990-2005, meanwhile all we saw were more and more japanese or south korean car
manufacturers on the road or at stoplights, more and more and more...voila, 2007/2008 rolls around, and gm needs a bailout because they misrepresented the numbers to the stockholder.

pretty simple. anyone can re-arrange numbers for how they see fit.

the metrics to judge a d on “yards allowed” in this modern day of football is one the dumbest stats in all if sports, if not the dumbest.

 

[Nikolai]

basically, i use the eye test in conjunction with the numbers. if the numbers don’t really align with what actually happens during then game, the yes, i still lean towards the eye test over the numbers.

it’s general motors telling us “all is well” from 1990-2005, meanwhile all we saw were more and more japanese or south korean car
manufacturers on the road or at stoplights, more and more and more...voila, 2007/2008 rolls around, and gm needs a bailout because they misrepresented the numbers to the stockholder.

pretty simple. anyone can re-arrange numbers for how they see fit.

the metrics to judge a d on “yards allowed” in this modern day of football is one the dumbest stats in all if sports, if not the dumbest.

Great example, and I agree 100% that judging a defense by yards allowed is a terrible method. Much better, but still not best is by points. I've become a big fan of yards/point as a measuring stick and have found that to be the most reliable predictor of future performance; or at least, its utility in a comprehensive analysis is far greater than yards or points allowed. I've gotten into comparing efficiencies between offenses, QBs in particular, and defense. It's working out pretty well so far.

 

[Deus Irae]

Stats are fine, and they can be quite meaningful. You just have to understand the context, and you have to understand that few stats have a 1:1 correlation to winning or losing.

The NFL turnover stats are a great example of this.

 

[Chris Stevenson]

Great example, and I agree 100% that judging a defense by yards allowed is a terrible method. Much better, but still not best is by points. I've become a big fan of yards/point as a measuring stick and have found that to be the most reliable predictor of future performance; or at least, its utility in a comprehensive analysis is far greater than yards or points allowed. I've gotten into comparing efficiencies between offenses, QBs in particular, and defense. It's working out pretty well so far.

the other thing is, when these supposed elite defenses get so much praise, do they have the depth to hang in a dome during a superbowl and against a pats level offense?

it is sort of why you need to build a lethal offense in this era, but at the same depth and versatility is more important on a defense than a few elite players like with what a philly has on that d.

we saw it with atlanta last year..i think jones and neal, two rookies, very good futures, nevee took that many snaps in their life and they had no depth behind their starters to give those guys a breather, out of fear of the drop off being too severe.

so, sure they are fast and atheletic, but it certainly didn’t matter...seattle, same thing...jax, same deal...

these gms are technically building teams in outdated fashion.

 

[Ring 6]

Stats on their own are no good without a check on the results. Some would call it a common sense check, but sometimes it's the opposite that's needed.

In my little "****ing with Stats"/prediction project, I'm learning a lot more about assumptions and the right ones to make or the wrong ones. For example, I had two possibilities for Bortles' performance; 197 yards or 264 yards, and I went with the "common sense" angle of leaning toward the lower number. Whoops.

I have to remember that it's not the end of the world if I challenge conventional wisdom, but also important not to get too cute. There's a balance in there and I'll get better, but it's been an interesting hobby so far.

So your 2 guesses were 35% apart, the difference between the average passing yards for a terrible Qb and close to leading the league.
See this is my point. You are using a statistical model that uses statistics against different opponents and in very different situations and trying to apply it in context of the defenses cumulative stats vs different opponents in different situation to come up with some kind of statistical guess that doesn’t account for matchups, game situation or anything unusual happening.
I can see it could be a fun hobby but football just isn’t built to use statistics to figure out how a game will go.
That was the point of my OP. If it’s a close game or the patriots are behind his much do their defensive stats mean when 57% of them occurred when the gane was basically on ice?

 

[Ring 6]

Stats are fine, and they can be quite meaningful. You just have to understand the context, and you have to understand that few stats have a 1:1 correlation to winning or losing.

The NFL turnover stats are a great example of this.

But there are so many variables built into cumulative stats such as down and distance, opponent, scheme of the opponent, and fame situation that the cumulative stats don’t hold much predictive value.

Like in this example if 57% of the patriots defensive plays occurred in a situation where they were so far ahead that they had an 80%+ chance to win those stats are very instructive on how they will play from behind or in a close game when scheme and probably personnel would be different.

It’s like using a defenses stats vs run and shoot to predict how they will do against
a 2 TE run heavy offense.

 

[Brewskies]

Here's a stat (one that is about to be broken) no QB that had the most passing yardage in a season has won a SB.

Go Brady

 

[oldrover]

The Browns played 1026 snaps of defense. On 57% of them they have less than a 20% probability to win, on 88% they had less than 40, and on a whopping 23 snaps, or 2% they have an 80% or better probability of winning.

This just in from a Browns fan:

 

[Ring 6]

Here's a stat (one that is about to be broken) no QB that had the most passing yardage in a season has won a SB.

Go Brady

Right because the 2 have no correlation at all.

 

[Deus Irae]

But there are so many variables built into cumulative stats such as down and distance, opponent, scheme of the opponent, and fame situation that the cumulative stats don’t hold much predictive value.

Like in this example if 57% of the patriots defensive plays occurred in a situation where they were so far ahead that they had an 80%+ chance to win those stats are very instructive on how they will play from behind or in a close game when scheme and probably personnel would be different.

It’s like using a defenses stats vs run and shoot to predict how they will do against
a 2 TE run heavy offense.

What you're actually arguing is that some stats (the ones you don't like) aren't accurate enough for your satisfaction. That's fine and all, and it's something of a point (essentially that, in sports, compiler a/k/a cumulative stats are often a lot less meaningful than other types of stats) but that's not new ground, and it's an essentially hollow argument. Nobody looks to total yardage in a game and says "I can't believe Miami won that game, when they were out gained 300-299!", and nobody looks at total yardage in a season as the signature stat for which team will win the Super Bowl.

But, if you want to get a decent start at figuring out which offenses/defenses were best, or worst, total yardage isn't a bad place to go. It's not perfect, you can usually find exceptions, and context is important, but it can be a good indicator.

2017 NFL Team Total Stats - National Football League - ESPN
2017 NFL Team Total Stats - National Football League - ESPN

 

[PatsFanInVa]

Per profootball reference

The Patriots played 1024 snaps on defense this year, which are all combined together to give total defensive stats and rankings.
They played 572 of them (56%) when they had above an 80% probability of winning. They played only 109 (11%) with less than a 40% probability of winning.

I'm not sure if that means their stats are overrated or underrated, because you could make an argument either way, but they are not apples to apples with other teams.

To compare, the Eagles played 38% with a 80+ probability and 18% with under 40%.

Jacksonville played 36% at 80+ and 30% at 40-.

The Bills played 20% at 80+ and 49% at 40-.

Finally for comic relief I will add the Browns.

The Browns played 1026 snaps of defense. On 57% of them they have less than a 20% probability to win, on 88% they had less than 40, and on a whopping 23 snaps, or 2% they have an 80% or better probability of winning.

I just finished NN Taleb’s “fooled by randomness” and am reading his (more famous) “the Black Swan.” So far the less famous of the two has been the better book.

He’s a skeptic, a professional trader to boot (although he’s trying to popularize skepticism much more than claiming to know, from the numbers, anything about markets. (His latest, “Antifragile,” has been widely panned.

All that to say, I agree about the limits of stats, particularly with the tiny sample size of a 16 game season.

You can’t quantify much in football. I do think, though, that it’s reasonable to conclude that the Pats’ performance over the last 16 years has been the outlier, and I do not think that it’s been by chance

PS, I think at this point that their perceived probability of winning plays a huge role. I don’t think it’s likely as high as the highest number the stats would give it, or as low as the lowest. Down 28-3, it’s much more useful to think w/2 teams, you have a 50-50 shot. Up by a huge margin, something like that thinking prevents letting teams back in the game.

 

[Chris Stevenson]

Here's a stat (one that is about to be broken) no QB that had the most passing yardage in a season has won a SB.

Go Brady

that basically means high flying teams with loaded offenses scored too fast
and gassed out their own defenses..

colts, saints, sadly the 2007-2012 pats

sucks to admit, but bb finally realized balance was more important than lots of fast scoring.

sucks that he actually thought it would work, but i am one who believes he was trying to cater to what brady wanted, and by drafting garroplo, it was a symbolic reverting back to more balance...note how we carry and care about using a fb again and a pure beast blocker like allen.

bb had that cranking in 2010, and then sort of got away from it in 2011 and 2012.

 

[maust]

Here's a stat (one that is about to be broken) no QB that had the most passing yardage in a season has won a SB.

That's more of a factoid than a stat, lol

 

[Bobsyouruncle]

that basically means high flying teams with loaded offenses scored too fast
and gassed out their own defenses..

colts, saints, sadly the 2007-2012 pats

sucks to admit, but bb finally realized balance was more important than lots of fast scoring.

sucks that he actually thought it would work, but i am one who believes he was trying to cater to what brady wanted, and by drafting garroplo, it was a symbolic reverting back to more balance...note how we carry and care about using a fb again and a pure beast blocker like allen.

bb had that cranking in 2010, and then sort of got away from it in 2011 and 2012.

It really doesn't mean that. Even if Brady won every SB since 2007 the QB with the highest yardage would not have won a SB. The idea there's been some philosophical shift since 2014 on passing doesn't show in the stats. It doesn't say much about the team offense. Last year he was a bigger part of the offense than 2013, 2014, or this year, averaging more yards per game.

In a lot of those SBs the QB with more yards on the season still won so it's certainly not a negative correlation. Brady had more yards than Wilson in 14. Manning more than Brady but beat us in 13. Eli had near identical yards as Brady in 12.

 

[sambam94]

points per game:
NE-28.6
Phil-28.6

points allowed per game
NE-18.5
Phil-18.4

 

[Nikolai]

So your 2 guesses were 35% apart, the difference between the average passing yards for a terrible Qb and close to leading the league.
See this is my point. You are using a statistical model that uses statistics against different opponents and in very different situations and trying to apply it in context of the defenses cumulative stats vs different opponents in different situation to come up with some kind of statistical guess that doesn’t account for matchups, game situation or anything unusual happening.

There is no way to account for something unusual happening. The benefit of having 18 games of data, and all the snaps that go with it, is that we can most certainly see some tendencies form. As for Bortles, I freely admit that my process is more or less in its infancy, and had I gone with the more aggressive model (which I rejected for reasons already stated), I would have been far closer to the actual output. The lesser yards took Bortles in a vacuum, and the higher output put his receivers' ability into the equation and I was wrong to dismiss that. I'm still figuring it out.

And nothing can make it ironclad, 100% correct, but I feel more confident in what this process could become than most predictive models I see...as long as I don't **** it up.

I can see it could be a fun hobby but football just isn’t built to use statistics to figure out how a game will go.
That was the point of my OP. If it’s a close game or the patriots are behind his much do their defensive stats mean when 57% of them occurred when the gane was basically on ice?

There's no way to account for it. Oddly, when it came to the total yards by each team in SBLI, I was pretty much spot on, but of course, the distribution of those yards was skewed and made the game something far different than what my projection suggested. At the end of the day, the Tom Brady and the Patriots did what they do and the Falcons did what they did. It was a violent swing in momentum, but the equilibrium was established by the end of the game. The same is true for the Jax game.

So while you can't account for specifics or twists or turns in a game, a sound analysis rooted in statistics can give a reasonable approximation of the final outcome. Of course, this kind of analysis leans on having many games from which to draw data. I'm not even sure if I could seriously apply a statistical analysis to something in Week 2 or 3. There's simply not enough data.

 

[Bobsyouruncle]

Teams with better offensive and defensive stats seem to end up in the playoffs. For instance NE and PHI were 4th and 5th in defensive scoring and 2nd and 3rd in offensive scoring.