Jump to content

Bills and Steelers historically "lucky" according to FiveThirtyEight article


Big Turk

Recommended Posts

Says the Bills should have an average of 4.7 wins, meaning we have 2.3 wins more than expected based on point differential, which is the same as the Steelers who should have 5.7 wins.

 

Bills are the 9th "luckiest" team after 9 games since 1960, Steelers the 8th based on this metric...

 

That being said the Bills area also ranked 7th in their ELO ratings at 1607 and they project an 11-5 record for us...meaning apparently they feel we will keep being lucky...

 

https://fivethirtyeight.com/features/the-steelers-and-bills-have-been-historically-lucky-so-far-the-chargers-have-not/

Edited by matter2003
  • Like (+1) 1
Link to comment
Share on other sites

Yeah I saw this yesterday. 
 

It seems to me after looking at the information quickly that teams with winning records are in that “lucky” bin and teams with losing records are in that “unlucky” bin. There is very minimal crossover for a losing team winning more than they should and a winning team losing more than they should and in all of those instances it’s pretty close to being 0 difference. 
 

Which I find kind of odd... but it also supports the concept that teams in the NFL are all close in ability/talent and wins are hard to come by in the NFL... or something like that. 
 

 

Edited by JGMcD2
  • Like (+1) 2
Link to comment
Share on other sites

Just now, Buffalo_Stampede said:

Ours is actually flawed because we've dominated basically every game we won but let teams back in the game. 

 

I agree with that to some degree and also it is highly skewed by one blowout loss to the Titans.  

  • Like (+1) 4
Link to comment
Share on other sites

27 minutes ago, matter2003 said:

That being said the Bills area also ranked 7th in their ELO ratings at 1607 and they project an 11-5 record for us...meaning apparently they feel we will keep being lucky...

Projecting 11-5 means 4-3 the rest of the way, which is in line with what our point differential says our record should be. So they are projecting regression.

Link to comment
Share on other sites

The concept of Pythagorean Wins would be useful over a very large sample size where game-script permutations are smoothed out but over 9 games it's pretty much useless.  In the Seattle game, for example, the game was 99.9% decided when we went up 41-20 but the Seahawks got a fluke TD and a garbage time TD that made the score look closer but probably did almost nothing to change the win probability at the time they were scored.  So the notion of that win being valuated as less of a win because the margin was 10 instead of 21 (or on the other side the KC loss looking closer than it should) is garbage.

 

EDIT:  It's absolutely true that if you ran a simulation of these 9 games say...1,000 times...our 7-2 record would be in the higher end of the range.  But saying that the AVERAGE result of those simulations would be a 4.7-4.3 record is asinine to anyone who's watched the games.

Edited by SageAgainstTheMachine
  • Like (+1) 2
  • Thank you (+1) 1
Link to comment
Share on other sites

1 hour ago, Buffalo_Stampede said:

Ours is actually flawed because we've dominated basically every game we won but let teams back in the game. 

No doubt luck IS involved, as the difference between a W or L can be a few plays, ref calls (or non-calls), lucky bounce, whatever.

 

But as Buffalo_Stampede wrote, it doesn't apply here. Ridiculous to base "luck" on close scores.

Link to comment
Share on other sites

5 minutes ago, uticaclub said:

What are these stats based on? Just random numbers thrown together to make random graphs doesnt move the needle for me

 

It's a barely more complex statistic than W/L record itself and the thing it pretends to illustrate (luck) isn't what it illustrates at all.

 

If I wanted to start an analysis of luck I'd look at truly high variance things like fumble recovery percentage, tip drills and wind conditions for FG kicking.

Edited by SageAgainstTheMachine
  • Like (+1) 2
Link to comment
Share on other sites

×
×
  • Create New...