FiveThirtyEight Elo rating system for best teams of all-time

[74_75_78_79_]

Any of you hear of this? Quite interesting list to say the least although quite a few of the real top ones belong. But '04 Pats #2?? '05 Steelers not only one slot above '78 Steelers, but '78 Steelers at...#35??

https://web.archive.org/web/20150930054 ... ng-to-elo/

[7DnBrnc53]

Any of you hear of this? Quite interesting list to say the least although quite a few of the real top ones belong. But '04 Pats #2?? '05 Steelers not only one slot above '78 Steelers, but '78 Steelers at...#35??

https://web.archive.org/web/20150930054 ... ng-to-elo/

Also, they have the 97 Broncos below the 97 Packers. WTH?

[Rupert Patrick]

1923 Canton Bulldogs at number 27 is very interesting. I don't buy it, because as I understand 1920's football (and I'm sure I will be corrected if I am wrong) it was not unlike late 18th century major league Baseball in that the teams and the leagues were fluid as players often switched teams as the schedule deemed necessary, for example guys would play for a different team under pseudonyms for a game for a check if his team was off. In this way, the strength of one team relative to others would be nearly impossible to determine because teams did not have consistent rosters from week to week. Also, there was a lot of barnstorming, and playing against "franchises" who only played a couple games.

[74_75_78_79_]

Any of you hear of this? Quite interesting list to say the least although quite a few of the real top ones belong. But '04 Pats #2?? '05 Steelers not only one slot above '78 Steelers, but '78 Steelers at...#35??

https://web.archive.org/web/20150930054 ... ng-to-elo/

Also, they have the 97 Broncos below the 97 Packers. WTH?

Yeah, and how about '83 Redskins #10, '83 Raiders # 80?

'90 Bills not only #113, but behind the '91 installment who's at #101.

#168 - '92 Eagles over...
#170 - '80 EAGLES?
#196 - '79 CHARGERS?
#197 - '68 COWBOYS?
#260 - '80 Steelers over...
#264 - '80 CHARGERS?
#267 - '81 NINERS? (disrespected once again)
#499 - '80 FALCONS??
#523 - '76 PATS???

Senseless.

[Rupert Patrick]

I think it is generally pretty good in rating the top ten teams, it does have issues. One particular problem is with the 1946 Cleveland Browns. It ranks the Browns (ELO Blend 1478) between the 1946 Cardinals (1501), who finished 6-5, and the 1946 Steelers (1439), who finished 5-5-1. One might infer from that ranking that the 1946 Browns relative to the 1946 NFL were about a .500 or so team. Three of the 1946 AAFC teams (Chicago, Buffalo, Miami) finished in the bottom two percent of teams from all time. The 1946 Miami Seahawks were the lowest rated team of all time.

[NWebster]

The rating is just math, it was invented 60 or so years ago to rank chess players and can be applied to any closed system. Essentially everyone starts at the same rating and then gains or loses points for winning or losing, fairly obvious. Over time winning or losing over higher rated opponents gains you more or loses you fewer points, again, pretty straight forward. For lower rated opponents a win gains you less and a loss loses you more, so far so good.

Where it has trouble with football, my opinion, is actually where lots of rating "systems" do, there are just too few games and too few different opponents played to not have randomness sink in and to be well connected to the quality of your opponents. How many times have we seen 7-1 or 6-2 teams and said, we'll they haven't played anyone real yet?

High level chess is more like baseball or basketball with multi game series so you have many more observations and randomness is less likely a factor. Clearly the 78 Steelers are the better team but Big Ben catches Dwane Harper's foot, what was that, one in 100, one in 1000, I don't know but the 78 team sure would win a 7 game series.

[NWebster]

Any of you hear of this? Quite interesting list to say the least although quite a few of the real top ones belong. But '04 Pats #2?? '05 Steelers not only one slot above '78 Steelers, but '78 Steelers at...#35??

https://web.archive.org/web/20150930054 ... ng-to-elo/

I think #2 is high as well, but the math sees a team with a 21 game winning streak that was only snapped by a team headed to a 15-1 regular season.

[Jeremy Crowhurst]

The rating is just math, it was invented 60 or so years ago to rank chess players and can be applied to any closed system. Essentially everyone starts at the same rating and then gains or loses points for winning or losing, fairly obvious. Over time winning or losing over higher rated opponents gains you more or loses you fewer points, again, pretty straight forward. For lower rated opponents a win gains you less and a loss loses you more, so far so good.

Where it has trouble with football, my opinion, is actually where lots of rating "systems" do, there are just too few games and too few different opponents played to not have randomness sink in and to be well connected to the quality of your opponents. How many times have we seen 7-1 or 6-2 teams and said, we'll they haven't played anyone real yet?

High level chess is more like baseball or basketball with multi game series so you have many more observations and randomness is less likely a factor. Clearly the 78 Steelers are the better team but Big Ben catches Dwane Harper's foot, what was that, one in 100, one in 1000, I don't know but the 78 team sure would win a 7 game series.

The big problem as I see it is that ELO ratings are, in part, quantifying luck, and in football luck plays a much larger role in determining the outcome than in chess.

The small sample size obviously is huge -- in chess, back when I was an active player at least, the proper formula for calculating ratings didn't even kick in until you had played 25 games. In the NFL, there are years when a team doesn't have the same lineup for more than 2 games in a season.

But it's a fun exercise.