≡ Menu

The early AFL was unstable, which probably isn’t too surprising. This was most clear in 1963, which must be one of least sticky league seasons in pro sports.  The Oakland Raiders, under new coach Al Davis, jumped from 1-13 to 10-4. The San Diego Chargers, who added quarterback Tobin Rote and had a breakout season from second-year wide receiver Lance Alworth, went from 4-10 to 11-3 and league champions.

Meanwhile, three of the eight teams in the AFL had huge declines.

  • The Houston Oilers had been the class of the early AFL, winning the title in ’60 and ’61, before falling in the championship game in ’62. But in ’63, Houston dropped from 11-3 to 6-8.
  • Houston lost in the ’62 AFL title game to Lamar Hunt’s Dallas Texans, who went 11-3 behind coach Hank Stram and quarterback Len Dawson. But  after moving to Kansas City in the off-season, the team went just 5-7-2.
  • Denver went 7-7 in 1962,  with a pass offense and a pass defense that was both about average.  But in ’63, both ranked last in the league, and the Broncos fell to 2-11-1.

The graph below shows each team’s winning percentage in 1962 (on the Y-Axis) and in 1963 (the X-Axis), along with a trend line.  In a stable lead, the teams would form a diagonal line that starts on the bottom left and goes up to the bottom right; i.e., as win percentage rises in Year N, it rises it Year N+1. Yet here, the trend line is the exact opposite.  That’s because there was a strong negative correlation (-0.49) between winning percentage in the two years.


So yeah, the AFL was really unstable in 1963. This isn’t just a function of it being a small league: by 1969, the AFL was extremely stable, as we’ll see in a minute. The impetus for this experiment? I was curious if the NFL is more or less sticky now than it used to be. In other words, is team performance more volatile or less volatile now than it used to be?

There is no perfect way to measure this, and there are some things I haven’t quite figured out how to handle.1 That said, here is what I did:

1) Calculate each team’s winning percentage in Year N, and pro-rate to 16 games to get wins per 16 games in Year N.

2) Calculate each team’s winning percentage in Year N-1, and pro-rate to 16 games to get wins per 16 games in Year N-1.

3) Square the difference, to give more weight to extremes.

4) Sum that result for all teams in that league for that season (Year N season).

5) Take the square root of that result, and then divide by the number of teams.

I have graphed the results below, which shows that the average team changes by about 3-4 wins per year. The AFL is in orange, while the NFL is blue.


In general, there does appear to be more volatility now. That would make it more impressive for some team like the Patriots to keep being dominant year after year than it would be for a team to do so in the ’70s. That’s probably not a surprising result, but it’s at least good to have some evidence of it.

  1. One thing I have been thinking about: If there is more parity now (not sure if that’s accurate), that means teams are closer to .500 to begin with. We know about general regression to the mean, so if there are fewer teams with extreme records, that would negate some natural volatility in team records (for example, if there used to be a ton of 10-2 teams, we know we would project them closer to say, 8-4 the next year; the same goes for an 11-5 team moving back to the pack, but that “move” would look less extreme. []
  • Ryan

    Is another way to measure volatility over time simply to repeat the correlation exercise done for ’62-63 for all years since? Perhaps you could add a trailing 5-year average to visually spot trends in this correlation.

    • I’m not sure I am totally following. What are you thinking?

      • Ryan

        Basically the chart Tom did but using winning percentage instead of SRS.

        • Gotcha. Tom’s chart is pretty cool — not sure if I can recreate that too easily.

  • Tom

    Chase – this is pretty cool, I’ve got some SRS data to round this out. There does seem to be a trend towards more instability, as the correlation coefficient between teams’ SRS rating from year to year +1 is getting lower. As both of our charts show (if I’m reading yours right), 1998-1999 really stands out:

    The Colts went from 3-13 to 13-3
    Rams went from 4-12 to 13-3
    Niners went from 12-4 to 4-12
    Atlanta went from 14-2 to 5-11

    And I do think that it makes what the Patriots are doing that much more impressive…just taking a cursory look, the 1970’s Steelers had a core group of great players, on offense and defense, from about 1972-1980. Not sure the Patriots have had anything like that, besides, of course Brady and Belichick. https://uploads.disquscdn.com/images/1efc0a0abceb43d52899068c1906cd7db21024e432088cccd6cb9d3ccc2f28a8.png

    • Yep. Also 1998 was kind of a weird year with 4 ancient QBs going to the title games, and none of them were good the next year (Elway retired, Testaverde got hurt, Chandler got hurt, Cunningham got benched; and Young might have been the best QB that year and he got hurt). So I think having 5 old QBs fall off played a big role, and the 49ers/Falcons were in your example already.

      Someone else made a good point in the comments that staying good was easier back then, but getting good was hard. That’s true, and one could make the argument that if the NFL was more top-heavy, it was even harder to win SBs? I don’t know if I follow — after all, 1 team has to win every year so I don’t know if the concept of harder makes sense — but in the AFC you had to go through some stacked teams, whereas maybe you don’t have to do that now?

      • Tom

        Well, the only thing I can think of, is that before free-agency, guys had longer contracts, and so they had to stay with teams longer? I don’t know enough about this side of football – signings, free agency, length of contracts, money, etc. – so I’m not sure how/if that would affect whether or not it’s harder to “get good” or “stay good”.

      • Adam

        I think it’s easier to win one SB in the cap era, but easier to win multiple SB’s in the pre-cap era.

        • Richie

          I think I agree with that theory, but I was trying to find a way to quantify it.

          If we say the cap era began in 1994, there have been essentially 16 different teams win the Super Bowl in 22 years. I am considering the same “team” if there is significant continuity from the last time a team won. For instance, the 1998 Broncos and 2015 Broncos count as different teams, because they are nothing alike. However, all of the Patriots winners are the same, because they all had Belichick-Brady and there was never any kind of rebuild between their wins.

          In the 22 years prior to 1994, there were between 12 and 14 different winners. I can’t figure out how to classify Washington’s 3 wins. Same coach for each, but different QB. Different RB’s. But the 82 and 87 teams shared 9 starters. The 87 and 91 teams only shared 7 starters.

  • Josh Sanford

    In response to Adam, below, about it being easier to win one, but harder to win two or more in a row: logically that MUST be true, since in the year that the winning team from the year before is trying to repeat, they have less of a grip on it, and it is more of a toss up for everyone else.

    To everyone else: when I was reading the body of the article, I was thinking “I can’t wait to get to the comments to see what everyone says about free agency.” Boy, was I wrong. No one has said anything about free agency?

    • Richie

      ” since in the year that the winning team from the year before is trying
      to repeat, they have less of a grip on it, and it is more of a toss up
      for everyone else.”

      What does that mean?

      • Josh Sanford

        For instance: 1975 = 20% chance of repeating. 2015 = 6% chance of repeating. In 1975, there’s 80% of the opportunity for winning the SB divided among rest of the whole league. In 2015, the rest of the league has a share of a 94% chance of winning. Just for instance

        • Richie

          Why does a team in 1975 have a 20% chance of repeating?

          • Josh Sanford

            I am not saying that the underlying theory is correct, I am simply pointing out that if it is true, then the corollary must also be true.