## Asterisk

Posted by Doug on June 9, 2006

I've gotten a couple of emails in the last few weeks alerting me to a situation that I'm sure many others were already aware of. But this is the first I'd heard of it. Friday is rant day, and I think this deserves a rant. Here is one of the emails:

Why are the New York Giants of 1930 listed in first place, while the Green Bay Packers won the title that year, and have a better record?

First let me say that this is a completely appropriate email, and that nothing I write in this post should be interpreted to mean that the emailer was in any way out of line.

I get emails like this every so often. Normally they alert me to a legitimate error in my data or in my programs. I fix it, thank the emailer, and move on. But this time is a little different. The page in question is the 1930 standings page, which does indeed show the following:

W L T PF PA

New York Giants 13 4 0 308 98

Green Bay Packers 10 3 1 234 111

So I went to *Total Football* to verify. Those are indeed the correct W-L-T numbers for each team, but *Total Football* lists them in the opposite order. The problem boils down to this:

- 10/13 = .769
- 13/17 = .765
- 10.5/14 = .750

Instead of being counted as half a win and half a loss, ties were simply discarded before computing winning percentage back in those days, which is Just Plain Wrong, and that's the nicest way I know how to put it. If you disagree, then consider that your system would rank a 1-0-15 team ahead of a 15-1-0 team. That's a contrived example, but it is illustrative. Discarding ties would be appropriate if tie games conveyed no information about which team was stronger. But that's not what tie games do. They convey information that the teams were equally strong on that day, which is a very different thing.

Someone please tell me if there is something that I'm missing here. Did forfeits used to count as ties or something like that?

The NFL eventually figured out the error of its ways, because ties are now apparently counted as half a win and half a loss in computing the winning percentages. As best as I can figure it, the change occurred between the 1971 and 1972 seasons. *Total Football* lists Washington's 9-4-1 record as .692 in 1971, which indicates that the erroneous system was still in place at the time. But it lists the Eagles' 2-11-1 mark in 1972 as .179, which is in compliance with the new (correct) way of counting ties.

Correct me if I'm wrong, but my understanding is that they didn't have a postseason back in 1930. The team with the best record was simply declared the champion. It should have been the Giants, but it was the Packers. Sports fans and pundits really love their asterisks, generally too much, but this is a rare case where an asterisk is completely appropriate.

As I said, I am quite sure I am not the first person to notice this. But it was something of a shock to me. I haven't yet decided whether I should "fix" the standings at the site.

The Packers actually got the short end of the same stick in 1932 when they finished the season at 10-3-1 but were denied the opportunity to participate in the first-ever NFL playoff game, which was played by the 6-1-6 Chicago Bears and the 6-1-4 Portsmouth Spartans.

I'm guessing there are other examples of this in NFL history as well.

ut uh. more later.

phew. i've been counting ties the wrong way in bloff. went through the history, the difference between 8.5/14 and 8/13 didnt change any seedings. THANK GOD. That 1-0-15 comment created a major "ah-ha" moment.

In my travels through football databases over the years, I also noticed the change in calculation of winning percentages. But I never noticed that 1930 season before.

I've always figured that the most appropriate way to organize standings is wins minus losses.

Sorry, wasn't done yet. The 13-4 Giants (+9) would then be clearly ahead of the 10-3-1 (+7) Packers.

That won't work all the time Vince. That would make a 10-0 team inferior to a 13-2 team.

Is a 10-0 team necessarily better than a 13-2 team?

To steal a line from Doug, it Just Plain Is.

The argument you presented was fair and accurate. However the conclusion was in error. To wit--The Green Bay Packers did have the better record in light of their .769 winning percentage, versus the New York Giants winning percentage of .765. This is too much math for my little brain. Thanks again!! P.S.--The Giants would have choked in any playoff game regardless. Re--1958, 1959, 1961, 1962, 1963, et al.

Actually, I'd disagree. I'd say 13-2 is better than 10-0.

I don't think there is a right answer to the 13-2 vs. 10-0 question. But my preferred solutions are as follows:

1. have every team play an equal number of games.

If that isn't possible, then

2. At the top of the standings, rank the teams according to the probability that a true .500 team would achieve that record (or better) by random chance. In other words, is a fair coin more likely to come up heads 10 out of 10 times, or 13 out of 15 times? Answer: 13 out of 15 is about four times more likely (.004 probability vs. .001 probability). You could say we have more confidence that the 10-0 team is better than average, so they should be ranked higher.

I'm not claiming this is a better answer than any other. It's just the one that feels the most right to me.

what if the 10-0 team lost to the 13-2 team? let me know, i've got a trade brewing.

Another way to test it would be to look at all 10-0 teams and see what their record is after 15 games. If the average is 15-0 or 14-1, we can say 10-0 teams are better than 13-2 teams. If the average is 12-3 or lower, we can say 10-0 teams are worse than 13-2 teams.

Indeed. Lets see if we can track down a few survivors from the 1930 Packers and have them finish there schedule. Should they have to play another vintage roster, or should we just throw them out there verses some current NFL teams. Why debate - settle it on the field.

I think you must abide by the rules at the time. Green Bay was awarded the championship that year, per the rules in effect then. Who are we to now declare the Giants as champions?

Consider our nation's electoral college system of voting. Suppose, because of the controversies of the last few elections, that this was changed to a straight up national head count. Do we then announce that Al Gore was the winner of the 2000 election? Of course not.

And neither should history be rewritten on behalf of the 1930 Giants. They are rightly or wrongly the champions of that season.

"what if the 10-0 team lost to the 13-2 team? let me know, i've got a trade brewing."

Just in case the sarcasm is lost on anyone, any team that loses is not 10-0, unless contrary to all standards losses are shown first.

"2. At the top of the standings, rank the teams according to the probability that a true .500 team would achieve that record (or better) by random chance. In other words, is a fair coin more likely to come up heads 10 out of 10 times, or 13 out of 15 times? Answer: 13 out of 15 is about four times more likely (.004 probability vs. .001 probability). You could say we have more confidence that the 10-0 team is better than average, so they should be ranked higher."

This fits intuition nicely, a 10-0 team would be better than a 13-2 team but worse than a 50-1 team.

It also explains the obvious. a 10-0 team is better than a 9-0 team but worse than an 11-0 team