Thursday, January 29, 2009

Has fighting really been dropping since the 90s? The numbers.

In my last post, I made reference to my assumption that while fighting may have increased this season, it's still down significantly from 15 years ago. I linked to some data from hockeyfights.com that went back as far 2001, then said this:

While I can't find any stats on fighting frequency in 1993, I feel pretty safe in saying it was significantly higher than even this season's numbers. If anyone can point to the numbers and they prove me wrong, I'll be glad to eat some crow.
An anonymous commenter, who I'm just going to go ahead and assume is Howard Berger, called me on it:

It isn't good enough for me for you to basically suggest, "I think I'm right, but if you prove me wrong, by all means. Until then, this argument stands." It's the kind of technique you'd call Damien on and it undermines your argument, in my opinion.
After permanently IP-banning him, I got to thinking about his comment. I decided to dig a little deeper and ended up finding what I think are some reasonable ballpark numbers going back to 1991.

So, once again courtesy hockeyfights.com, here are some estimates. I took their yearly fight log for each season, and counted the number of fights listed. I call this an estimate, since I'm not sure how complete their records are (although my guess is they're pretty complete) and there may be some altercations included where both guys didn't get majors. Still, when it comes to fight data online hockeyfights.com is pretty much the undisputed champ, so I feel pretty good about these numbers.

(Update: the owner of hockeyfights.com confirmed that they're "very confident" that the data is complete, with the exception of the occasional tweak.)

Here's what the numbers show. All are for regular season only.

1990-91 - 772 fights
1991-92 - 792 fights
1992-93 - 642 fights
1993-94 - 835 fights
1994-95 - 506 fights (short-season due to lockout)
1995-96 - 785 fights
1996-97 - 907 fights
1997-98 - 838 fights
1998-99 - 660 fights
1999-00 - 573 fights
2000-01 - 684 fights
2001-02 - 803 fights
2002-03 - 668 fights
2003-04 - 789 fights
2004-05 - No season
2005-06 - 466 fights
2006-07 - 497 fights
2007-08 - 664 fights
2008-09 - 784 fights (projected)

In graph form (minus the lockout years):


So was my assumption about fighting dropping over the past 15+ years correct? Mostly, with a few caveats.

There's clearly been a downward trend since the early 90s. The drop doesn't actually seem to start in earnest until the late 90s, which is later than I had assumed. There's a big drop in 1998, a climb back up, and then fighting falls off a cliff after the lockout.

I made the point that even if fighting reaches its projected level for this season (which is unlikely since fighting usually declines late in the season), it would still fall far below the levels seen in the 90s. That turns out to be true, although perhaps not as significantly as I would have expected.

Beyond that, my reference to 1993 actually turned out to be wrong, at least as far as the 1992-93 season goes. That year saw fewer fights than even last season, which seems surprising. It was also an out lier, far below the average of other years in the early 90s.

Why? What happened in 1992-93? Answer: The instigator rule. The rule came into effect that year, and initially caused a significant reduction in fighting. Those who were fans back then will remember how haphazardly the rule was applied in that first season, with referees trying to call it on virtually every fight. Sometimes it seemed that if your gloves hit the ice a fraction of a second early, you got an extra two.

The rule helped keep fighting low that season, but the effect was temporary. After some tweaks to the rule and how it was called, fighting went back to previous levels.

One additional point to consider: While this list includes total fights league-wide, the NHL has been adding teams since the early 90s. There were only 21 teams in 1999-91, compared to 30 now. More teams means more games, so all else being equal we'd expect the numbers to be trending upwards.

So for example, this year's projected 784 fights would be an average of 0.64 fights per game. That would actually equal the 0.64 average of 1992-93 (when there were 24 teams, and the league played 84 games), and would be far below the per-game averages in other years from the early 90s despite the grand total being comparable.

The verdict: My general claim of fighting levels dropping over the past two decades was right. My specific claim about 1993 turns out to be wrong (it's even, but not higher) because I had the misfortune of picking a year that turned out to be an aberration due to the instigator.

I suppose that means that I do have to eat a little bit of crow. But I think the overall point stands. Fighting has been trending downwards for a long time, and the numbers back it up.




9 comments:

  1. Great post, but I've heard that there are these things they have nowadays, on computers even, where you can actually visualize the data so it can be grasped more easily, while also allowing for a line indicating any sort of trending that may be going on. This all might be a hoax however.

    ReplyDelete
  2. I wonder if instead of the "fights per game" stat, if you went with "fights per player" instead, if it makes any difference?

    With 6 more complete rosters now than there were in the early 90's, and what appears to be somewhat fewer fights than those days, on a per-capita basis, players are fighting way less. No?

    ReplyDelete
  3. Congrats, you've just done more in-depth research in the last 24 hours than the lead hockey writer for the biggest newspaper in Toronto has done all year.

    Interesting stats to say the least, I would have thought fights would have been significantly higher in the early 90's, but your point about fewer teams is well taken.

    I wonder about these stats though just based on PIMs. Look at the league leaders from 07/08 vs. 92/93, two seasons with apparently roughly the same fights per game. It doesn't even compare. Carcillo went insane and led the league by over 100 PIMs, but he would have only been 5th in 92/93. The 2nd most penalized player in 07/08 would be 60 PIMs short of even cracking the top 10 in 92/93.

    How were people getting THAT many PIMs in 92/93 without getting into fights? Does the 10 min instigator account for it? Is it possible that hockeyfights.com just doesn't have complete records for the pre-internet era?

    ReplyDelete
  4. How were people getting THAT many PIMs in 92/93 without getting into fights?

    Interesting question. The instigator misconducts probably played a role. But it's also certainly true that individual enforcers fought more often back then than now.

    You can see this at hockeyfights.com:
    http://www.hockeyfights.com/leaders/players/1/reg2008

    Change the year and you can see the leaders going back to 1991.

    It was common for guys to fight 30+ times back then, and unheard of now. In fact, prior to last year there were two straight seasons post-lockout where nobody fought 20+ times. In 1992, 11 guys had 20+.

    ReplyDelete
  5. Great post, but I've heard that there are these things they have nowadays, on computers even, where you can actually visualize the data so it can be grasped more easily, while also allowing for a line indicating any sort of trending that may be going on.

    Added. Don't ever say I don't look out for you.

    ReplyDelete
  6. Awesome, much appreciated. And sorry for the sarcasm. Now how about one indicating average fights per game? Just kidding. Sort of.

    Keep up the great work.

    ReplyDelete
  7. I agree the trend has been down though it would be better to do an analysis based on fights per game to account for shortened seasons or few teams. Also I would toss out of the analysis the two years after the lockout. The crackdown on obstruction created a lot of 5 on 4 and 5 on 3 situations which kept enforcers on the bench more. I like everyone else has done a post on fights. I found, not surprisingly, that 20% of players produce 85% of fighting majors.

    ReplyDelete
  8. Fascinating stuff.

    If fight totals go up again next year (a big "if" when one looks at 20 years of data) it will mark the first time in 20 years that fighting has increased in four consecutive seasons.

    Not sure if that says anything, just thought it was an interesting pattern amongst the very up and down data.

    ReplyDelete
  9. Good post, but you should tried this site, http://dropyourgloves.com/Top.aspx, is better.
    Here are the fights per game statistic:
    1990-91 - 1.04
    1991-92 - 1.02
    1992-93 - 0.71
    1993-94 - 0.85
    1994-95 - 0.88
    1995-96 - 0.80
    1996-97 - 0.93
    1997-98 - 0.87
    1998-99 - 0.66
    1999-00 - 0.57
    2000-01 - 0.61
    2001-02 - 0.70
    2002-03 - 0.56
    2003-04 - 0.70
    2004-05 - No season
    2005-06 - 0.41
    2006-07 - 0.44
    2007-08 - 0.59
    2008-09 - 0.70 (projected)

    ReplyDelete