Norris trophy (99.8% complete historical results, 1954-2023)

[Hockey Outsider]

There's a potential issue with how I've been presenting the Norris trophy shares (see post #4). Recent defensemen seem to be ranked too high, and but I couldn't figure out why. For example, Chara is ranked between Chelios and Coffey. Karlsson ranks about Potvin, while Hedman and Doughty rank above Robinson and Park. None of that feels right.

The reason is, up until 1995, there were only three spots on the awards ballot. The total number of votes added up to 1.8x. (Voters were able to award 5 points to first place, 3 points to second place, and 1 point to third place. 5 + 3 + 1 = 9, and 9/5 = 1.8). From 1996 onwards, the total number of votes add up to 2.6x. Thus, there are significantly more votes to go around from 1996 onwards. I think this explains why some recent players appear to be overrated on my list.

I've scaled the data so that each year is worth the same amount. (In other words, a player's voters get scaled back by 44% after 1996 - 44% is 2.6 divided by 1.8). The main counter-argument is a unanimous win from 1995 and earlier would be worth 0.56 vote shares (1 / 1.8), and a unanimous win from 1996 and later would be worth 0.38 vote shares (1 / 2.6). So, for a player from 1996-present, a win would be worth less under this system. But, because the ballot has room for three rather than five players, that player would be able to get votes much more frequently over the course of his career.

So, for discussion purposes only, I'll present the career Norris shares under the new method:

Ray Bourque	4.72
Bobby Orr	4.23
Doug Harvey	3.48
Nicklas Lidstrom	3.45
Paul Coffey	2.16
Denis Potvin	2.02
Chris Chelios	1.99
Pierre Pilote	1.92
Al MacInnis	1.59
Zdeno Chara	1.54
Larry Robinson	1.54
Brad Park	1.48
Erik Karlsson	1.41
Brian Leetch	1.28
Bill Gadsby	1.27
Victor Hedman	1.19
Scott Stevens	1.12
Chris Pronger	1.11
Drew Doughty	1.11
Borje Salming	1.07
Red Kelly	1.07
Rod Langway	1.04
Tim Horton	1.01
Mark Howe	0.96
Shea Weber	0.95

Intuitively, these results seem to make more sense. Chara, although still an all-time great, is now clearly below Coffey and Chelios. Karlsson still looks great - but he's well below Potvin. Intuitively, this approach seems to make more sense.

 

[Hockey Outsider]

Here's an attempt to prove that the new approach makes more sense. Here's a summary of the top 25 defensemen under the new approach, where they ranked under the old approach, and how they ranked in the Top 100/200 projects:

Player	New	Old	Change	Top 100
Ray Bourque	1​	1​	0​	3​
Bobby Orr	2​	3​	1​	1​
Doug Harvey	3​	4​	1​	2​
Nicklas Lidstrom	4​	2​	-2​	4​
Paul Coffey	5​	7​	2​	10​
Denis Potvin	6​	9​	3​	6​
Chris Chelios	7​	5​	-2​	8​
Pierre Pilote	8​	10​	2​	11​
Al MacInnis	9​	11​	2​	15​
Zdeno Chara	10​	6​	-4​	17​
Larry Robinson	11​	16​	5​	7​
Brad Park	12​	17​	5​	9​
Erik Karlsson	13​	8​	-5​	21​
Brian Leetch	14​	15​	1​	20​
Bill Gadsby	15​	21​	6​	18​
Victor Hedman	16​	12​	-4​	24​
Scott Stevens	17​	20​	3​	13​
Chris Pronger	18​	13​	-5​	12​
Drew Doughty	19​	14​	-5​	22​
Borje Salming	20​	25​	5​	16​
Red Kelly	21​	26​	5​	5​
Rod Langway	22​	27​	5​	23​
Tim Horton	23​	28​	5​	14​
Mark Howe	24​	31​	7​	19​
Shea Weber	25​	18​	-7​	25​

The biggest gainers are defensemen who spent all of their career before 1996, and who had many years where they contended for the Norris trophy, with relatively few wins (ie Park, Howe, Gadsby). This is exactly what we'd expect. All those 2nd place finishes for Park (as an example) were worth less under the old system, since there were 44% fewer votes to go around.

The correlation between the new approach and the HOH ratings is 0.71. The correlation between the old approach and the HOH ratings is 0.47. The higher the number, the stronger correlation. This suggests that the new method is much more strongly correlated to the HOH consensus. (Part of that is due to Karlsson and Hedman adding to their legacies since early 2021, when the Top 200 list was made. Even if we excluded them altogether, the correlation is still much higher under the new system - 0.74 to 0.57).

This is pretty good evidence that the new method that I've presented makes more sense.

 

[Hockey Outsider]

A final comment on this topic:

It also raises an interesting question - which defensemen are ranked significantly higher or lower than expected, when we compare the Norris voting results to the HOH top players lists? Obviously, there isn't going to be a perfect correlation. A player needs to be healthy to accumulate a high total - thus Bobby Orr ranks 2nd in career vote shares, despite obviously being the greatest defenseman ever. And playoffs are excluded from this analysis entirely. Still, there should be some correlation.

Biggest gainers (HOH list vs Norris trophy shares)

• Red Kelly (5th vs 21st) - this one is easily explainable. The Norris trophy was only introduced in 1954, midway through his prime. Based on the all-star data, Kelly likely would have won the Norris trophy decisively in 1951, 1952, and 1953. HOH takes this into account, but the raw data doesn't. (Also, not that this changes things much, but Kelly would probably be ranked 6th rather than 5th if we're only looking at his time as a defenseman - the HOH list takes into account his eight years as a centre in Toronto, and that likely moves him above Potvin in the overall ranking).
• Tim Horton (14th vs 23rd) - a big gap, but Horton was an excellent playoff performer (which isn't taken into account by the Norris trophy data), and as a primarily defensive defensemen, he's the type of player that voters tend to underrate (it's easier to just pick the player with the flashy point totals).
• Chris Pronger (12th vs 18th) - two of the reasons I mentioned for Horton apply here - Pronger was a stellar playoff performer, and he was primarily a defensive defenseman (so it was easier for voters to overlook him). He also struggled with injuries throughout his prime (he missed 15 games in 1999, 30+ games in 2001, virtually the entire 2003 season, and 16 games in 2007 - he would have won the trophy had he been healthy). Actual accomplishments are paramount, but the HOH lists consider the context as well.

Biggest losers (HOH list vs Norris trophy shares)

• Erik Karlsson (21st vs 13th) - primarily explained because the HOH list is from early 2021. Since then, Karlsson added a third Norris trophy (and was a Pearson trophy finalist). I'd also argue that Karlsson's playoff resume is weaker than most players on this list (he was specular in 2017, but has played in less than 70 games) - so most of the others get a boost relative to him. He's also all-offense, which is a bit easier for the Norris voters to recognize (and the HOH list implicitly tries to correct for this).
• Victor Hedman (24th vs 16th) - this one is purely due to timing. Since the list was completed, Hedman added two more years as a Norris trophy finalist, and was the #1 defenseman on a Stanley Cup winner and a Stanley Cup finalist. He'd surely rank higher today.
• Zdeno Chara (17th vs 10th) - Chara, like Pronger and Horton, is primarily a defensive defenseman. So, if anything, we'd expect the Norris voters would have underrated him. The best explanation I can come up with is he peaked during a relatively weak era for defensemen. (He was a finalist six times in the span of ten seasons between 2004 and 2014). Yes there was stiff competition at the top from Lidstrom, Pronger and a few others - but a lot of weak defensemen (in an all-time sense) placed in the top five - Phaneuf, McCabe, Aucoin, Boyle, Campbell, Visnovsky, Yandle, Green, Beauchemin, Redden, Timonen (not to mention borderline HOF players like Zubov, Gonchar, Pietrangelo, Letang, Subban, etc). That probably resulted in Chara sneaking towards the top of the race more than he should have, had he debuted a decade earlier (or later).

There shouldn't be a perfect correlation (otherwise our list is doing nothing more than parroting the data). But this tells me that the HOH lists have a solid factual basis, and the biggest discrepancies have reasonable explanations.

 

[jigglysquishy]

Bobby Orr ranks 2nd in career vote shares, despite obviously being the greatest defenseman ever.

I'd also add, the inherent nature of the system means high peak seasons get undervalued. You can't get more than 100% of votes.

So each of the 6 years of Bobby Orr getting over 93% of possible votes isn't weighed heavier than a standard dominant Norris. Bobby Orr could have been even better in 1970 than he was, but he literally could not get more votes.

I'd also argue Orr has 3 or 4 years he should be been the unanimous winner. But there were always a few clever votes outside 1970.

In general, the new method makes much more sense.

 

[seventieslord]

I was surprised to not see Duncan Keith in that list.

 

[Hockey Outsider]

I'd also add, the inherent nature of the system means high peak seasons get undervalued. You can't get more than 100% of votes.

So each of the 6 years of Bobby Orr getting over 93% of possible votes isn't weighed heavier than a standard dominant Norris. Bobby Orr could have been even better in 1970 than he was, but he literally could not get more votes.

I'd also argue Orr has 3 or 4 years he should be been the unanimous winner. But there were always a few clever votes outside 1970.

In general, the new method makes much more sense.

I agree with this pretty much word for word.

I've toyed around with a few ideas for how to reward dominant performances. But I've never come up with anything satisfactory. Any approach that I've tried has been completely subjective. I'd rather keep the system as objective as I can and talk about the context, rather than fudge the numbers.

 

[Hockey Outsider]

I was surprised to not see Duncan Keith in that list.

The "issue" with Keith is, although he's won two Norris trophies, he was never again a finalist. That's unusual. 13 players have won 2+ Norris trophies, and every single one of them has been a finalist at least once in a year where they didn't win (and usually a lot more than once). The voters have only put Keith in the top five three times (two wins and a 4th place finish). Yes, he does have several 6th and 7th place finishes, but there aren't that many votes to go around that low on the ballot.

Part of that goes back to my previous comment about voters usually favouring defensemen with big numbers - thus Phil Housley and Sergei Gonchar look better than we'd expect, and Serge Savard looks worse. I just checked but Keith has only topped 45 points three times in his entire career. If he stopped being such a responsible, disciplined blueliner, I'm sure he could have scored 50+ points consistently. That probably would have helped him for the Norris trophy - but it surely wouldn't have helped the Blackhawks win games.

I agree Keith should rank higher, but a big part of his legacy is his playoff performances. Surely he's one of the top ten playoff performers of the post-lockout era (which, of course, isn't taken into account with the Norris trophy).