Thursday, September 18, 2008


Hitting Analysis of 2007-2008 Part 2

If you'd like to read updates about training camp, Tom Gulitti is your man to go check out. He is, as usual, on point and he has all you need to know and more about the scrimmages, who will line up at center between Rolston and Elias on the first line (at first), the free AHL preseason game at the Rock, and that Brian Rolston has a no-trade clause. Hopefully, should Rolston play as well as he can play, a trade will not be necessary regardless of the clause. Anyway, go to Fire & Ice and just keep on scrolling down. Likewise, check out Tibbs at Beast of the East as he aggregates all the Devils articles out there himself.

However, I want to talk about hits. I made an attempt at a hitting analysis from last season similar to the shooting accuracy analysis I did a few months ago. I put off the follow-up to Part 1 because I wasn't quite confident in the accuracy of the hit results I got game to game from the NHL's game reports (pulled from I noticed that some games somehow had way more hits than others. For example, the Devils' seventy-seventh game of the season was against the Rangers at MSG and a total of 104 hits were recorded. In comparison, the Devils' next game was against Philadelphia at the Rock, and despite how those games get, only 21 total hits were logged. I found it hard to believe that there would be that big of a difference in hits recorded, especially against two hated rivals. Different scorers at different games can truly yield extremely high variability. I can understand why Hockey Analysis attempted to adjust the real time super statistic of hitting (among others).

So basically, you may want to take all of this with a grain of salt since I have to assume NHL's reports are correct. I don't know how well the adjustment works nor how it would work on a game to game basis. That said, I think it's an interesting exercise all the same.

Anyway, I forge on because what I am looking at is how do the Devils do when it comes to hits overall? Do they win more games when they out hit their opponent, or does it seem like it's not that significant? Do they seem to hit more on the road? Let's get a basic overview, first:

Overall Hit Stats 07-08

Now, that is something right there. From what I've compiled from NHL's reports, the Devils outhit their opponents more often than not in all games and when they do win. More impressively, is that they do so while at The Rock to the tune of approximately 75.61% of all 41 home games. Based on this empirical evidence alone, and with our assumptions, we can conclude that the Devils are definitely not a soft team as they pile up the hits, especially at home. As an aside, the Devils actually outhit their opponents in 21 of the 36 Devils losses (or non-wins to be precise), which is good for a rate of approximately 58.33%.

Again, using our assumptions we find out that this makes sense when we breakdown the average hits per game by the Devils and their opponents. In this case, I determined the averages between Devils wins and losses, Devils at the Rock and on the road, and games where the Devils did and did not outhit their opponents. I expected the Devils to have a higher average of hits than their opponents when they outhit their opponents and especially at home.

Breakdown of Hit Averages

From what we see here, my expectations were only half right I took a percent difference to highlight how significant the differences in averages were. I was correct in expecting that the difference in average hits by both sides would favor New Jersey at home. Interestingly, when the Devils win or lose, the difference in averages just under a hit in favor of New Jersey. Given that the Devils outhit their opponents in about 58% of their wins and 58% of their losses, this is a reasonable result. What is truly stark is the difference in average hits by the Devils in games at home and on the road. It's truly massive and given that both sample sizes are the same, 41 games each, it perhaps could be said that the Devils take up their physical game when away from home. Unfortunately for them, so do their opponents, it seems. In addition, check out the averages when the Devils outhit and get outhit. The Devils actually average more hits in the games where they get outhit by their opponents. Again, assuming our assumptions are valid, that's a testament to how tough the Devils are - they don't roll over, it's just that their opponents were more vigorous in taking the body.

Based on the chart, the significant differences appear among our empirical data when the Devils are at home (a difference in averages of 2.756 hits), when the Devils are outhitting their opponents, and when the Devils are being out hit. In those last two cases, the differences in averages is roughly around 4.5 hits (up to 4.667 in the former, down to 4.412 in the latter).

However, given that the Devils average 1.51 more hits when they are being outhit, this brings me back to the issue of the assumption made earlier. I cannot rule out scorer bias when it comes to hits being recorded from game to game. I gave an example of this earlier, but the previous chart provides further suggestion. While the Devils aren't certainly soft, the fact that the Devils' average so many more hits on the road is questionable. From what I've seen of the team, they don't seem to be that more physical on the road than at home and while I'm sure there are games where they have been, I certainly have some doubts that it would translate over the whole season. Likewise, I don't think traveling to the Rock incapacitates opponents from throwing the body by half of what they do at home. Granted, the opponents are teams of varying physical ability, effort on a given night, etc. However, overall, that their average based on NHL's numbers doubles up when they go home definitely raises an eyebrow. Combine that with the earlier example and I cannot come up with a definitive solution.

If we assume there isn't anything wrong with NHL's numbers, we can conclude from the empirical evidence that the Devils were certainly not soft last season. They got more hits than their opponents at home, they stepped up their hitting on the road, and they get more hits than their opponents more often than not in both wins and losses.

While the 58.7% of the Devils' wins come as they outhit their opponent, that 58.3% of their losses come in the same state means that I don't think that there is a correlation between wins and outhitting the opponent. Furthermore, there wasn't a large difference in average hits in wins or losses; those were only significant when the Devils were at home aside from one outhitting the other. Because the initial assumption about NHL's numbers - the scorers recording real time super stats being correct - is suspect, a complete conclusion cannot be reached. However, based on what numbers we do have access to, I am confident in stating that the Devils were definitely not a soft team lacking in physical play. Even when they were being outhit, they average a solid number of hits - moreso when they are leading other team in hits. While the loss of Vitaly Vishnevski will hurt in the hit department and the team in general isn't a hit machine, I don't expect the Devils to become physical doormats in the coming season based on what they showed last season.

Labels: , , ,

Of course, if you find anything wrong or anything lacking, please feel free to e-mail me or leave a comment here.

Regarding significant average differences; if I really, really wanted to spend the time getting the standard deviation of hits in all those cases as well as assuming a normal distribution (which is likely not the case), I could have done a comparison. Instead, I just went with percent difference as an estimate.
Great analysis. It's impressive that you point out that, 1, if the devils outhit their opponents by a hit in their wins and losses (Table 2), hits might not mean anything; 2, the scorer could make a big difference, as we can imagine NHL hiring a constant group of scorer around NJ to do the Devils game and these scorers under-report hits (or set a tough standard on hits) consistently.

However, I'd like to point out that even if we find a strong correlation btw hits and wins, it might not mean causation. It could be that the Devils were pumped up before some random games, and they ended up playing better defence, scoring more and hitting more.

Nevertheless, even if we want to take these data seriously, perhaps we could have a separate category for shootout win/loss. After all, these games are "ties" and number of hits might not translate into individual shootout performances.
Thanks Rod. You're absolutely right, correlation NEVER means causation.

I don't know about separating out "ties." For starters, they yield results in the standings. Also, a team that has been outhitting the other team could conceivably wear them down and have them play worse than usual - which would have an effect in a longer game and in a shootout. It's possible, is what I'm saying.

I do agree the that consistent scorers from the NHL would help in making the RTSS more reliable. A better guide to judge what is a hit or a takeaway or a giveaway and such would help in reducing subjectivity. Basically, some instructional training may help. Then again, for all I know, the NHL has such a program and the scorers just rely on their own views regardless of the training.
Post a Comment

<< Home
Blog Directory - Blogged