The Pittsburgh Penguins recent struggles have been the hot buzz on the internet ever since returning from the Olympic break, and everybody has a theory about what is wrong and how they should fix it. Most of these seem to come back to possession, pointing out the recent downward trend that has seen the Pens drop to a 50.1 Close FF%.
There has been a lot of recent complaints about the D, blaming the poor possession on them and assuming that when some players return from injury the numbers will automatically pick back up. However, we looked at the numbers recently and determined that the Defenseman themselves have fairly limited control over generating shot attempts, as well as some interesting information about who is responsible for generating offense while they are on the ice, which in turn debunked some of the complaints about Bottom 6 scoring. We also looked at the general effectiveness of using advanced metrics to predict playoff success, which is the basis for the current worries about the Pens possession numbers because people fear it means their playoff chances are doomed. Then we took that a step further and looked at the correlations for the East and the West separately.
So what we are going to do this time is a bit similar to that, but instead of trying to predict the future, which is sketchy at the best of times, we are just going to take a look at the Pens current season, what influences the outcome of the game, and how playing in different situations influences the results. Having thrown the spreadsheets together a couple days ago the correlations include up through game 75 this season, which should be more than enough of a sample size to account for outside effects. The team record, on the other hand, I was able to simply add in the most recent game, which gives us 76 total games to work with.
We are going to look at the team's record for the season and how it correlates to GF%, SF%, FF%, and CF%. We are looking at those values in overall, 5-on-5, and 5-on-5 Close situations. And in order to look at how the different situations influence correlation, we are looking at the values based on the IIHF 3-point scoring system, the NHL loser point scoring system, the old NHL tie system, and a straight up Win-Loss system.
So far this season the Penguins team record is 39 Regulation Wins, 9 Overtime Wins, 5 Overtime Losses, and 23 Regulation Losses. For simplicity, all further standings will be listed in that order, as it is in the IIHF standings, so the Pens overall record would be 39-9-5-23.
The Penguins had 43 games with a GF% over 50% and 26 games with a GF% under 50%. At 5-on-5 that drops to 32 over 50% and 30 under 50%, while in Close situations it is 36 over 50% and 30 under 50%.
If you noticed that doesn't add up to 76 games, it is because there were games which ended in a 50% tie and were decided in the shootout, of which the Pens went 0-5-2-0.
The overall GF% has the highest correlation to determine who wins or who loses, which is obvious because the winner is the one who has more goals at the end of the night. However, because of the shootouts, it is not a perfect correlation.
In games with an overall GF% over 50%, the Pens record is 39-4-0-0. In games with an overall GF% under 50%, the Pens record is 0-0-3-23. The correlations are IIHF .7301, NHL .6737, Tie .7290, and W-L .6350.
In games with a 5-on-5 GF% over 50%, the Pens record is 30-0-1-1. In games with a 5-on-5 GF% under 50%, the Pens record is 2-7-1-20. The correlations are IIHF .5513, NHL .4419, Tie .6039, and W-L .3951.
In games with a Close GF% over 50%, the Pens record is 32-2-1-1. In games with a Close GF% under 50%, the Pens record is 2-5-3-20. The correlations are IIHF .6139, NHL .5149, Tie .6346, and W-L .4981.
Obviously, these numbers are meaningless. We already know the team that scores the most wins. However, I wanted to include them as a reference. The difference the overall GF% is below a perfect 1.000 is explained by the shootout, since those games ended in an even 50% GF%. However, in comparing that number to the 5-on-5 and Close GF% we can get an idea of how special teams success influenced the outcome.
The Penguins had 42 games with an SF% over 50% and 32 games with a SF% under 50%. At 5-on-5 that drops to 37 over 50% and 36 under 50%, while in Close situations it is 37 over 50% and 32 under 50%.
In games with an overall SF% over 50%, the Pens record is 21-5-3-13. In games with an overall SF% under 50%, the Pens record is 17-3-2-10. The correlations are IIHF .0020, NHL .0036, Tie .0014, and W-L .0031.
In games with a 5-on-5 SF% over 50%, the Pens record is 23-3-2-9. In games with a 5-on-5 SF% under 50%, the Pens record is 16-5-3-12. The correlations are IIHF .0289, NHL .0176, Tie .0327, and W-L .0193.
In games with a Close SF% over 50%, the Pens record is 24-4-1-8. In games with a Close SF% under 50%, the Pens record is 11-5-3-13. The correlations are IIHF .0918, NHL .0712, Tie .0984, and W-L .0690.
Overall the correlations are low enough as to indicate that the shot differential is utterly meaningless in determining who wins and who loses. If we are attempting to apply this to the playoffs then we are only worried in W-L, in which case we see the Pens 26-16 when outshooting their opponents and 20-12 when being outshot, not a big difference there. However, it improves when looking at 5-on-5, 26-11 when outshooting their opponents and 21-15 when being outshot. And even better still at Close, 28-9 when outshooting their opponent and 16-16 when being outshot. It would appear that leading shots increases the Pens chances of winning, but being outshot doesn't necessarily limit their ability to win. And score effects to appear pretty clearly, with 5-on-5 higher than overall and Close higher still.
The Penguins had 39 games with an FF% over 50% and 37 games with an FF% under 50%. At 5-on-5 that plummets to 30 over 50% and 45 under 50%, while Close situations improves to 36 over 50% and 29 under 50%.
In games with an overall FF% over 50%, the Pens record is 19-5-3-12. In games with overall FF% under 50%, the Pens record is 20-4-2-11. The correlations are IIHF .0000, NHL .0012, Tie -.0003, and W-L .0012.
In games with a 5-on-5 FF% over 50%, the Pens record is 16-4-2-8. In games with a 5-on-5 FF% under 50% the Pens record is 22-5-3-15. The correlations are IIHF .0225, NHL .0200, Tie .0195, and W-L .0251.
In games with a Close FF% over 50%, the Pens record is 23-4-1-8. In games with a Close FF% under 50%, the Pens record is 10-4-3-12. The correlations are IIHF .0968, NHL .0820, Tie .0984, and W-L .0812.
Like SF%, the correlations are low enough to be practically meaningless to try to use shot attempts to determine who will win or who will lose. The Pens are 24-15 when outshooting their opponents and 24-13 when being outshot, practically no difference whatsoever there. At 5-on-5 though we have the strange 20-10 while outshooting their opponents and 27-18 when being outshot. But much better in Close situations, 27-9 when outshooting their opponents and 14-15 when being outshot. Like SF% we see score effects at work, as 5-on-5 gets a little better but Close is the best. Interestingly, overall the SF% is a little better, and 5-on-5 its a tough call, but then in Close FF% does get the edge. We will have to keep that in mind and look into that further.
The Penguins had 33 games with a CF% over 50% and 40 games with a CF% under 50%. At 5-on-5 that drops to 30 over 50% and 44 under 50%, while Close improves to 39 over 50% and 33 under 50%.
In games with an overall CF% over 50%, the Pens record is 14-5-3-11. In games with an overall CF% under 50%, the Pens record is 24-4-2-10. The correlations are IIHF -.0126, NHL -.097, Tie -.0269, and W-L -0.113.
In games with a 5-on-5 CF% over 50%, the Pens record is 15-3-2-10. In games with a 5-on-5 CF% under 50%, the Pens record is 23-5-3-13. The correlations are IIHF -.0002, NHL .0000, Tie -.0008, and W-L .0003.
In games with a Close CF% over 50%, the Pens record is 24-5-2-8. In games with a Close CF% under 50%, the Pens record is 13-3-3-14. The correlations are IIHF .0704, NHL .0680, Tie .0682, and W-L .0649.
Once more we see that the numbers are low enough to be of no use in trying to determine who will win and who will lose. The Pens are 19-14 when outshooting their opponents and 28-12 when being outshot. At 5-on-5 this becomes 18-12 while outshooting their opponents and 28-16 when being outshot. Then Close gets back where we expect to see it with 29-10 while outshooting their opponents and 16-17 while being outshot. We see how important score effects are by looking at Close situations, but the overall and 5-on-5 numbers are way out of whack. We also see that CF% is actually lower than both SF% and FF%. This definitely needs to be looked into further.
One of the issues with trying to compare to Points earned or simple W-L is that it provides a much more narrow field. So let's try to compare the possession metrics to overall GF%. We want to use the overall GF% because it is the closest representation of who wins and who loses, being just slightly taken off track because of the shootout. That way we have a wider array of data points with which to determine our correlations.
SF% .0085, 5-on-5 SF% .0609, Close SF% .1801.
FF% .0015, 5-on-5 FF% .0421, Close FF% .1629.
CF% -.0136, 5-on-5 CF% .0010, Close CF% .1234.
So there you have it, it jumps out more clearly now. The score effects are incredibly important when it comes to shot attempts, with the Close values all being significantly higher correlations than overall or 5-on-5. So when looking at possession always use Close data, while scoring you want to use the overall numbers.
However, we do also see more clearly that, for the Penguins at least, CF% is not the ideal metric, and in fact just plain old shots are better. However, being familiar with the Pens system we can make some guesses to explain these results. The Pens defensive system is built around a style that emphasizes players blocking shots, so having a low CF% is not particularly indicative of not playing well. As for why SF% is slightly better than FF%, my best guess is that the Pens offensive system emphasizes shot quality over shot quantity, so shots that actually reach the net are more important that just being able to fling anything and everything towards the net.
Interestingly these numbers are quite close to what we found by examining 6 years worth of data in the East. Those numbers gave us FF% .1942 and CF% .1844, so we can assume that this season isn't just an anomaly because we have 6 years of data that gave us similar results. Which is that there is a low correlation between possession and winning for the Pens. And in particular it would appear that while outshooting is better, getting outshot isn't necessarily a bad thing.