his is my first foray into FanPosts and also a first foray into what I hope will be a meaningful discussion of PDO statistics. I will gladly welcome constructive criticism of any kind.
For those not familiar with the PDO stat, a definition and history of the stat can be found here. One subtle but possibly important distinction here is that the PDO values I use below are for all situations. The definition referred to above identifies the statistic to be for even strength only.
I became obsessed, to a degree, with the PDO stat earlier this year. It all started with a Noon Number that J.P. posted on Feb 21, 2013. While trying to understand that particular Noon Number, I made this graph
NOTE: The X-Axis scale is from 0.7 to 1.3 and each bar represents 0.01 in PDO. My apologies.
that showed how strongly correlated a team's PDO was with winning a game. The graph shows how many times a team with a particular PDO won, loss or "tied" a game. The resulting graph was not particularly surprising because it wasn't much different than just counting how many goals each team won at the end of the game to see who won. It did surprise me, though, that getting a PDO for a game of 1.03 was enough to pretty much guarantee a win.
However, the realization that the Caps only required consistent mediocre performance to become a powerhouse along with previous overheard discussions about PDO and "luck" made me wonder whether PDO could identify when a team was on a hot streak (or slump) that was out of character for the talent level on the team. It also made we wonder whether PDO could give a hint as to the outcome of Stanley Cup playoff games at season's end.
To start, I began to compile team PDO statistics for the teams this season. Now that we are approaching the playoffs and a discussion has begun as to who the Caps should play in the first round, I decided to do a little analysis to see whether one opponent might be a better choice to make it through the first round than another. The results have been interesting (if they actually mean anything) because they seem to run counter to what most people think.
For starters, lets look at the PDO computed over the course of the season for the playoff contenders in the Eastern Conference:
These values, at first glance, look very exciting. The Caps are on par with Pittsburgh for the year. But we also see that Toronto is actually higher than both the Caps and Pittsburgh. Is that some sort of fluke? Let's look at plots of some of the Eastern Conference playoff team PDOs for the season to see how they got there:
What we can see is that Toronto and Pittsburgh have settled into the 1.020 to 1.030 range for the last 30 games or so. The Caps have been creeping up steadily and it is not quite as clear where they may settle out to. But do these numbers really mean anything? Could they just mean the teams have been lucky? Further research on the internet into PDO turned up this fine article. It does a pretty good job of putting limitations on how much PDO changes due to luck. All three of these teams seem very unlikely to be in the range they are without a high skill level playing a significant role. In fact, when I calculate the z-score for all three where the standard deviation due to chance is 0.011, they are over 2 standard deviations away from the average (Toronto is almost 3 SD away). This means to me that there is less than a 5% chance that the Caps have a high PDO by pure luck. The fact that the PDO has remained high and climbing implies to me that there is essentially a zero chance that their high PDO is luck.
So we have proof that the Caps are a skilled team. That is reassuring but not terribly enlightening.
One of the items I found fascinating about the article mentioned above, was the ability to predict how a team would fare in the standings based on their PDO. I was curious how the Caps fared so I used the coefficients from the article and plotted out the projected points and compared it to the Caps existing standings points throughout the season. (NOTE: Since the "Panther point" throws the equations off, anytime the Caps went into overtime or a shootout, they were awarded 1.5 points and this is referred to as "Adjusted Pts" in the plots)
Interesting. The Caps have underperformed in the standings from where they should be with their PDO. Their PDO implies that they should really be in the 65 point range in the standings (right up there with Pittsburgh, Chicago, Anaheim, Boston, etc).
One of my concerns, though, is whether a season long PDO is hiding a problem in a shorter term. With so many games thrown into the average, any recent trends are going to be hidden in the final result. So I went back to the data and created sliding windows of 10, 20, and 30 games each. For example, I took the PDO for games 1-10, then the PDO for games 2-11, etc. When I plot out the PDO values together for the Capitals, I see the following:
Again, the results are extremely promising. It also shows me that the 10 game sliding window is almost too noisy to make any sense of. The 20 game sliding window seems the easiest to see recent trends and it shows a very promising upward trend to the present day. Let's look at the PDO plots for the other Eastern Conference playoff teams:
They look to be holding pretty steady and pretty skilled. Not much of a change when Crosby got injured. An impressive team to say the least.
Ooohh! Looky looky. The bear appears to be tiring out. That could spell problems for them in the playoffs.
Next is Montreal:
Another team losing steam.
How about the Islanders:
Holding steady and possibly turning it up a notch for the playoffs. However, their PDO is pretty pedestrian.
How is Ottawa doing:
Another team that appears to be losing steam as the playoffs approach.
Holding steady. Consistency is definitely one of their strong points.
and finally, what about Winnipeg:
Losing momentum and the lowest PDO of the bunch. No surprise if they miss the playoffs entirely.
So, do these numbers mean anything going into the playoffs? I decided to use the same calculations on last year's playoff matchups and produced the following table of data:
|Delta Season PDO||SW20
|Delta SW20 PDO||SW20 Slope over last 5 games||SW20 Slope over last 3 games||1st
As you can see, the 20 game sliding window did a reasonably good job of identifying potential winners. It identified 5 of the eight pretty clearly, in my mind. St. Louis, Phoenix, Nashville, Los Angeles and New Jersey. It appears to miss the NY Rangers / Ottawa match up completely. By the numbers above, Ottawa should have taken out the Rangers in 5 or 6 games instead of losing in 7. The Boston / Washington series looks too close to call and the same goes for the Pittsburgh / Philadelphia series. Both of which went the full seven games.
The outlier of Ottawa and the Rangers annoyed me. So I decided to dig a little deeper. If folks remember correctly, Ottawa lost their star forward, Alfredsson, due to a concussion weeks before the playoffs. He did not return until game 7 of the NYR-OTT series. While it might be easy to say his impact on the PDO did not register in the 20 game sliding window, I have no proof of it. In fact, I couldn't find a significant change in either shooting % or save % after he was injured. Also, Gaborik, of the Rangers, was injured during the playoff series and yet the Rangers still managed to win.
One of the things about PDO that I found a little disturbing, is that approximately 90% of the value is based on the goalie's performance (and/or team defense) and only 10% is based on the offensive ability of the team. I decided, on a lark, to make the two values equal and create a weighted PDO where 55.6% of the save % is used and the shooting % is multiplied by 5. We can no longer assume the league average for the weighted PDO is 1.0 and I have to admit I was too lazy to try to calculate what the league average would be. Nor can we use the same deltas above as a reference to what may be a significant difference between two teams weighted PDOs.
Anyways, I calculated the weighted PDOs for last year's playoff teams and got the following table of data:
|Weighted PDO||Delta Weighted PDO||SW20 Weighted PDO||Delta SW20 Weighted PDO|
As we can see, not the Ottawa / Rangers series looks a heck of a lot closer than it did before. If we use 0.1 as a cut-off between a toss-up and a leaning, we see now that the NYR / OTT, WSH / BOS and PIT / PHI series are all toss-ups. New Jersey is still a winner against Florida. In the Western Conference, Los Angeles still defeats Vancouver and Nashville still defeats Detroit. The other games become toss-ups.
So, using this methodology, how does the Eastern Conference currently stand?
|Weighted PDO||Delta Weighted PDO||SW20 Weighted PDO||Delta SW20 Weighted PDO|
It looks like the Caps could walk all over the Islanders.
What if the Caps were matched up against others?
In conclusion, I don't think I want to see a Caps / Toronto series. Ottawa would actually be my first choice with this criteria.