Photo Credit: © Tom Szczerbowski | USA Today

Goal Differential’s Great, But How Did You Get There?

In the course of rebuilding the franchise, the Toronto Maple Leafs have built something special. Watching as they have amassed the quality players that are deployed each night, and even the ones that are only deployed some nights, has been a joy to witness.

This season, this carefully-crafted roster has enough points to sit 3rd in the Atlantic Division behind the pair of one legendary Tampa Bay Lightning team, and one red-hot Boston Bruins team. However, at a league-wide level, the Toronto is 5th, with Tampa obviously at number one and Boston in 3rd. That is, again, in terms of points.

Points don’t always tell the whole story, though.

There is a popular statistic, as mentioned in the title, to gauge the quality of a team, which is goal differential. The number is calculated quite simply: take a team’s total goals for and subtract that team’s total goals against. Its intuitive nature leads to popularity, and being directly available in the standings on NHL.com doesn’t hurt either.

On an anecdotal level, it seems consistently true that the top goal differential teams in the first half of the season tend to find themselves in playoff spots at the end of the season. The historical data proves this point as well. Though few studies exist, this piece from Kyle Gipe at TheHockeyWriters used goal differential compared to expected league scoring rates to predict future wins, with moderate success.

There’s another, less popular metric that isn’t so intuitive, and it can disagree at times.

It’s known colloquially as “PDO”, an acronym with no meaning. How the metric is calculated is almost as simple as goal differential: take a team’s total save percentage (team shots-on-goal saved, divided by team shots-on-goal against) and add that team’s total shooting percentage (team goals scored, divided by team total shots-on-goal for). This is typically taken only at even-strength, whereas goal differential includes goals from all situations (5-on-5, shorthanded, powerplay, empty net).

An explanation of this metric’s importance can be found, to a slight exaggeration in my opinion, in this piece by ‘Hawerchuk’ on Arctic Ice Hockey. It’s a very short piece by ‘Hawerchuk’, who created the website behindthenet.ca, a former go-to location for hockey analysts for the formerly new advanced statistics like Corsi and Fenwick.

The boilerplate thinking on PDO is that it is a way to measure luck in hockey. Average shooting percentage and average save percentage are almost completely constant season over season. Teams that reach abnormal levels of either are extremely likely to come back to normal — this is termed in a colloquial as ‘regressing to the mean’.

Of course, some oversimplification occurs here. The truth is more complex. There will undoubtedly be teams that have naturally higher shooting percentages. Some teams are loaded with high quality shooters that defy the typical averages — Steven Stamkos, Patrik Laine, and Auston Matthews come to mind. There will also be teams that have above-average goalie tandems, that are able to consistently perform above the 91.5% mean save percentage.

What Does This Mean for Toronto?

Well, Toronto has an excellent goal differential this year. They currently sit at +50 goals, which is second in the league behind, again, the dominant Tampa Bay Lightning at +84. Close behind on their tail is the Calgary Flames, who are second in the league at +39.

Goal differential is a statistic that Babcock is known to love. This quote from Elliotte Friedman’s 30 Thoughts column (from before there were 31!) evidences this:

During a Saturday morning chat, Mike Babcock was asked about the top teams in the league. He grabbed the daily stat sheet and circled the five with the best goal differential — Pittsburgh, Tampa Bay, Chicago, Nashville and St. Louis. “There you go,” he said. Note to NHL: Babcock thinks that statistic should be on the front of the package. It’s on Page Two.

The Leafs also have a very high PDO. At this time, they have the highest PDO in the league, unfortunately, with 102.9 (Sv% 92.7, Sh% 9.98).

It shouldn’t be hard to see that a team with a high PDO is also likely to have an inflated goal differential. High save percentage will mean low goals against, and high shooting percentage will mean high goals for. Either or both of those things will mean a high goal differential.

While there’s obviously a high shooting talent in Toronto, and a quality starting goaltender, for a team to be good you want to be able to have a similar goal differential by having far higher shots for — meaning that you’ll score the same number of goals, it’ll just take you more shots to do it. Similarly, you want to have fewer shots against, so your goaltender can have an average save % and not allow any more goals.

This is the case with the aforementioned Calgary Flames. They sit behind the Leafs — by a fair margin, but still, next in rank — but have an essentially even PDO at 100.2. This is only to say that having a good team (as measured by goal differential) doesn’t require that you have a high PDO.

The concern here is that it’s possible the Leafs’ goal differential is inflated by way of their high PDO. In fact, you can directly calculate how much it’s inflated.

The Leafs are 2.9% over even in PDO. If they had an average shooting percentage of 8.5%, their 1704 shots-on-goal would have led to about 145 goals for instead of their present 170. If they had an average save percentage of 91.5%, their 1821 shots against would have led to 155 goals against, instead of their present 128. That would equate to a goal differential of -10. Even if you say Frederik Andersen and Garret Sparks could be good for a higher baseline save percentage, and the scoring talent in Toronto is worthy of a higher baseline shooting percentage, it’s still very difficult to argue that their present goal differential is not inflated to some degree.


To point this out is not to say that the Leafs are a bad team. At least, that’s not the intent I had in writing this. What importance can be drawn from these numbers is that the Leafs need to work on taking more shots, and allowing fewer. It’s the same philosophy that Corsi and Fenwick live off of: having a positive shot share is going to help you win games.

At some point, the Leafs’ shooting could cool to average. It could even cool to below average. As we head into the playoffs, it’s really not the time for a cold streak to happen. The sole reason why I bring it up is this: if the Leafs suddenly forget how to score/save goals over the next couple months, PDO is probably why.

I certainly intend to look more closely at how PDO and goal differential are linked. While goal differential itself is highly linked to success, the question is PDO, and thus randomness, highly linked to goal differential. It certainly could be, and will take a little bit of analysis for me to answer that question definitely. My hypothesis is certainly that it does have a strong link, and that goal differential may come with a huge caveat where it is successful in predicting future success, unless it comes from a high PDO.

All in all, the Leafs are shaped for a brutal route to the Stanley Cup yet again, and will have hard-fought battles to get through Boston and, likely, Tampa Bay. Hopefully, we learn that this talented roster is, in fact, able to sustain a high PDO through the playoffs and not go cold. Time is, as always, the only way to tell.