Perhaps my favourite part of the off-season is crafting potential line combinations. When hockey is absent, it allows us to anticipate the coming season, to figure out how to arrive at the most optimal setup. Trevor did a great job looking at this from the Sens’ perspective, so this week I’ll be focusing on the numbers: instead of what do we think the lineup will be, I’ll be looking into what I think the opening lineup should be.

Over the course of this three-part series, I’ll use a few different methods to determine some possibilities. Part one will be a method I’ve used before: chemistry.

A quick refresher on what I mean by chemistry: it’s a description of how much better two players would play together, as compared to other options. I measure this by taking the Corsi% of the combo together, and figuring out how much better/worse that is compared to how each player performed apart from each other.

For example in this most recent season, Derick Brassard and Mike Hoffman had a Corsi% of 57.52% when playing together, which is really good. When Brassard was playing without Hoffman, his went down to 52.52%, meaning he was better off with Hoffman than without. The same applies to Hoffman, as without Brassard his Corsi% fell to 47.47%.

Using this data we can find their chemistry score by adding up the differences:

Chemistry Score = (CF% Together - CF% Player 1 Without Player 2) + (CF% Together - CF% Player 2 Without Player 1)

Chemistry Score = (57.52 - 52.52) + (57.52 - 47.47)

Chemistry Score = 15.05

This is a classic example of a pairing with good chemistry. Positive scores mean good chemistry, negative scores the opposite. In this specific case 15.05 is considered high, meaning it might be a good idea to play Hoffman and Brassard together.

We can apply this to the rest of the roster to figure out which combinations might be the most efficient.

Defencemen

The table below shows chemistry scores for all Sens defence pairings in 2016-17, with a minimum 100 minutes together and apart. It’s important that we apply the TOI limit to both sides, to limit any small sample sizes. I chose 100 minutes as the cutoff because I’ve found that after that point scores become less affected by TOI.

Here’s the scores:

Chemistry Scores - Defencemen

Using this data, the Sens’ most optimal defence pairings last season would’ve been:

Claesson - Karlsson
Methot - Ceci
Phaneuf - Wideman

As interesting as this might be, this doesn’t tell us what it should be for next season, and we’re really just analyzing the past. The repeatability of chemistry scores is yet to be tested, and aside from that the Sens don’t even have the same players!

The most prominent switch is that with Methot gone, Erik Karlsson will need a new partner. Johnny Oduya and Thomas Chabot are both options and not listed in the data, although looking at the table it appears that Karlsson’s ideal partner should still be fellow countryman Fredrik Claesson. This makes sense, as Steady Freddy would give Karlsson a mobile partner to play off of, thus lessening the burden for the captain. It increases Karlsson’s impact because it allows him to have an extra passing option, the opposite of Methot who would steer clear of the offensive zone. With a chemistry score as high as 14.55, the highest on the list, I don’t see how that can be surpassed by Oduya or Chabot.

Then we arrive at the second pairing. Because Karlsson’s so good and takes up the next few spots, the next best pairing the team has available is Dion Phaneuf and Chris Wideman. Together they could create an offensively inclined duo as Ottawa’s second and third highest scoring blueliners last year.

That leaves Cody Ceci on the third pairing. This might not be so bad, and you could make the case he should take a step back and play some weaker competition after getting caught too often in the defensive zone last year. He would presumably start alongside Oduya, although this could all change depending on what results we start to see. I would avoid putting him next to Mark Borowiecki, as together their chemistry score was a frightening -5.05.

Using chemistry scores, the Sens’ optimal defence pairings should be:

Claesson - Karlsson (14.55)
Phaneuf - Wideman (4.41)
Oduya - Ceci (N/A)

Forwards

The forwards are a bit more tricky, simply because there’s three of them on a line. To account for this, I took all possible three-forward combinations that met the TOI requirements and the average of their chemistry scores. For example, Dzingel and Brassard had a score of 9.28, Dzingel and Stone had a score of 1.36, and Brassard and Stone had a score of 1.07. If you take the average of the three, you get a new line score of 3.90, which would presume that the line of Dzingel-Brassard-Stone has positive chemistry.

The following table has all of the Sens’ possible combinations, thanks to Guy Boucher’s tendency to blend forward lines.

Chemistry Scores - Forwards

With that data, we can make an optimal top-9 unit of:

Hoffman - Brassard - Stone (11.64)
Smith - Turris - Ryan (-0.82)
Dzingel - Pageau - Pyatt (7.27)

It’s worth noting that although a line may have a high chemistry score, that doesn’t necessarily mean they’re better. It just means that the combination is a more efficient use of talent, which is seen in the third line above. Line 2 combined for 2.34 goals per 60 minutes in 2016-17, whereas line 3 combined for only 1.77.

I should also note that although the second line has a negative score, it’s needed to enable the two high chemistry lines.

Now we have to work in the missing players, similar to how we managed the defencemen. Alex Burrows and Nate Thompson should be on the 4th line. The biggest notable omission is Clarke MacArthur, who spent most of last season on the IR. He could be a potential aid to that negative second line, as he posted a positive score next to Turris during his last full season in 2014-15, as well as during the most recent playoffs. He’s performed worse next to Ryan in both those scenarios, however, so maybe some lines should be shuffled accordingly. But for now, in terms of available options, Ryan may end up being the best one. That would shift Zack Smith down to the 4th line.

Hoffman - Brassard - Stone (11.64)
MacArthur - Turris - Ryan (N/A)
Dzingel - Pageau - Pyatt (7.27)
Smith - Thompson - Burrows (N/A)

That concludes part one of the three part series that will run tomorrow and Friday. What do you think of the line combos? What would you do differently? I created a viz tool, so you can play around with the data for different teams and positions.

Huge thanks to B_T (the awesome man behind Natural Stat Trick) for providing the data.