This season, I’ve gone back to my roots. By which I mean that I’ve gotten to start back up an ongoing — albeit a little sporadic — project of tracking microstats from the Lehigh Valley Phantoms games, all as a nice way to get pretty granular about how things are going for the team this season, in a season where the stakes feel like they’ve never been higher. It’s all a process indeed, but it’s coming together and it has come together thanks in large part to the work of my lovely data tracking partner in crime Brad Keffer (everyone say thank you, Brad!).
Over the next handful of days — across various outlets — we’re going to be diving into this first batch of data, looking at what it says and what it means for how things are going for the Phantoms players on an individual level, as well as the team on a collective level. But if the idea of a deep dive into the minutiae of this data sounds a little bit scary and daunting, worry not, because we’re here to work through a bit of an explainer together.
So, without any further ado, let’s talk about what we’ve tracked and what the heck it even means. Feel free to have a peek at the numbers in these spreadsheets here, and let’s get right into it.
Why 5-on-5
Okay, just kidding, we have a little bit further of ado. We’ll begin with a note that all of the data we see through this project was tracked at 5-on-5 and 5-on-5 only. Why? We’re glad you asked.
Broadly speaking, there’s the most basic level of the fact that, at 5-on-5, teams are more or less on equal footing and each side has an equal opportunity to create offense. Whoever does or does not move the needle in that regard tells us something, but using that equal ground as a starting point is the most useful way of evaluating. It also helps us adjust in some ways for usage. Obviously based on the hierarchy of lines and pairs, minutes won’t be doled out completely evenly, but this helps us strip away the outliers, if you will — that is, players who are given time on the power play will inflate a player’s impacts if we look at all situations, by virtue of being on the ice while more chances are being created for their team, while those played on the penalty kill will have their numbers dragged down by being on the ice for more chances against. Sticking to 5-on-5 lets us strip out as many confounding factors as possible, and whatever happens from there will paint a picture for us.
Shot impacts and what they mean
The main scope of this project is that of shot impacts, both on an on-ice and individual level. And we do this through a few different areas. You’ll see shots on goal tracked, the same as you would find on the league reports (though theirs are across all situations). The shots that make it on goal are obviously valuable to track, but they’re don’t give us the full picture, so we’ll also be keeping track of the totals in shot attempts, or, every shot that is taken that makes its way on net, or misses the net, or is blocked. You’ll see those noted as CF, CA, and CF% (that is, Corsi-For, Against, and percentage, which is a somewhat needlessly fussy way to note what’s just every shot that’s sent even in the direction of the net). And all of this helps us to see a broader picture of how much offense is being created or allowed, and to get a general sense of possession, who has the puck most often when a certain player is on the ice.
But it’s also true that all shots are not created equally, and while we’re limited in terms of what we’re able to capture as far as pre-shot movement goes, we can also give the shots a bit different of weight depending on where they come from. Because, as we can intuit, as a general rule, a shot from closer to the net had a greater chance of beating a goalie than one from very far away, and it’s useful to know not just who’s creating the greatest volume of shots, but also who’s creating a good proportion of high-quality shots.
And the way we measure this is by, first, envisioning the offensive zone side of the ice. Coaches love to talk about protecting the house, or home plate, and that’s an area of the ice that we’re honing in on here. If we picture the faceoff dots, and then imagine drawing two lines straight up to the top of the circles, connecting them with a horizontal line, and then returning to the dots and drawing two more lines diagonally to go from the dots down to the corners of the net, we have a shape that roughly looks like, well, an upside down house or home plate base. And now that we’ve done all of the heavy lifting to visualize this zone, we’ll get to the point — any shot taken from this area is a more or less medium-danger shot, and they’ll be recorded as scoring chances (either as a Scoring Chance For (SCF) or a Scoring Chance Against (SCA). On top of this, any shot taken in the slot from the level of the faceoff dots or below is a high danger chance, and will be recorded as a High Danger Scoring Chance For (HDCF) or a High Danger Scoring Chance Against (HDCA).
We’ll track these on an individual level for each player, see how many of each type of chance they create on their own, but we’ll also look at the on-ice side as well — that is, when each player is on the ice, how many of each type of chance do the Phantoms create, and how many are put up against them, all as a way of approximating play driving and possession. The percentage differentials will want to hover around the average of 50 percent, and any averages put up above that tells us that the team is generally creating more than they give up when that player’s on the ice, and anything below would suggest they’re getting caved in a bit.
Introducing shot assists
For long-time followers of these various tracking projects, we’re introducing something new! You may see on the spreadsheets columns for primary (SAs1) and secondary (SAs2) shot assists (which are a little clunky as abbreviations, and I’m allowed to say that because I made them up).
But as far as the stat goes, the concept is simple — just like every goal will have the last two touches preceding it added as assists, if applicable, every shot that gets put on net will have the two touches preceding it noted (again, if applicable).
And this is a way to help us approximate playmaking contributions in an expanded way. After all, assists tell us something, but relative to all of the other events that happen over the course of a game, there are just so few goals and assists that only a small portion of a picture is painted. This allows us to see who is feeding offense well, independent of the results that the shooters are having.
A note on the completeness of the dataset
By which we mean… it is not complete.
Here is when we drop in an acknowledgement that we’re working with a somewhat imperfect dataset here, and we must, based on the limitations of the tracking apparatus. That is, these games are all tracked from a video feed, and occasionally, weird things happen with that feed. It may drop out and a couple of minutes have disappeared and can’t be recovered. The camera may be focused on a faceoff circle, or a battle in the corner, or is stuck from the angle broadcasting from the International Space Station and a shot that would otherwise be tracked happens out of frame. The mascot might wander down through the seats and block out part of the frame for a hot second.
Things slip by, are missed, but we do the best with what we have. Eventually, a large enough dataset will be amassed that the handful of events that were missed will hardly bear any weight, and everything will more or less even out. That’s what we’re aiming for anyway. We’re just about at that point, but we’d be remiss if we didn’t at least acknowledge — begrudgingly — that we’re not perfect.
And that’s just about everything covered! Everyone is experts now, congratulations, gang.
