Welcome to Beyond the Basics!
My name is Zack Capozzi, and I run LacrosseReference.com, which focuses on developing and sharing new statistics and models for the sport.
The folks at USA Lacrosse Magazine offered me a chance to share some of my observations in a weekly column, and I jumped at the chance. Come back every Tuesday to go beyond the box score in both men’s and women’s lacrosse.
I listen to my readers. Don’t doubt that for a second. This is not some Ivory Tower operation here. I’m a lacrosse fan first and foremost, and it makes my day when I get a reader email with feedback on this column. And it makes my week when they give me a suggestion for a future topic.
And that is just what we have here today; your first-ever reader-inspired Beyond the Basics. Per their suggestion, we are going to dive into EGA (expected goals added), why it’s important and how it’s used to build the most objective Tewaaraton Watch List you’ve ever seen.
EGA: EXPECTED GOALS ADDED
OK, EGA stands for Expected Goals Added. For those familiar with baseball stats, it’s the lacrosse version of WAR. It’s a stat that takes everything a player does (that shows up in the box score) and puts it into a single number. EGA penalizes players for negative plays (turnovers/penalties) and gives credit for positive plays. The credit given is based on how often each type of play is expected to lead to a goal (hence the expected-goals name). So, a ground ball is worth more than a missed shot because ground balls lead to goals more often than missed shots do. (If you want to dig deeper, I wrote a more detailed article on the topic.)
Before EGA, how would one compare two player stat lines? The short answer is that you really couldn’t unless they only differed on one dimension. Is three goals, two assists and five turnovers a more valuable performance than four ground balls and one goal? The bias toward points as the end-all metric would probably make you think that the five-point effort is more impressive. It may not be.
With EGA as a guide, we can account for everything a player does and get a much deeper understanding of the relative contributions. As a glaring example, let’s compare EGA with points, which has been the dominant single-number statistic in lacrosse since time immemorial. One issue with points is that it double counts assists and goals. Only one goal was scored but two points were awarded. Is an assisted goal more valuable than an unassisted goal? I would argue the opposite, and it’s certainly not worth double. Because EGA starts with play-by-play as the raw data, it distinguishes between assisted and unassisted goals, giving half credit to both the scorer and the player who recorded the assist.
The second issue with points as the dominant statistic is that there is much more that goes into winning games than scoring. A focus on points masks the contributions of all the players that worked to create those goals. What about the player that caused the turnover and won the ground ball in the first place? And how do you account for a player with a ton of points AND a ton of turnovers. All these things matter. The process is as important as the result. Points only cares about results. EGA values the entire process.
We can look at some of the highest-usage players in Division I Men’s lacrosse to see where EGA detects more value than points alone would suggest.
THE EGA RECIPE
The origin of EGA was actually as an input to my Win Probability model. The calculation needed a way to determine whether a team’s chances of scoring a goal were high or low. Otherwise, the only inputs to the model would be time and score; it wouldn’t be able to update the win probability after a ground ball was picked up as an example. I needed a way to determine, over the next stretch of the game, how many goals we would expect each team to score based on the most recent plays.
As an example, let’s assume there are 100 ground ball pick-ups recorded in the play-by-play. We then look at the next 60 seconds of play after each GB and count the number of times that the GB-winning team scores compared to the number of times the opponent scores. As you can imagine, the team that won the GB is more likely to score a goal, but not always. Let’s say the GB-winning team scores 24 goals over the next 60 seconds and the opponent scores five. That would give us a net goal margin of 19. Divide that by the 100 original ground balls and we can estimate that a ground ball is worth .19 expected goals. Here are the numbers for each play type that a player can record:
-
Missed Shot: 0.19
-
Ground Ball: 0.19
-
Faceoff Win: 0.18
-
Pipe Shot: 0.16
-
Blocked Shot: 0.10
-
Assisted Goal: 0.04
-
Saved Shot: 0.03
-
Unassisted Goal: 0.02
-
Forced Turnover: -0.17
-
Unforced Turnover: -0.17
-
Penalty – 2 min: -0.31
-
Penalty – 30 sec: -0.33
-
Penalty – 1 min: -0.36
Note: Obviously, a player who scores or assists a goal gets the EGA credit for that play as well. The values above are tied to the chances of scoring a next goal since it’s also used to adjust the win probabilities.
Once we have that worked out, all we need to do is count the number of times that each player is credited with each play type, multiply by the play values and sum the total. There you go. The recipe for EGA in a nutshell. You can probably do it for yourself if you are capturing these box score values. (If you do try it for your team, I’d love to hear what you discover.)
The results are often surprising. After I posted this week’s top EGA games, I had several comments about why Jake Taylor wasn’t on it. After all, EGA is supposed to identify the players who added the most value, and surely the eight goals he scored in the thrashing that Notre Dame gave to Syracuse was valuable. And don’t get me wrong, great game for the kid and a great story, but his EGA numbers were not as high as you might expect. The reason is that, of those eight goals, seven were assisted, so he gets half credit. (Pat Kavanagh actually had a higher EGA in that game.)
And that is sort of the point, EGA gets you a clearer understanding of the true value each player contributed.