Game control metrics love what Brown did against Quinnipiac
Ranking teams within any group is an exercise that can never be simple. On the face of it, it doesn’t seem like a difficult thing to do; wins and losses, point differential, common opponents all offer building blocks. We just have to pick which ones we value most, fit them together, and bam: rankings. But this is a classic case of the devil being in the details.
Wins and losses typically determine the champion in most leagues, but even in a perfectly balanced round robin league, not all wins are created equal. Ranking based on something like point differential can potentially overweight garbage-time points. Using opponents’ records or opponents’ opponents’ starts to take rankings too far away from what happened on the field. In other words, all of the building blocks have flaws, and when you start to refine them away, you end up faced with a slew of trade-offs that are difficult to reconcile.
Introducing LacrosseReference’s brand new rankings!!!
I’m kidding of course; we may try our hand with rankings in the future, but for now, we are going to stick to raw statistics. The building blocks in other words. We started this post talking about ranking systems, but didn’t mention one of our favorite ranking system building blocks: game control. To simplify, game control is a way to measure how easily a team wins games. A team that scores 5 goals to start the game, then steadily increases their lead throughout is more impressive than a team that struggles early, mounts a big comeback, and squeaks out a win at the end. Both are wins, but one win is more impressive. This also shortcuts the issue with aggregate point differential overweighting garbage time. If you are up 10 and score with 1 second left, your game control metrics don’t change (you were going to win anyway).
Of course, it doesn’t account for variation in quality of opponents, so you have to live what that or further refine by using some of the other building blocks. But again, that is the premise of this post: rankings are hard. Game control doesn’t solve that.
A few things about our approach
Before we get into the results, a few things about how we’ve implemented game control. First, we are using a cut-off of 75% win odds as “in control.” This means that if you go up early, and lead by a goal the whole game, but your win odds never break 75%, you won’t be looked on kindly by the model.
For reference using the Lacrosse Reference win probability model, a home team up 1 goal doesn’t hit 75% win odds, on average, until there are between 4 and 6 minutes left in the game. If they are up 2 goals, they will generally hover around 75% no matter when it happens, until about the end of the 3rd quarter; win odds with a two goal lead will increase steadily after that. A team up 3 goals, who I think we would be ok saying is in control, will always be above 75%. See our earlier write-up of our methodology for calculating win odds.
Second, in sorting the teams, we’ve excluded the first quarter from the average. So the average control percentage that you see below is calculated by taking the team’s average game control percentage from quarters 2 through 4. This doesn’t affect the actual rankings in a meaningful way, but I do think it’s a more reasonable way to calculate the average. It comes down to what perfect means. If you include the first quarter, where it’s almost impossible to crack 75% based on pure uncertainty, then the best score you can get is .75. If you exclude quarter 1, you can get pretty close to 1.000. Just preference, but I prefer a rating system that goes from 0 to 100.
Too Early to Tell
So how does this infant season look when viewed through our game-control colored glasses? Obviously, when some teams have only played one game, any metrics that rank teams are going to be a bit squishy, but to introduce the concept, we thought it was still worthwhile to show the current list. We’ll update this list periodically as we get into conference play and the teams become (theoretically) more evenly matched. So without further ado, the top 15:
Team | Games | Win% | Q1 Avg | Q2 Avg | Q3 Avg | Q4 Avg | Q2-4 Avg |
---|---|---|---|---|---|---|---|
Brown | 1 | 1.000 | 0.141 | 0.750 | 1.000 | 1.000 | 0.917 |
Marquette | 1 | 1.000 | 0.000 | 0.609 | 1.000 | 1.000 | 0.870 |
Princeton | 2 | 1.000 | 0.165 | 0.570 | 0.949 | 1.000 | 0.840 |
Virginia | 2 | 1.000 | 0.036 | 0.650 | 0.810 | 0.967 | 0.809 |
Maryland | 2 | 1.000 | 0.000 | 0.317 | 0.936 | 1.000 | 0.751 |
North Carolina | 3 | 1.000 | 0.067 | 0.503 | 0.764 | 0.914 | 0.727 |
Canisius | 1 | 1.000 | 0.000 | 0.213 | 1.000 | 0.907 | 0.707 |
Michigan | 4 | 1.000 | 0.081 | 0.313 | 0.799 | 1.000 | 0.704 |
Towson | 1 | 1.000 | 0.000 | 0.162 | 0.895 | 1.000 | 0.686 |
Binghamton | 1 | 1.000 | 0.000 | 0.231 | 0.841 | 0.827 | 0.633 |
Penn State | 3 | 1.000 | 0.000 | 0.253 | 0.676 | 0.911 | 0.614 |
Richmond | 3 | 1.000 | 0.000 | 0.340 | 0.685 | 0.786 | 0.604 |
Harvard | 2 | 1.000 | 0.000 | 0.024 | 0.741 | 1.000 | 0.588 |
Army | 3 | 0.667 | 0.062 | 0.500 | 0.681 | 0.573 | 0.585 |
Yale | 1 | 1.000 | 0.000 | 0.000 | 0.705 | 1.000 | 0.568 |
(A couple things to note in this table. “Games” refers to the number of games for which game logs are available online. Games without play-by-play can’t factor into our models, so they are excluded. “Win%” is taken from Inside Lacrosse, so it is accurate as of Feb 22. The four individual quarter averages indicate the percentage of time within each quarter that the team was in control of their games. The final column is an average of the Q2, Q3, and Q4 averages. For example, a value of .917 in the last column means that Brown was “in control” of the game for 91.7% of the 45 minutes played after the first quarter.)
A few things stand out. First and foremost, the game control model loved Brown’s opening day shellacking of Quinnipiac. Granted, it’s one game, but they are one of only 12 teams to have gotten above the control threshold in the first quarter. And overall, they were in control for 91% of the game and 100% of the second half. Only Marquette was able to stay in control for the entire second half.
Also, 48 teams were never once in control of their game. Given that we’ve played so few games and that some of the matchups were a bit lopsided, this isn’t particularly surprising. Obviously, this list will whittle down as the season goes along, but it does give us an opportunity to have a “last” team to ever be in control of a game. I hate to harp on the negative, but perhaps this could be an interesting anecdote to follow.
Last thing that stood out were a few teams that seem to be underperforming or overperforming their core metrics. Vermont, for example, has a 3-0 record, but they’ve been in control for just 30% of their total game time. This suggests that with a few bounces going a different way, they may not have that sterling record. Contrast that with Army, 14th on our list, ahead of 11 undefeated teams. Basically, Army had a bad day against Rutgers, but have been fairly dominant since. Could portend good things to come for the Black Knights.
See the full D1 game control standings here:
Team | Games | Win% | Q1 Avg | Q2 Avg | Q3 Avg | Q4 Avg | Q2-4 Avg. |
---|---|---|---|---|---|---|---|
Brown | 1 | 1.000 | 0.141 | 0.750 | 1.000 | 1.000 | 0.917 |
Marquette | 1 | 1.000 | 0.000 | 0.609 | 1.000 | 1.000 | 0.870 |
Princeton | 2 | 1.000 | 0.165 | 0.570 | 0.949 | 1.000 | 0.840 |
Virginia | 2 | 1.000 | 0.036 | 0.650 | 0.810 | 0.967 | 0.809 |
Maryland | 2 | 1.000 | 0.000 | 0.317 | 0.936 | 1.000 | 0.751 |
North Carolina | 3 | 1.000 | 0.067 | 0.503 | 0.764 | 0.914 | 0.727 |
Canisius | 1 | 1.000 | 0.000 | 0.213 | 1.000 | 0.907 | 0.707 |
Michigan | 4 | 1.000 | 0.081 | 0.313 | 0.799 | 1.000 | 0.704 |
Towson | 1 | 1.000 | 0.000 | 0.162 | 0.895 | 1.000 | 0.686 |
Binghamton | 1 | 1.000 | 0.000 | 0.231 | 0.841 | 0.827 | 0.633 |
Penn State | 3 | 1.000 | 0.000 | 0.253 | 0.676 | 0.911 | 0.614 |
Richmond | 3 | 1.000 | 0.000 | 0.340 | 0.685 | 0.786 | 0.604 |
Harvard | 2 | 1.000 | 0.000 | 0.024 | 0.741 | 1.000 | 0.588 |
Army | 3 | 0.667 | 0.062 | 0.500 | 0.681 | 0.573 | 0.585 |
Yale | 1 | 1.000 | 0.000 | 0.000 | 0.705 | 1.000 | 0.568 |
Johns Hopkins | 3 | 1.000 | 0.000 | 0.294 | 0.554 | 0.823 | 0.557 |
Notre Dame | 1 | 1.000 | 0.000 | 0.000 | 0.648 | 1.000 | 0.549 |
Ohio State | 3 | 1.000 | 0.000 | 0.181 | 0.652 | 0.756 | 0.529 |
Syracuse | 2 | 1.000 | 0.009 | 0.465 | 0.500 | 0.586 | 0.517 |
Hobart and William | 2 | 0.500 | 0.105 | 0.536 | 0.533 | 0.402 | 0.490 |
Boston U | 3 | 1.000 | 0.000 | 0.259 | 0.538 | 0.647 | 0.482 |
Rutgers | 1 | 1.000 | 0.000 | 0.000 | 0.511 | 0.932 | 0.481 |
Lehigh | 3 | 0.667 | 0.061 | 0.195 | 0.509 | 0.669 | 0.458 |
Mercer | 2 | 0.500 | 0.000 | 0.135 | 0.479 | 0.669 | 0.428 |
Denver | 2 | 1.000 | 0.000 | 0.012 | 0.457 | 0.763 | 0.410 |
Marist | 3 | 0.667 | 0.000 | 0.053 | 0.457 | 0.717 | 0.409 |
Delaware | 3 | 0.333 | 0.000 | 0.127 | 0.589 | 0.396 | 0.371 |
Duke | 4 | 0.500 | 0.027 | 0.165 | 0.325 | 0.479 | 0.323 |
Vermont | 3 | 1.000 | 0.028 | 0.310 | 0.301 | 0.299 | 0.304 |
Providence | 2 | 0.667 | 0.000 | 0.000 | 0.305 | 0.598 | 0.301 |
Bryant | 3 | 0.333 | 0.000 | 0.025 | 0.409 | 0.459 | 0.298 |
Robert Morris | 3 | 0.667 | 0.000 | 0.000 | 0.320 | 0.536 | 0.285 |
Monmouth | 3 | 0.667 | 0.012 | 0.047 | 0.293 | 0.460 | 0.267 |
Navy | 2 | 0.333 | 0.000 | 0.220 | 0.169 | 0.347 | 0.245 |
Fairfield | 2 | 0.500 | 0.000 | 0.000 | 0.272 | 0.298 | 0.190 |
High Point | 3 | 0.333 | 0.000 | 0.000 | 0.185 | 0.358 | 0.181 |
Furman | 4 | 0.000 | 0.000 | 0.144 | 0.097 | 0.271 | 0.171 |
Hofstra | 1 | 1.000 | 0.000 | 0.000 | 0.075 | 0.203 | 0.093 |
Manhattan | 3 | 0.333 | 0.000 | 0.132 | 0.124 | 0.000 | 0.085 |
Stony Brook | 1 | 1.000 | 0.000 | 0.000 | 0.000 | 0.204 | 0.068 |
Air Force | 3 | 0.333 | 0.000 | 0.000 | 0.000 | 0.161 | 0.054 |
Bucknell | 3 | 0.667 | 0.000 | 0.000 | 0.000 | 0.120 | 0.040 |
Wagner | 3 | 0.333 | 0.000 | 0.000 | 0.013 | 0.079 | 0.030 |
Sacred Heart | 2 | 0.500 | 0.000 | 0.051 | 0.012 | 0.000 | 0.021 |
Albany | 1 | 0.000 | 0.000 | 0.037 | 0.017 | 0.000 | 0.018 |
Cleveland State | 2 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
Quinnipiac | 1 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
Villanova | 2 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
Jacksonville | 2 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
UMBC | 3 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
Massachusetts-Lowell | 2 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
Mount St Marys | 3 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
Dartmouth | 1 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
Siena | 3 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
Cornell | 1 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
Drexel | 1 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
Massachusetts | 2 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
VMI | 1 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
Colgate | 2 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
Georgetown | 2 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
NJIT | 2 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
Lafayette | 2 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
Holy Cross | 2 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
Hartford | 1 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
Bellarmine | 2 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
Detroit | 3 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
Saint Joseph’s | 1 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |
Loyola MD | 1 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 | 0.000 |