LacrosseReference really got started during the 2016 NCAA tournament.  During the Maryland vs Syracuse quarterfinal game, Maryland scored with 11 minutes remaining in the 4th quarter to increase their lead to 10-6.  We started thinking, what are the odds that Syracuse comes back in this game?  I remember seeing win probabilities during ESPN.com’s coverage of MLB games 5 or 6 years ago; with each pitch they would adjust to show the odds of each team winning.  A quick bit of Googling didn’t turn anything up for the game I was watching (although FiveThirtyEight’s win odds were inspirational), and we were off and running.

What would it take for us to produce and publish live win odds?  It really came down to three things:

• A database of previous games to use as a sample population (way more challenging than expected)
• An estimate of the value of individual plays to enable forecasting score and time (moderately difficult)
• A wordpress account (easy)

UPDATE: In 2021, I launched LacrosseReference PRO, which is like KenPom for lacrosse. One of the products is the GameGuide, which gives you the insights you need for all upcoming games, men’s or women’s.

Anyway, if we could get those three things, then after every play in a game, you’d be able to update the time and expected score to get to a current win probability for each team.  Fortunately, for most games, all that data is available online.  And here’s what it looks like for that Maryland vs Syracuse tournament game.

So when Maryland scored that goal to go up 10-6, our model would have told me that the odds of them pulling out the eventual victory was 90.8%.  In fact, the worst odds they had the entire game were 45.3% when Syracuse scored the first goal of the game 3 minutes in.  We can talk all day about how this would have affected my watching experience; I don’t think that I would have been any more confident, even with these numbers.  I lived through Maryland-Duke basketball – 10 points in 54 seconds after all.  I do think that this model will give people watching a given game a greater appreciation for great plays, especially those that have a large effect on the win probabilities. 

Anyway, using this model, our intention is to live log a selection of games so that as you are watching them, the win odds chart will update automatically and in real-time.  In addition, for any game that we have the play logs for, we can go back and figure out what the win odds chart looks like.  I’d love to quantify which teams had the statistically most improbably comebacks (or those that stole defeat from the jaws of victory).  

Hopefully this adds some tangential entertainment value for those watching live games.  I know it will for us.  And it’s a good introduction into what we hope is a burgeoning lacrosse analytics community.