Lacrosse Analytics

[laxreference]

As always, here is the daily digest for Feb 15, 2020.

[laxreference]

It's been an offensive-minded start to the season, and there are some gaudy stat lines on this week's list of the top individual statistical performances. But I was honestly surprised to see that scoring wasn't up that much since last year. It's early, so take this with a grain of salt, but as far as goals per game:

2015- 20.5
2016- 20.2
2017- 20.7
2018 - 20.6
2019 - 22.9
2020 - 23.6

Not that big of a jump. Will be curious to see how it changes over the season.

[laxreference]

The top movers over the weekend in terms of NCAA tournament probability were:

Cornell +21.2% (now 64.8%)
Robert Morris +19.5% (47.4%)
Virginia +12.8% (79.9%)
Army +11.7% (54.1%)
Yale +11.1% (79.6%)
Duke +10.9% (69.0%)
Loyola +10.0% (61.8%)

Note: this assumes the committee relies heavily on RPI as a criterion again.

[Farfromgeneva]

It's been an offensive-minded start to the season, and there are some gaudy stat lines on this week's list of the top individual statistical performances. But I was honestly surprised to see that scoring wasn't up that much since last year. It's early, so take this with a grain of salt, but as far as goals per game:

2015- 20.5
2016- 20.2
2017- 20.7
2018 - 20.6
2019 - 22.9
2020 - 23.6

Not that big of a jump. Will be curious to see how it changes over the season.

Min side there’s a primer on the model available that I ultimately need to read but could you Cliff note what drives the scores a little? For example, and I really couldn’t care less but to the eyeball it seems odd: ToMo kid Smith was 8,2 and top the model against Canisius but he took 22 shots so good (above 1/3) but not astounding shot % compared with my guy Jason Knox who was 10,1 on 12 shots (83%) and he has 3gbs, 1TO, 1CT so arguablly a strong overall contribution and the scores have a healthy gap. I really don’t mean to make it “my guy should be higher” just where my domain knowledge is strongest but on the surface I didn’t see smith being more efficient or having other meaningfully superior non scoring contributions. I’m guessing it has something to do with when the scoring game in the game and outcome overall but it still struck me as odd that a guy who takes 22 shots, even with scoring 8 goals would be #1 in offensive efficienty

[ICGrad]

Note: this assumes the committee relies heavily on RPI as a criterion again.

I was under the impression that the committee relied on whichever criteria guarantees a barely-.500 Hopkins a shot in the tourney; i.e., last year it was quality wins, though when someone pointed out Cornell had more of those it became top 10 quality wins. If they had had a better RPI that would have been the criteria.

[laxreference]

Min side there’s a primer on the model available that I ultimately need to read but could you Cliff note what drives the scores a little? For example, and I really couldn’t care less but to the eyeball it seems odd: ToMo kid Smith was 8,2 and top the model against Canisius but he took 22 shots so good (above 1/3) but not astounding shot % compared with my guy Jason Knox who was 10,1 on 12 shots (83%) and he has 3gbs, 1TO, 1CT so arguablly a strong overall contribution and the scores have a healthy gap. I really don’t mean to make it “my guy should be higher” just where my domain knowledge is strongest but on the surface I didn’t see smith being more efficient or having other meaningfully superior non scoring contributions. I’m guessing it has something to do with when the scoring game in the game and outcome overall but it still struck me as odd that a guy who takes 22 shots, even with scoring 8 goals would be #1 in offensive efficienty

Great questions. Always happy to describe the philosophy behind the numbers. Your comments had two main themes. First, Knox had more points and a strong over stat line with the GBs and CT. Second, Knox was more efficient in getting 10 goals on 12 shots.

The first one is the easiest to explain, and is the main cause of the discrepancy. Here is the key stat: Knox's 10 goals were split between 8 assisted and 2 unassisted; Smith's 8 goals were 7 unassisted and 1 assisted. And the reason that's a big deal is that the model splits the "value" of a goal between the guy who got the assist and the guy who scored the goal. For unassisted goals, the goal scorer gets the whole "value" added to his total.

Now, you could argue that an offense with a lot of assists is better than an offense where one guy goes and gets a bunch of goals through hero-ball. And while I don't have an empirical argument one way or the other, since this is an individual ranking list, unassisted goals get you higher on the list than assisted goals. Is that fair to someone like Knox? Who knows.

Second factor is the efficiency of the performance. Knox was undoubtedly more efficient. And Hobart has been more efficient than ROMO overall this year (1st in efficiency vs 11th for RoMO). But the model was created was to look at how often different type of plays lead to goals. By that logic, missed shots have a positive value because they are followed by goals for the offense more often than they are followed by goals by the defense. As a result, a missed shot increases a player's EGA. (It should be pointed out that the play values include pre-shot clock games, so eventually, when we can use just post-shot-clock games, that may no longer be the case).

I don't disagree that a more efficient performance is desirable to a less efficient performance. But for me, the benefits of being able to use the "play-values" to aggregate a player's contribution is worth the downside of favoring a player with a high-volume, low-efficiency performance.

Hope that helps. What do you think?

[laxreference]

Note: this assumes the committee relies heavily on RPI as a criterion again.

I was under the impression that the committee relied on whichever criteria guarantees a barely-.500 Hopkins a shot in the tourney; i.e., last year it was quality wins, though when someone pointed out Cornell had more of those it became top 10 quality wins. If they had had a better RPI that would have been the criteria.

This thread has a strict no-conspiracy-theories rule. Ok, not that strict, but I'll leave these discussions to the more partisan threads.

[laxreference]

Updated list of Live Win Probability Links

- Detroit vs Bellarmine

[laxreference]

Updated list of Live Win Probability Links

- Brown vs Sacred Heart
- Detroit vs Bellarmine

[laxreference]

Updated list of Live Win Probability Links

- Marist vs Army
- Fairfield vs Merrimack
- Bryant vs Boston U
- Brown vs Sacred Heart
- Detroit vs Bellarmine

[laxreference]

Updated list of Live Win Probability Links

- Saint Joseph's vs Delaware
- Michigan vs Canisius
- Holy Cross vs UMass-Lowell
- Villanova vs Maryland
- Richmond vs Navy
- Marist vs Army
- Fairfield vs Merrimack
- Bryant vs Boston U
- Brown vs Sacred Heart
- Detroit vs Bellarmine

[Dip&Dunk]

I guess I thought probability of win would be more predictive. This looks like, to give a nautical analogy, steering by your wake.

[laxreference]

Updated list of Live Win Probability Links

- Princeton vs Colgate
- Saint Joseph's vs Delaware
- Michigan vs Canisius
- Holy Cross vs UMass-Lowell
- Villanova vs Maryland
- Richmond vs Navy
- Marist vs Army
- Fairfield vs Merrimack
- Bryant vs Boston U
- Brown vs Sacred Heart
- Detroit vs Bellarmine

[laxreference]

I guess I thought probability of win would be more predictive. This looks like, to give a nautical analogy, steering by your wake.

Haha, well I look at it this way. Would you be shocked if a .263 hitter and a .312 hitter got back to back base hits? That's the statistical equivalent of both Marist and Villanova winning today. Based on the probabilities, there was an 8% chance of that happening since MD had a 74% chance of winning and Army's was 69%.

[laxreference]

As always, here is the daily digest for Feb 18, 2020.

[laxreference]

I did an analysis of the relationship between turnover rates and saved shots. I wanted to see if there was any useful information to be gleaned by looking at possessions and ending one of three ways:

1. Goal
2. TO
3. Saved/Missed shot

I feel like you hear about turnover rates and you definitely hear about offensive efficiency, but this third outcome is somewhat less analyzed. Turnovers are usually described as an error from the offense, while saves are typically thought of as an accomplishment by the goalie.

But an offense that minimized turnovers by gifting easy saves because of poor shots wouldn't have an obvious flaw (i.e. a high turnover rate), yet might be a horrendous offense. I wanted to dig in and see if the numbers help describe the trade-offs involved.

[laxreference]

Updated list of Live Win Probability Links

- Lehigh vs VMI

[laxreference]

Updated list of Live Win Probability Links

- Towson vs Cornell

[laxreference]

Here are the statistical previews for all of today's games. Lots of tournament implications today.

[joewillie78]

The top movers over the weekend in terms of NCAA tournament probability were:

Cornell +21.2% (now 64.8%)
Robert Morris +19.5% (47.4%)
Virginia +12.8% (79.9%)
Army +11.7% (54.1%)
Yale +11.1% (79.6%)
Duke +10.9% (69.0%)
Loyola +10.0% (61.8%)

Note: this assumes the committee relies heavily on RPI as a criterion again.

Although as a Big Red fan, I like seeing that our chances went up 21%, but we kept seeing this last year and I believe after we beat ND on the road, our chances were like 80%, and we all know how that ended, so like I said, Im pleased our chances went up, but until we either win the AUTOMATIC bid, or when the bids come out that we got and at-large, I will continue to be of the mindset that the NCAA will find a way to screw us like they did last year.

GOBIGRED
Joewillie78