Lacrosse Analytics

[laxreference]

I tabulated the cumulative probability of each team in each conference getting a bid to the NCAA tournament. ACC leads the way with 2.7 expected bids. Ivy nipping on their heels. Obviously, this is greatly affected by the number of at-large bids truly up for grabs after the conference tournaments, so I would think about this as a combination of conference strength and the way that schedules are expected to influence RPI.

ACC: 2.7
Ivy: 2.5
Big Ten: 2.3
Patriot: 1.7
Big East: 1.6
CAA: 1.5
NEC: 1.3
SoCon: 1.2
Amer East: 1.1
MAAC: 1.1

[Laxfandad]

Suggest looking at http://mylaxrankings.com/ which is based off MyHockeyRankings.

FACT: With 75 teams, each team is not capable of playing everyone.
FACT: Teams (call them Old Boys Club) are not open to adding weaker teams to their schedule.
FACT: The teams that hold this power dictate on their terms the when/where a weaker team will play them (often on short rest and at their facility)
I know this from actual discussions with D1 Coaches

This analytics formula uses a schedule/strength of schedule/rpi type value BUT also incorporates the addition of a Goal Differential factor. This in its pure form is the summary of your goal differential in games (+s and -s) then divided by your game total. Usually, no rankings come out until each team has played a minimum of 5 games. Now I know many of you reading are jumping to the "run up the score" on weaker teams. The formula tenders the Goal For Differential (I am using this from Ontario Lacrosse Association box lacrosse) where 10 goal deltas are reduced to just 7 goals.

Hypothesis: What IF a Hampton Lax team beat everyone they played and by just 1 goal in each game. The SOS would have a GFD added and they would probably rise above a team that has a strong RPI but say went 8-7 like a Hopkins in 2019. Rightly so, shouldn't they be given an "At-Large" bid to the tourney. They should. Their strength of schedule is not there, but they would be reward by a +Goal Differential added to their weaker schedule value. I've done this math on teams from 2019.

Penn State last year would be the top team.
High Point would have finished much higher than Hopkins (rightly so). This formula negates the idea of a bad loss. Their 1 goal loss to St John's was one simple missed play. One shot on goal. Arguably, one could have said that Hopkins had 3 or 4 Bad losses...but it didn't play out that way.

The beauty of this system is that it rewards wins, scoring and good defense. If a loss occurs and it is by just 1 goal or 2, this doesn't completely destroy a season.

In OLA Box Lacrosse, the goal is to invite the top 16 teams to the Ontario Gold Medal playoffs where they split into two pools of 8. Each team plays each other. Then the top 3 from each pool play for the Gold, Silver and Bronze medals in the province. In Canada, a Province is like a US State.

I suggest that someone test out the math. Automatic Conference Qualifiers are the Red Herrings in this process. Answer is to expand the tourney with more play-in games. I'd like to see up to 20 (4 play in games).

[notentitled]

Are you trying to make John's Hopkins fans aware of a fact they already know? LOL

The idea of math system is welcome. But ...

[laxreference]

Suggest looking at http://mylaxrankings.com/

This analytics formula uses a schedule/strength of schedule/rpi type value BUT also incorporates the addition of a Goal Differential factor.

So, it sounds like this system discounts win-loss records a bit and increases the weight of goal differential? Do I have that right?

[wgdsr]

I tabulated the cumulative probability of each team in each conference getting a bid to the NCAA tournament. ACC leads the way with 2.7 expected bids. Ivy nipping on their heels. Obviously, this is greatly affected by the number of at-large bids truly up for grabs after the conference tournaments, so I would think about this as a combination of conference strength and the way that schedules are expected to influence RPI.
ACC: 2.7
Ivy: 2.5
Big Ten: 2.3
Patriot: 1.7
Big East: 1.6
CAA: 1.5
NEC: 1.3
SoCon: 1.2
Amer East: 1.1
MAAC: 1.1

no offense, your math is way off. acc i'd be shocked if they had more than one year where they had only 3 teams in since they've expanded past 4 teams. i'd guess most years the number is 4, with the occasional 5 spot (though 5 with the b1g around not likely).
take the over on 2.7.

[Laxfandad]

Suggest looking at http://mylaxrankings.com/

This analytics formula uses a schedule/strength of schedule/rpi type value BUT also incorporates the addition of a Goal Differential factor.

So, it sounds like this system discounts win-loss records a bit and increases the weight of goal differential? Do I have that right?

The system doesn’t discount win-loss records. Wins will contribute to +s and losses to -ves in the Goal For Didferential. If that is an overall +ve number then a team’s RPI/SOS will get a boost, conversely a -ve will lower an RPI/SOS.

In theory, if a team rolls through their weaker RPI/SOS the increase from their GFD could raise them above a mediocre Big10 or BigEast team with a stronger schedule. The key is to apply a proper scaling factor so all teams GFD do not dominate the RPI/SOS. RPI is a number like 0.613 Penn State’s GFD last year was something crazy like 1.3 (2x to the RPI). You want the GFD to be scaled in proportion to the RPI. Something like GFD divided by 3 or 4.

[laxreference]

Ok. Need some help. I want to do some analysis on scheduling hypotheticals. For example, Hopkins' schedule this year includes Navy, but drops Virginia. I'm curious about the potential RPI effects and I'd like to see how much of a difference scheduling changes like this make.

So the question to the group: what are the scheduling changes that your team made that you think might have a positive or negative effect on RPI? What are the games you wish your team had on the schedule that they don't, and which non-conf game would you remove?

[laxreference]

Pretty sweet win probability chart for the 1st game of the year. Hopefully a harbinger of things to come...

IMG_20200201_134147_217.jpg (28.65 KiB) Viewed 3786 times

[laxreference]

Here is the list of Live Win Probability Links

- Duke vs Air Force
- Rutgers vs Quinnipiac
- Boston U vs Providence
- Maryland vs High Point
- Penn State vs Lafayette
- Jacksonville vs Detroit
- North Carolina vs Colgate
- Cleveland State vs Marquette

[laxreference]

Here's the daily digest for yesterday's game. Highlighted the 3 best games based on a combination of excitement, importance, and degree-of-comeback.

[laxreference]

For 2020, I'm going to show the RPI table with each of the 3 RPI components listed out separately. That will give you the ability to see how a team's RPI is constructed of W/L Record, Opponent's W/L Record and Opponent's Opponent's W/L Record. I will be curious to see, once the season progresses far enough, whether there are any stark differences in how various team's RPIs are made up.

Here is the 2019 RPI rankings , updated with the new columns.

[old salt]

Ok. Need some help. I want to do some analysis on scheduling hypotheticals. For example, Hopkins' schedule this year includes Navy, but drops Virginia. I'm curious about the potential RPI effects and I'd like to see how much of a difference scheduling changes like this make.

So the question to the group: what are the scheduling changes that your team made that you think might have a positive or negative effect on RPI? What are the games you wish your team had on the schedule that they don't, and which non-conf game would you remove?

For Navy : 2020 vs 2019

Retained MD (hopefully permanently) & entire PL
JHU returns (hopefully permanently) / Syracuse dropped
Richmond added / Princeton dropped
Furman added / Vermont dropped
Manhattan added / UMBC dropped

Probably a slight drop in SoS, but worth it.
MD, JHU & Furman for sentimental & attendance reasons. Likely 2 QW opportunities.
Richmond interesting nearby addition, might be a QW opportunity.
Manhattan (home game) good opener for new staff.

I like it & hope we can retain MD AND JHU every year, going forward.
Switch out the other 3 NC slots for variety.
Would like to get Air Force 2 of every 4 years, giving every Cadet & Mid a home & home game.
As VMI continues to improve, would like to get them back in the mix (home game only)

[laxreference]

Nice to have the weekly top performances list back in action. Here is the full write-up of the week 1 stars, per EGA.

I think it's safe to say that Penn State is picking up where they left off. And generally, given some of the gaudy scores we saw, it's not surprising that this list has a lot of guys in the teens who might normally show up in the top-5 with similar stat lines in future weeks.

Here is a quick infographic of the top-20. Post itself has the details.

PlayerGame_visualization(20200203).png (56.87 KiB) Viewed 3673 times

[Dip&Dunk]

So I could understand this better, could you explain how AF's Q. Peene's 4G 1A in a win over Duke would have less value than say J. Winston's 4G 0A for JU in a loss to Detroit Mercy?

[NElaxtalent]

Would like to get Air Force 2 of every 4 years, giving every Cadet & Mid a home & home game.

Fully agree on your AF suggestion, at a minimum.

However, I'd love to see an annual CiC style round robin. Makes too much sense. Great experience for all . . . build mutual respect.

[laxreference]

So I could understand this better, could you explain how AF's Q. Peene's 4G 1A in a win over Duke would have less value than say J. Winston's 4G 0A for JU in a loss to Detroit Mercy?

Without digging into the stat lines in more detail, I'll tell you the main reason that two stat lines can look similar, but have different EGA values, is assisted vs unassisted goals. An assisted goal's EGA value is split half-and-half between the assister and the goalscorer. Might that be it?

[NElaxtalent]

So I could understand this better, could you explain how AF's Q. Peene's 4G 1A in a win over Duke would have less value than say J. Winston's 4G 0A for JU in a loss to Detroit Mercy?

Without digging into the stat lines in more detail, I'll tell you the main reason that two stat lines can look similar, but have different EGA values, is assisted vs unassisted goals. An assisted goal's EGA value is split half-and-half between the assister and the goalscorer. Might that be it?

Believe I read other stats like GBs, TOs, FOs factor in a slight +/- adjustment.

[laxreference]

Yep, that's exactly right. Things that tend to lead to goals have positive EGA. Things that generally lead to opponent goals have negative EGA. Each player's EGA value is the sum of all the contributions noted in the game play-by-play logs.

[laxreference]

Win probabilities are live for the Ohio State vs Detroit game. Good one early on.

[laxreference]

I dug a bit farther into the NCAA tournament probabilities from the simulations that I ran earlier this year. I had shared the average number of bids for each conference from across the simulation runs, but I was curious about whether they range of outcomes was narrow or wide. After all, the same average value can come from many different shaped distributions.

I was surprised to see just how wide the spreads were. The Big Ten for example, earned anywhere from a single bid to having 5 teams make it. And that is with all the same schedules and starting points. The actual probabilities were:

1 bid: 20% of all simulations
2 bids: 42%
3 bids: 30%
4 bids: 8%
5 bids: 0.5%

The distributions for the rest of the conferences are in the post itself.

I was surprised that there were so many simulations where a top conference gets a single bid, but I suspect it has to do with the fact that the strength of the conference plays such a role in RPI. If one or two teams isn't what we thought they were, then it's possible for the whole conference's at-large chances to suffer. Anything can happen I suppose.