Fun with Stats

Submitted by Sebastien Belanger on Wed, 2017-11-08 15:17

Forums:

For those of you new to Parity and enjoy statistics, you should know that all of the game stats are available here: https://docs.google.com/spreadsheets/d/1F46H8ZRGP8Jzj1zSW0PT8HerBZ_BlHI5...

As someone who enjoys data, I get way too much fun out of crunching and re-crunching the numbers. Will you find funny annecdotes on game play or philisophical observations on the point of the league by playing with numbers? Absolutely not but you will find interesting metrics that aren't immediately apparent when looking at the "official" individual/team salary figures here: https://parity-server.herokuapp.com/

New this year is a "Normalized Salary" which seems to be equal to "Salary/Points Played" multiplied by 15. Not sure why 15 was chosen but this makes a lot of sense. I did a quick rank of players by "Salary" and compared them to a rank by "Normalized Salary" and found some pretty significant differences (below).

Top 5 Players Undervalued this Week (Salary_Rank - Normalized Salary_Rank)

Name & Salary Rank Delta 

Kevin Hughes  -18

An Tran           -14

Jon Rowe         -14

Liam Parker     -13

Josee Guibord -13

 

Top 5 Players Overvalued this Week (Salary_Rank - Normalized Salary_Rank)

Name & Salary Rank Delta 

Hope Celani    18

Tim Kealey      16

Jaime Boss      16

Michael Wong 13

Simon Berry    13

 

What does this meant? Probably that if you play more, while you might get more points but you probably also get tired which would make you less efficient. Josee shows up on the first list in position #5 but also in the second list in position #10 so while it's a fun analysis, it obviously has its flaws.

Simon Berry

Wed, 2017-11-08 18:15

Good to know I am an over rated player...

Chris Sullivan

Wed, 2017-11-08 19:21

Seb,

It is impossible for Josee to show up on both lists, they are the inverse of each other. (perhaps you are comparing Josee with substitute Josee?)

Normalized salary is just salary/point multiplied by a constant (15) to reflect a normal amount of points played. "Salary" is just gross salary and gives players who play more points an obvious advantage.

If Player A and Player B play exactly as well as each other and Player A plays twice as many points, they will have the same salary/point (and the same normalized salary), but Player A will have twice as much gross salary.

Your lists should just be a list of the players who played the most points and least points. All the players in the first list played in our game, which only had 31 points played. This is why salary/point is a better measure.

However, I prefer to choose whatever number reflects better on myself in the moment and go with that. Another strategy is to multiply all salaries by 1000000 and be filled with self-awe. It's kind of like what happened to pinball scoring in the 80s.

Simon, you are not over rated, unless somewhat rated you better than awesome.

#mathgeeksrule

Kevin Hughes's picture

Kevin Hughes

Thu, 2017-11-09 11:12

Should I add the number of points played to the app or the spreadsheets? Seems like people might want to see this data

Sebastien Belanger

Thu, 2017-11-09 11:15

Absolutely! You can do it in your spare time ;-)

Christopher Keates

Thu, 2017-11-09 11:41

The sheet needs to have some useful views and stuff added to it still, but the game 1 dump has points played.

Link:

https://docs.google.com/spreadsheets/d/1F46H8ZRGP8Jzj1zSW0PT8HerBZ_BlHI5...

App

John Haig

Thu, 2017-11-09 11:30

Math is fun an all, but if we're going to apply this and form a team's strategy, wouldn't it make sense to chuck it more? 

If a player's salary is essentally their salary per point,  a strategic GM might want a team that scores efficently in terms of salary "earned" in order to avoid inflating their players' salaries and forcing trades.  A great way to achieve this goal would be to minize the number of passes required per goal.  This may lead to more goals per game as two pass points tend to be quick, but in the end it should lead to more efficiency in terms of salary dollars generated per game, as players' salaries are directly related to salary per point, not per game.

The lesson here is that in order to help my team succeed without being broken up due to high salaries, is that I should be chucking more, right?

#chuckearlychuckoften  #joggingforbombs

Christopher Keates

Thu, 2017-11-09 11:44

> If a player's salary is essentally their salary per point,  a strategic GM might want a team that scores efficently in terms of salary "earned" in order to avoid inflating their players' salaries and forcing trades.

Some of the more successful GMs over the course of Parity have actually noticed this and built strategies around it. Pew pew pew! Some of the salary changes are intended to minimize this as much as possible (even though good frisbee should still score in as few passes as necessary).

Also, every team is going to get blown up repeatedly (I hope).

Mehmet Karman

Thu, 2017-11-09 16:21

He should have been a GM!  Chucking in an effort to score efficiently can also have its downsides - i.e. widespread mocking on the throwaway leaderboard - much depends on how talented one is at chucking...

John Haig

Fri, 2017-11-10 08:11

Don't let them get to you.  They simply aren't on your strategic level.  They probably went to McGill, or *shudder* Carleton.  The joke will be on them in a week or two when those salaries have ballooned from all of those needless dumps and swings.

Frederic Caron

Fri, 2017-11-10 08:44

They will heckle you if you don't throw it anyway. The best way to get around both issues is to throw the hammer! :p

Amos Lee

Wed, 2017-11-15 10:19

Some of the least successful GMs have also built strategies around it. But when you see the look on 11 of your teammates' faces when you ask them to play zone indoor or to half-court press the other team, you realize you didn't want to be THAT much of a villain.

Alex Bush

Thu, 2017-11-09 17:09

Top 5 "undervalued players" and their points played distribution :

Kevin Hughes - 15 O, 1 D 94% O

An/Jon - 5 O, 12 D 29% O 

Liam Parker - 6O, 11 D 35% O

Josee Guibord - 12 O, 5 D 71% O

*note: Jon, An, and Liam were on Sultimate Fight Club which won 21 - 10. This blow out means they were on D more often.

Bottom 5 "overvalued" and point spread:

Hope Celani - 10 O 15 D 40% O

Tim Kealey - 9 O, 15 D 38% O

Jaime Boss - 15 O, 13 D 54% O

Michael Wong - 13 O, 9 D 59% O

Simon Berry - 10 O, 15 D 40% O

 

I think all you've really stated is that "Given possession of the disc, people are going to accrue salary." Kevin played all but 1 point on offense, so he very likely touched it on all those points. An, Jon, and Liam played a blow-out 21-10, so I also can say with certainty that they had opportunity to accrue salary.

 

Tim, Simon, and Hope played less than 40% of their points on defense, meaning they had a much lower chance of seeing the disc. It's not surprising that they had fewer opportunities to accrue salary.

We know that in ultimate, offense has all the advantages and that there are defensive points played where you never get a possession. Those defensive points where offense was perfect should not count in "salary per point played" for players since they did not have any opportunity to accrue salary.

Statistics can tell you a lot of information, provided that they are meaningful. Comparing someone's salary rank with their salary per point rank does not "overvalue" or "undervalue" players. 

 

John Haig

Fri, 2017-11-10 08:06

Those defensive points where offense was perfect should not count in "salary per point played" for player since they did not have any opportunity to accrue salary.

So am I not supposed to try and get blocks during me D points?  Am I just suposed to wait for Mehmet to throw it out the back again?  Failure to accrue salary doesn't mean lack of opportunity.
 

Frederic Caron

Fri, 2017-11-10 08:53

If you want a really high salary and climb the D leaderboard get every D you can. But if you want to be valued and not a burden to your GM, only D what you have to. I have let disc go by me in the past after beating someone to a spot or whatever. Still a turn, just no salary increase.  

Kevin Hughes's picture

Kevin Hughes

Fri, 2017-11-10 15:21

I am open to suggestions about how to account for this. We know how many O and D points each player played. Splitting salary into O and D buckets is difficult but probably doable. If someone can prove this is a valuable way to do it I'll look into getting it into production.

Jim Robinson

Fri, 2017-11-10 16:53

Do we know how big of a problem this is?  What percent of points do not have a turnover?  Once there is a single turnover should we now consider this a point played since there was an opportunity to accrue value? 

Christopher Keates

Fri, 2017-11-10 20:41

Over 24 games players get close to 400 points played and trend towards a 50/50 split. It's not a big deal.

One game looks weird because sample size is small. That won't last.