PNW district difficulty projections

[SoftwareBug2.0]

I thought it would be interesting to guess how competitive the different Pacific northwest district events would be. To do this I used teams 2013 winning percentages. In the first column below is listed the 24th best winning percentage of teams signed up for that event. In the second column is the average winning percentage.

(0.227,0.385,Eastern Washington University)
(0.333,0.398,Central Washington University)
(0.333,0.429,Oregon State University)
(0.333,0.434,Auburn Mountainview)
(0.333,0.463,Mt. Vernon)
(0.347,0.441,Shorewood)
(0.375,0.453,Glacier Peak)
(0.416,0.476,Auburn)
(0.444,0.498,Wilsonville)
(0.450,0.496,Oregon City)

For reference, the worldwide average winning percentage is 0.452. The average winning percentage for a team that attended a northwest event last year was 0.468. If you're wondering why the average is below .5 it's because the teams that play more matches tend to win more.

It would probably be more scientific to do this comparison with OPR. This is left as an excercise for the reader.

So how are you feeling about the events you're going to? My team's first two events are Oregon City and Wilsonville, so that ought to be interesting.

[julianpowell]

We will be at Glacier Peak, Mt Vernon, and Shorewood. I'm really excited about the district system, I think teams will benefit a lot from the smaller event sizes and greater number of matches. And the level of competition at the championship should be awesome.

[Navid Shafa]

This is really cool! I love geeking out over stats.

Originally Posted by SoftwareBug2.0

In the first column below is listed the 24th best winning percentage

Any reason for 24th in specific, or are you just trying to point out the variation in the data? I think this might be a tiny bit misleading, because this number is likely to be higher at events with more registered teams.

Originally Posted by SoftwareBug2.0

For reference, the worldwide average winning percentage is 0.452. The average winning percentage for a team that attended a northwest event last year was 0.468. If you're wondering why the average is below .5 it's because the teams that play more matches tend to win more.

This logic seemed flawed at first. I had to think about it a while, but It makes since that the average is less than half. I'm not sure which plays a bigger factor though, teams who play more, or a lower average due to the fact that teams who win a large percentage of their matches are a significant minority.

Originally Posted by SoftwareBug2.0

(0.227,0.385,Eastern Washington University)
(0.416,0.476,Auburn)

Although I think Auburn has a slightly deeper roster, there are some teams I'm going to watch out for when we are at Eastern. I like the fact that the overlap isn't too high, I enjoy getting to play more matches with more teams!

Originally Posted by SoftwareBug2.0

It would probably be more scientific to do this comparison with OPR. This is left as an excercise for the reader.

Now you have me wanting to try to predict district standings based on all previous year's performance in our region. I don't think I'll use OPR, because the model changes a fair amount each year. This might unintentionally weight some years higher than others... PM me, perhaps we can work something out together

Hope we get to play with you again this year at some point. See you at district champs! *Crosses Fingers*

[Jeffrafa]

Originally Posted by Navid Shafa

Any reason for 24th in specific, or are you just trying to point out the variation in the data? I think this might be a tiny bit misleading, because this number is likely to be higher at events with more registered teams.

Knowing Eric, the logic is simple - by looking at the 24th best team, you are looking at the performance of the worst team to make it into finals, if alliance selections were to perfectly follow robot performance/abilities.

It is a way of seeing how deep the field is of competitive robots - if the 24th best robot is pretty good, then there will be a sufficient number of good teams to fill all the alliances, and the last few picks may still be pretty good robots - if the 24th best is pretty poor compared to the average, then not all alliances will get a good 3rd pick, which may shift the advantage to lower alliances which get 3rd pick earlier in the selection. Of course this ignores factors like robot specific abilities, and forming alliances off complimentary characteristics - but its a simple way, given limited information, to try and see how good the last picked robots are.

It is true that the 24th best would likely be better at events with more teams, but part of the intent with districts is to make sure all the events have the same (or very similar) number of teams - if one event is significantly smaller than the others, then those teams have an unfair advantage in the point system - they are more likely to rank high, win awards, and end up on alliances than at an otherwise equal larger district event - this is the primary driver for offering a 3rd play, to make sure all the events fill up and the odds are similar. This year there is a slight advantage for teams attending Mt. Vernon, or Oregon State University (as a 1st or 2nd event), otherwise the team count is pretty close across PNW district events

For reference, here's the current team count along with Eric's data:

31 (0.227,0.385,Eastern Washington University)
35 (0.333,0.398,Central Washington University)
29 (0.333,0.429,Oregon State University)
32 (0.333,0.434,Auburn Mountainview)
28 (0.333,0.463,Mt. Vernon)
33 (0.347,0.441,Shorewood)
32 (0.375,0.453,Glacier Peak)
35 (0.416,0.476,Auburn)
32 (0.444,0.498,Wilsonville)
35 (0.450,0.496,Oregon City)

[Navid Shafa]

Makes sense