How do you rank?

[allen0977]

Team 1251, the TechTigers, have put together a "Relative Power Ranking" system that analyzes scores and records of teams competing at various regionals and ranks teams relative to each others performance. Come check out how your team ranks among other teams participating in your local regionals, across multiple regionals, and those attending nationals.

We pulled team names from each team's website, and tried to be as accurate and current as possible. Please let us know if your team goes by a different name so we can fix it in our database.

We could only compile data from the 19 regionals that contained data on the US FIRST website, if anyone has data from regionals that we are missing, it would be greatly appreciated.

http://1251.techtigers.com/PowerRank/FirstRPI

Thanks,
-Allen

[Raul]

It looks like your power ranking is very close to being based on winning percentage. But since there are teams with higher percentages ranked lower, I am curious what formula that includes scores you used to figure the ranking.

[allen0977]

Originally Posted by Raul

It looks like your power ranking is very close to being based on winning percentage. But since there are teams with higher percentages ranked lower, I am curious what formula that includes scores you used to figure the ranking.

We determine a strength of schedule by looking at each teams opponents winning percentage and also those opponents opponents winning percentage ..

(3*(winningPercent) + 2*(OpponentsWinPercent)+1*(OpponentsOpponentsWinPe rcent))/6

that helps weight your score, and then we look at margin of victory in individual games to give more/less points which could move a team up/down the ranking. Everything changes with each new match entered, even if it is a team you never faced (as long as one of your opponents faced any team in the new match)

[Kris Verdeyen]

Originally Posted by allen0977

We determine a strength of schedule by looking at each teams opponents winning percentage and also those opponents opponents winning percentage ..

(3*(winningPercent) + 2*(OpponentsWinPercent)+1*(OpponentsOpponentsWinPe rcent))/6

that helps weight your score, and then we look at margin of victory in individual games to give more/less points which could move a team up/down the ranking. Everything changes with each new match entered, even if it is a team you never faced (as long as one of your opponents faced any team in the new match)

You could subtract out alliance partners by doing:
(3*(winningPercent) + 2*(OpponentsWinPercent - PartnersWinPercent)+1*(OpponentsOpponentsWinPercen t-PartnersPartnersWinPercent - OpponentsPartnersWinPercent + PartnersOpponentsWinPercent))/6

... at least I think that'd do it.

I'm not sure how you'd be able to put in the relative strengths of the regionals. Average total match score might do it.

Anyway, great job.

[spears312]

Awesome software. It's kinda cool to look at the rankings of all the teams at nationals in some sense. There is just a little typo with heatwave's rank! (j/k). Another piece of great work from you guys, hope I get the chance to talk to you all again at nationals.

[Kims Robot]

Great Data!!

If you are looking for some results for FLRC, partial UNOFFICIAL results are here: http://www.chiefdelphi.com/forums/sh...ad.php?t=35742

I have also attached our match history as recreated by Doug.

[Corey Balint]

All weekend I had been talking to 1251 and just discussing the game and their strategy. Come saturday i asked them(since they were in first) how their scouting was coming along. They said greatly and that they had a "system" set up. Me been nosey, wanted to know more. They kept it secret until we ended up aligning with them (we ranked a 100 at Palmetto, 5 points higher then any other, and 3 spots ahead of 1251 themselves-had to brag). I was amazed. I had always wanted to set something like this up, but my lack of intelligence seems to hinder that progress. Props to 1251 on a great system.

[Barry Bonzack]

::sniffs arms:: hey, I don't rank!


This is a very cool system. I was wondering when someone would come out with a formula that calculated SOS. I decided to take this site for a test run and used it in picking my top Fantasy First teams... along with a mix of teams that I saw at UCF and Lone Star Regionals and gut feeling. I like the cool color code to the left of the teams names that says what division the teams are in for the Championship, that was useful as well. I'm pretty proud of my team's rank by this system. Very cool guys.

Now I wonder when someone will rank teams judging by their community service and public outreach programs. :-D

[dmurdz]

If this is to be used to rank the teams from the best one that won the most it would be great.

But I believe that ypu need to consider other factors when choosing who the BEST all around team, factors that would take hours and hours to find and calculate.

But still Good Job!

[David Guzman]

Originally Posted by dmurdz

If this is to be used to rank the teams from the best one that won the most it would be great.

But I believe that ypu need to consider other factors when choosing who the BEST all around team, factors that would take hours and hours to find and calculate.

But still Good Job!

Did You read the first post???

It is not only based on winnings and loses it also takes into acount your opponents and your partners, if you look at the ranking you can see that teams that have less wins then another team can still rank higher then teams with more wins.

Dave

[ChuckDickerson]

First, I would like to say thank you to Team 1251 for a job well done! The effort put into this ranking system is evident and the outcome is a very nice product useful to all teams. I have never been a big fan of the W-L-T/RP system that FIRST uses at the competitions especially with the RP based on the losing score/lowest un-penalized score. I have never really felt that it has been effective at accomplishing the objective. The FIRST system does not really seem to do a good job at ranking a team based on the strengths and weaknesses of the teams that you randomly play with and against. The system used in Team 1251’s relative power ranking does at least make an attempt at factoring in these strengths and weaknesses even if there is minor data entry errors discovered in team information. It is unfortunate that the raw data is unavailable from some of the early regional competitions. Without that we cannot really paint a complete picture. The FIRST W-L-T/RP system, however, is fairly straightforward and easy to compute rapidly at competitions. With this in mind I have a few thoughts, questions, and comments.

1) I am unclear as to if this relative power ranking system is (or can be) computed in near real time or is it something that takes quite a while to compute due to the interdependence of the power rankings of the teams at the competition. If a team was to supply corrected information for their particular team or even accurate data for a whole regional is it a big deal to update/compute the rankings after the data is entered? Once the “compute ranking” button is pressed are the results fairly immediate or is it more of a go get coffee come back and they are ready or more of a start it tonight before bed and they are ready in the morning kind of thing? Is this a system that could be used instead of the real time FIRST W-L-T/RP system for ranking teams at competition? Is it even appropriate?

2) How is the rank determined between teams with equal relative power rankings determined?

3) How difficult would it be to compute the same rankings (overall and per regional) without the elimination rounds being included? I would be interested to see how close this ranking system would be to the FIRST standings using the W-L-T/RP system after qualifications. Could it be a simple button on the web page to include or not include elimination rounds or even just compute the rankings both ways and display them side-by-side for comparison?

4) It would be helpful in the overall ranking page to also be able to see which regionals each team attended and the ranking points for that team at that regional in an additional column. For Example: 3, Archimedes, PINK, 233, 30, 3, 0, 97, FL=90 CO=99. This way you know Team 233 went to 2 regional competitions (FL and CO) and that they got relatively stronger. If this additional data were on the same overall page it would prevent a lot of jumping back and forth between pages.

5) The ability to sort by divisions at nationals might be helpful as well. This way you could get a short list of just the teams in your division.

6) With all the talk on CD about best, strongest, etc. regional competitions it seems like you have the data to compute a relative difficulty level of each of the regional competitions here as well. It would be an interesting statistical exercise.

Finally and most importantly, I want to say to all teams that numbers are just numbers. Don’t put too much faith in ANY ranking system. This ranking system at best points out which teams were relatively strong at the competitions they have already competed in. Use it as just that, an indication of what teams might be strong in Atlanta. Anything can and probably will happen in Atlanta. For those teams going to Atlanta please remember to have fun. Don’t get bogged down in the rankings. Relax, enjoy yourself, do the best you can, and remember that these are the things that make tomorrow’s memories.

Good luck to all teams going to Atlanta! With so many people traveling from so many far away places please travel safely.

Sorry for the long post!

-Chuck

[techtiger1]

Wow First off thanks to everyone for there imput especially for the last post very well put. The power ranking system has proven to be fairly accurate for us at the regionals we have been to. Hopefully it contines to be! A very special thanks to Sebas and Mr.Barriere for making this work and being so diligent about this thank you all very much.

[allen0977]

Originally Posted by DeepWater

1) I am unclear as to if this relative power ranking system is (or can be) computed in near real time or is it something that takes quite a while to compute due to the interdependence of the power rankings of the teams at the competition. If a team was to supply corrected information for their particular team or even accurate data for a whole regional is it a big deal to update/compute the rankings after the data is entered? Once the “compute ranking” button is pressed are the results fairly immediate or is it more of a go get coffee come back and they are ready or more of a start it tonight before bed and they are ready in the morning kind of thing? Is this a system that could be used instead of the real time FIRST W-L-T/RP system for ranking teams at competition? Is it even appropriate?

While at UCF and Palmetto, we generated this data Real Time. After each match we'd add the data to the db, and then refresh the page to see the new results. One regional takes a little less than a minute to compute. You can generate results real time for multiple regionals on the site by selecting them on the right. 2 regionals takes a little over a minute. 8 Regionals is on the order of 4-5 minutes, 19 regionals takes 30 minutes. As you add more teams it increases the data hash exponentially and thus takes much longer. 50 teams isn't too bad, 700 is a little out rageous

Originally Posted by DeepWater

2) How is the rank determined between teams with equal relative power rankings determined?

This is quite rare with so much unique data per team, but in the even that they are exactly the same, it sorts by team number.

Originally Posted by DeepWater

3) How difficult would it be to compute the same rankings (overall and per regional) without the elimination rounds being included? I would be interested to see how close this ranking system would be to the FIRST standings using the W-L-T/RP system after qualifications. Could it be a simple button on the web page to include or not include elimination rounds or even just compute the rankings both ways and display them side-by-side for comparison?

This is not difficult at all. I have the mechanism in place just hadn't generated the pages yet. With the multiple requests for this, I think I'll have to do that soon

Originally Posted by DeepWater

4) It would be helpful in the overall ranking page to also be able to see which regionals each team attended and the ranking points for that team at that regional in an additional column. For Example: 3, Archimedes, PINK, 233, 30, 3, 0, 97, FL=90 CO=99. This way you know Team 233 went to 2 regional competitions (FL and CO) and that they got relatively stronger. If this additional data were on the same overall page it would prevent a lot of jumping back and forth between pages.

This would be real hard to do real time, since the data changes dramatically each match, even when your team isn't even competing. But with static data like previous regionals, I should be able to store that data and output it .. would the "Relative Power" value be the number you'd like, or would the team rank be more appropriate?

Originally Posted by DeepWater

5) The ability to sort by divisions at nationals might be helpful as well. This way you could get a short list of just the teams in your division.

I was trying to figure out a good algorithm to do this, haven't thought through it just yet.

Originally Posted by DeepWater

6) With all the talk on CD about best, strongest, etc. regional competitions it seems like you have the data to compute a relative difficulty level of each of the regional competitions here as well. It would be an interesting statistical exercise.

This data must be cached somewhere in the program, it'd just be a matter of figuring it out. Maybe I should compute the win/loss percentage per regional and average margin of victory then run each regional through the program as if they were individual teams. That might have some interesting metrics

Originally Posted by DeepWater

Finally and most importantly, I want to say to all teams that numbers are just numbers. Don’t put too much faith in ANY ranking system. This ranking system at best points out which teams were relatively strong at the competitions they have already competed in. Use it as just that, an indication of what teams might be strong in Atlanta. Anything can and probably will happen in Atlanta. For those teams going to Atlanta please remember to have fun. Don’t get bogged down in the rankings. Relax, enjoy yourself, do the best you can, and remember that these are the things that make tomorrow’s memories.

Myself and the TechTigers whole-heartedly agree with this. Data is data, and statistics can be very misleading. As a Statistics minor in college, I spent much time analyzing data and recognizing falsehoods in the nature of the data. This is provided as a fun way to "measure" teams, but it is the unpredictable nature of the game that makes it fun to do this. FIRST does a wonderful job a leveling the playing fields with the game design and alliances. The only thing statistics are good for is predicting the past (and even that is arguable). We wish the best of luck to all teams at Atlanta, and are anxious to meet with you all there,

-Allen

[MattB703]

It seems like this ranking could be more accurate if it also included wins/losses in the elimination rounds.

[NoRemorse]

That is a very neat system. I am very impressed! I wrote somthing very similar to fill out my bracket in March!