Rigged partner assignments?

[dez250]

Originally Posted by SilverStar

What are the "few other items"?

I dont exactly remember every line in the algoritm but there are many spots in it which are there to prevent bugs... These are more or less copies of prior statements in there to work as a fall back so it checks 2 times to make sure you have the time between matches and you have only 2 blue bots and 2 red bots per match and such like that.

Are these algorithms public? If they are can you give us a link. The information you already gave was very useful.

[Kevin Kolodziej]

The title of the thread asks if matches are rigged. I do not believe this to be the case at all.

I can say for sure (at least from experience) that there are groups that are supposedly random that play together. In 2002 at the Midwest Regional, we (Team 71, at the time) always played right before Wildstang. Thats not very random. This year we (1064) would play with a team in one match, then against them in the next match - it happened three times.

As far as time goes, I like the idea, but I don't like what it does to the matches. And still, 71 ended up playing two matche spaced 5 rounds apart at the Midwest this year.

Back in 99 and 2000 at least, it did happen that we would play a few matches close together and have a long break...but with only 35 teams at the regional, a long break was never more than an hour...and we had 13 matches. Only rarely did a team see another team more than once on the field. I miss those days....

Kev

[Joe Ross]

Looking at the top 5 seeds at the chesapeake regional, it doesn't look like the parter assignments are rigged. Since the team number is a pretty good indication of the amount of time a team has been in FIRST, I simply averaged the team numbers of each of the top 5 seeds parters and opponents

Rank	Team #	Average Partner	Average Opponent
1.	222		488		694
2.	1405		868		749
3.	1083		708		733
4.	25		645		777
5.	303		486		760

If anything, the two newest teams had worse partners and better opponents then the 3 teams that had been around for more the 5 years.

Obviously, this is a very small sample. It would be better to do it on a smaller regional, where there are more matches to average out the differences.

[piotrm]

I'm not sure of the constraints of the base algorithm other than the obvious attempt to put some time between a team's matches.

Anyway, it doesnt matter much since that algorithm's output is not specific to any team. It generates matches with teams numbered 1 though N, the number of teams. These are pregenerated for various numbers of teams. During the actual events the teams that are present at an event are randomly assigned to those N numbers and thats that.

I do remember people said that at one event they played some practice matches and then the next day the qualification matches paired them with and against the same exact teams. This would happen if the seed wasn't changed between generation of practice matches and qualification matches. Since the pregenerated list of matches was the same (same number of teams present), the same seed made the teams be assigned to the same numbers as they were in the practice matches.

Hope this explains some things and if anyone has a really good and fast algorithm that you think would be good, suggest it to FIRST (and I would also like to hear about it).

[Joe Ross]

Originally Posted by piotrm

Hope this explains some things and if anyone has a really good and fast algorithm that you think would be good, suggest it to FIRST (and I would also like to hear about it).

This whole issue was discussed at length last year at length. http://www.chiefdelphi.com/forums/sh...ad.php?t=19012 The most interesting posts are Nate Smith's at the very end.

[KenWittlief]

it seems like this is not a very good way to go- if you are aligned with 3 or 4 other excellent teams then your matches have a better chance of running up high scores, which affects your placement

and if you are always playing with or against very simple bots that cant score many points, you are not going to be ranked very high at the end of the qual matches

surely FIRST, with all the brainpower at its disposal, can find a way to make sure teams are not always playing in matches together all day long and ensure that enough time is allowed between matches to at least get a new battery and make any quick needed repairs.

there has to be at least ONE person in FIRST who is good at math :^)

im mean, come on now - even if you slip the teams by one match (2 minutes) that means one of the teams you played with last time will play one match before you next time, and one will play one match behind you next time - if you do that all day long you will play with different teams in every match - some are on a 40minute schedule, some are on 44, some are on 48 and some are on 52 - to make up for the difference you switch ends on saturday and move the other way.

[MikeDubreuil]

Originally Posted by KenWittlief

it seems like this is not a very good way to go- if you are aligned with 3 or 4 other excellent teams then your matches have a better chance of running up high scores, which affects your placement

and if you are always playing with or against very simple bots that cant score many points, you are not going to be ranked very high at the end of the qual matches

surely FIRST, with all the brainpower at its disposal, can find a way to make sure teams are not always playing in matches together all day long and ensure that enough time is allowed between matches to at least get a new battery and make any quick needed repairs.

there has to be at least ONE person in FIRST who is good at math :^)

im mean, come on now - even if you slip the teams by one match (2 minutes) that means one of the teams you played with last time will play one match before you next time, and one will play one match behind you next time - if you do that all day long you will play with different teams in every match - some are on a 40minute schedule, some are on 44, some are on 48 and some are on 52 - to make up for the difference you switch ends on saturday and move the other way.

Yeah I agree.... the constraint to this whole system is how much time is "enough." What's the constant's value set at in the program?

[piotrm]

Originally Posted by Joe Ross

This whole issue was discussed at length last year at length. http://www.chiefdelphi.com/forums/sh...ad.php?t=19012 The most interesting posts are Nate Smith's at the very end.

Well it looks like it has been discussed but doesn't seem to have been resolved. One "Nate Smith" mentions he has the solution to all the problems but I fail to locate the algorithm used and considering that thread is kind of old, if the algorithm was now being used by FIRST, we shouldn't be having all these match list problems now. Or am I missing something?