"Random" match Schedules

[Gabe]

For this amount of trouble, lets step away from algorithms and any other sort of computerized system, as any kind of solution is bound to result in patterns, which we want to avoid to have a fair game.

This is my solution, and a very simple one. For each regional FIRST should buy popsicle sticks for as many teams as are attending. If there are 48 teams for the SV regional, FIRST buys 48 popsicle sticks. Each stick will have one team number written on it. Then they are all placed into a corrugated cardboard box with a lid. You shake the box really hard for five seconds, and then open it. Without looking, a person picks out sticks three at a time, forming alliances. After all of the sticks have been picked and having recorded the alliances, all 48 sticks are placed back in the box and the process is repeated as many times as necessary.

For added effect I suggest that the choosing of alliances be done in plain view of all of the teams. This will generate a lot of excitement as well as doing away with any doubt of having truly "random" match schedules.

(I think this is a realistic and fair way of chosing alliances.)

[Adam Richards]

Originally Posted by Gabe

How long would it really take for FIRST to write team numbers on pieces of paper and draw them out of a hat? For this amount of trouble, let us go to a simple solution.

This is my solution, and a very simple one. For each regional FIRST should buy popsicle sticks for as many teams as are attending. If there are 48 teams for the SV regional, FIRST buys 48 popsicle sticks. Each stick will have one team number written on it. Then they are all placed into a corrugated cardboard box with a lid. You shake the box really hard for five seconds, and then open it. Without looking, a person picks out sticks three at a time, forming alliances. After all of the sticks have been picked and having recorded the alliances, all 48 sticks are placed back in the box and the process is repeated as many times as necessary.

For added effect I suggest that the choosing of alliances be done in plain view of all of the teams. This will generate a lot of excitement as well as doing away with any doubt of having truly "random" match schedules.

(I think this is a realistic and fair way of chosing alliances.)

Only one problem: how long would this take for 100+ matches? A few hours, surely.

[Gabe]

The regional directors, as well as the rest of FIRST, has a little less than two weeks to figure out a solution. This is plenty, doing away with my original idea of doing it on actual competition day (which now that you point it out is a waste of time). If it takes time to end up with a fair game, so be it.

[Luckyfish05]

I hate to point this out, but there are regionals next weekend. Sooner is better than later in this case. If they're going to do something they'll probably (hopefully) do it this week.

[Arille]

Referring to what Joe Matt said in the VCU Regional Thread...

Originally Posted by Joe Matt

NEW 2008 GAME HINT: Not just interact with the field and what you know about, but have to interact with other robots what they do. Put yourself in the minds of your future alliance partners and opponenets. Not just about driving a robot onto a ramp.

Instead of thinking of this randomization (or lack thereof) as a flaw in the system, maybe, just maybe, it was intended to prep us for the 2008 game? If it is going to be needed that we know our alliance teams in and out, why not give us a taste of this collaboration in 2007 (even though it would've been nice knowing it far in advance.. )? Maybe in 2008 we'll have solid alliance teams (I doubt it, but its possible).....?

I know, I know: it's a long shot. (And probably not the reason at all.) But 'tis a funny thought!



SVR March 17!!

[Cory]

Originally Posted by Gabe

For this amount of trouble, lets step away from algorithms and any other sort of computerized system, as any kind of solution is bound to result in patterns, which we want to avoid to have a fair game.

This is my solution, and a very simple one. For each regional FIRST should buy popsicle sticks for as many teams as are attending. If there are 48 teams for the SV regional, FIRST buys 48 popsicle sticks. Each stick will have one team number written on it. Then they are all placed into a corrugated cardboard box with a lid. You shake the box really hard for five seconds, and then open it. Without looking, a person picks out sticks three at a time, forming alliances. After all of the sticks have been picked and having recorded the alliances, all 48 sticks are placed back in the box and the process is repeated as many times as necessary.

For added effect I suggest that the choosing of alliances be done in plain view of all of the teams. This will generate a lot of excitement as well as doing away with any doubt of having truly "random" match schedules.

(I think this is a realistic and fair way of chosing alliances.)

1) This is a waste of time. Volunteers and/or FIRST staff have much better things to be doing with their time.

2) You'll end up with teams playing a match, and then a second match, without enough time between the two. Hence the need for an algorithm.

Last year's algorithm, or plenty that have been posted by users here on CD would be fine. Anything but what we have now.

[Wayne C.]

Originally Posted by Psychoflood

we played 25 in at least half of our matches today. IMHO, i was very entertained by the eliminations at the regional, they were some very intense matches so i'm not complaining, but it seemed to me that they put a low number with a mid-ish number, and a rookie team/high number (<2000) I don't see how you can expect a fair and or "random" match from the same groups/sets of teams based on the teams number and not their robots specific ability/skill/efficiency this year

As the other half of that situation I wholeheartedly agree with you. We go to play ALL the teams in the event. We looked forward to playing WITH the Hawaiian Kids at least once yesterday and that never happened.
And we know that MORT knows how to kick our "bots defensively. There was nothing THAT wrong with the old random match generator. It certainly beats the nonsense we saw this weekend. The old format might not have made everyone happy but the new algorithm has succeeded in making just about everyone unhappy.

Not random, EVERY, and I literally mean EVERY game we played had at least team 948 (They became the number one seed). Because of these "random" match schedules we ended at the opposite end of the rankings.

[GaryVoshol]

Let's not contribute too many opinions on what was in the minds of the committee who designed this algorithm.

I don't think anyone consciously said, "Let's put vet against vet and rookie against rookie." Rather, the "max time between matches" constraint was emphasized over all other constraints. Then they simply decided on team number as a primary sort - they could have done it alphabetically, or could have done it randomly. It wouldn't matter. Once "max time between matches" was decided as having priority, teams would end up seeing the same teams over and over again.

The popsicle stick in a bowl thing can easily be set up as a computer simulation. But then we have decided that "randomness" is the highest priority, and time between matches means nothing.

Teach the computer could to do a popsicle stick picking. For Round1, totally randomize it. To pick for Round2 make a constraint that there has to be at least "X" matches before a team must compete again. "X" would vary based on the number of teams in a regional. It should be at least 3, but in the larger regionals could be 4 or 5. To pick the first X matches in Round2, the popsicle sticks of any team that played in the last X matches in Round1 would be set aside. Once the first X matches are picked, all the remaining popsicle sticks are thown into the drum, and the rest of the matches of the round can be drawn. Repeat for all matches up to lunch time. Then totaly randomize again, and finish Friday - or maybe set the X constraint for the first match of the afternoon to 1. Totally randomize again and pick for Saturday.

This wouldn't work as well for the small regionals of 36 teams or less, because it would effectively divide the pool into two, one group playing the first half of the round and other playing the second half. Perhaps there the constraint could be set at 2, with a 1-match-length break between each Round.

This would have to be somewhat adjusted for regionals with numbers not exactly divisible by 6.

Get working, FIRST contractors!

[AcesPease]

Originally Posted by Wayne C.

As the other half of that situation I wholeheartedly agree with you. We go to play ALL the teams in the event. We looked forward to playing WITH the Hawaiian Kids at least once yesterday and that never happened.
And we know that MORT knows how to kick our "bots defensively. There was nothing THAT wrong with the old random match generator. It certainly beats the nonsense we saw this weekend. The old format might not have made everyone happy but the new algorithm has succeeded in making just about everyone unhappy.

I would like to make a move for any system that lets us face as many different teams as possible. I am even willing to play two rounds in a row if this is needed, we have more than one battery, so unless you have a breakdown this should not be a major dissadvatage. The seeding match system does not need to be "random" but it should give the teams the widest exposure to each other. Many of us go to these events to meet as many other teams as possible. Playing the same teams 7 or 8 times greatly narrows that experience.

[Bongle]

Originally Posted by Keith Watson

The PNW Regional data shows a pattern. Not to take away from every team who performed well, but is this really the "intended" outcome? Note that the rules clearly state the schedule cannot be changed after it is published.

That pattern isn't really that shocking. Any team that has faced the top ranked teams in a match will tend to have a lower ranking. Any team that has faced them multiple times will have an even lower ranking. The bad part here is not that teams that face good robots multiple times end up with a poor ranking, it is that teams HAD to face the same good robot 4+ times.

Anyway, watch this space for redos of the stats I posted yesterday, but including the final day.

[Dominicano0519]

the matches in nj all have a certain order to them.

ex.

our match schedule was
1
11
21
31
41
etc....

that was a good thing( no back to backs)

however
we had the same team against us every single match

and the lower number team on the opposing alliance was our alliance partner in the next match

an example would be our first three matches

1: we had 11( lowest number in NJ) and opponents were 25(next lowest) and 486( the constant opponent)

2:we had 25(opponents from 1) and they had 41 and 486

3: we had 41 and they had 75 and 486

this trend just kept on going and going and going.

also since the pits were arranged by numbers i noticed something else while scouting for the following matches i never had to talk to anyone in my row or in the middle rows. all of my opponents and allies were in the same two rows every single match.

i don't think this is right. it should be the luck of the draw( maybe that explains the whole red alliance is better than blue idea)

[Travis Hoffman]

Hey, by using the search feature (orange bar at the top of the page), you would find there is already a very well traveled thread topic on this matter.

Please post in the following thread.....

http://www.chiefdelphi.com/forums/sh...threadid=55178

[Dominicano0519]

Originally Posted by T. Hoffman

Hey, by using the search feature (orange bar at the top of the page), you would find there is already a very well traveled thread topic on this matter.

Please post in the following thread.....

http://www.chiefdelphi.com/forums/sh...threadid=55178

uhh did u notice at what times they were written?

apparently we started writing at almost the same time, just since my post was longer i took longer to write

[Bongle]

Ok, rather than spamming with a few billion lines of text, I'll just do the summaries:

There are three main stats here:
Competitior Pool Size - The size of the set of all robots a given team played against
Ally pool size - The size of the set of all robots a given team played with
Opponent pool size - The size of the set of all robots a given team played against

All the pool sizes are given as a percentage of how many robots were at that regional. That way, you can say "ooo, at St Louis last year I could expect to play with or against 88.5% of the robots at the regional". Note that the lower numbers for the bigger regionals is predictable, since the more robots you have at a regional, the fewer matches you'll have and therefore less chance to 'meet' them all (also its a bigger pool to meet from anyway).


2006 - Aim High
NASA VCU Regional
Average Matches Played: 8
Robots at regional: 64
Competitor Pool Avg Size: 60.6%
Ally Pool: 26.5%
Opponent Pool: 36.6%

Pacific Northwest Regional
Average Matches Played: 10
Robots at regional: 46
Competitor Pool Avg Size: 83.7%
Ally Pool: 43.1%
Opponent Pool: 54.8%

Great Lakes Regional (I know it hasn't happened this year yet, but its hard to find 2006 results)
Average matches played: 8
Robots at regional: 63
Competitor Pool Avg Size: 61.3%
Ally Pool: 26.9%
Opponent Pool: 36.6%

St Louis Regional
Average matches played: 12
Robots at regional: 40
Competitor Pool Avg Size: 88.5%
Ally Pool: 54.1%
Opponent Pool: 66.5%

2007 - Rack n Roll with fun happy matching algorithm
NASA VCU Regional
Average Matches Played: 8
Robots at regional: 66
Competitor Pool Avg Size: 36.3%
Ally Pool: 25.7%
Opponent Pool: 26.2%

Pacific Northwest Regional
Average Matches Played: 8.1
Robots at regional: 54
Competitor Pool Avg Size: 44.4%
Ally Pool: 31.4%
Opponent Pool: 32.0%

Jersey Regional
Average Matches Played: 7.2
Robots at regional: 59
Competitor Pool Avg Size: 40.3%
Ally Pool: 25.8%
Opponent Pool: 23.8%

St Louis Regional
Average matches played: 8.6
Robots at regional: 45
Competitor Pool Avg Size: 52.4%
Ally Pool: 39.3%
Opponent Pool: 36.4%

So as you can see, the # of robots you meet in gameplay drastically dropped from previous years. Something I noticed though is that there has to be substantial overlap between the ally and opponent sets in 2007*, something that wasn't present in 2006. This means that much more than in previous years, teams would face one team as an opponent as well as play with them as an ally.

*That is to say, summing the size of the ally and opponent sets in 2006 results in a number only barely larger than your total competitor size, implying that few robots are in both ally and opponent sets. In 2007 however, ally + opponent is often much larger than the competitor pool.

Also note that St Louis's big drop is probably helped by the fact that more robots came to it, AND they ran far fewer matches than last year.