PDA

View Full Version : #1 seeded teams


Tim Delles
03-20-2006, 12:27 PM
So i decided to go through each regional just to see how many number one seeds went on to win the regional. well out of the 18 that have been played so far, only 9 #1 seeded teams have won the regional. Now this makes me ask does it pay to be the number 1 seeded team anymore?

BAE
non-#1 seed won

NJ
#1 seed won

VCU
non-#1 seed won

Pacific NW
non-#1 seed won

Arizona
non-#1 seed won

Buckeye
non-#1 seed won

FLR
#1 seed won

Florida
non-#1 seed won

GLR
#1 seed won

Pittsburg
#1 seed won

St. Louis
non-#1 seed won

UTC
non-#1 seed won

Boilermaker
non-#1 seed won

Chesapeake
#1 seed won

Detroit
#1 seed won

Mid-West
#1 seed won

Peachtree
#1 seed won

SVR
#1 seed won

*** this data came from the FIRST website. If it is incorrect please let me know or post it***

Rick TYler
03-20-2006, 12:30 PM
Now this makes me ask does it pay to be the number 1 seeded team anymore?

You ask, "does it pay to be the number 1 seeded team anymore?" The answer is, "it depends."

Elgin Clock
03-20-2006, 12:31 PM
Chesapeake
No Data

#1 seed won.

Joe Ross
03-20-2006, 12:38 PM
So i decided to go through each regional just to see how many number one seeds went on to win the regional. well out of the 18 that have been played so far, only 7 #1 seeded teams have won the regional. Now this makes me ask does it pay to be the number 1 seeded team anymore?

Since more #1 seeds won then any other seed, there is still an advantage.

It would be interesting to see a histogram of all seeds and their number of wins, and compare that to the last few years.

K.Shaw
03-20-2006, 12:45 PM
the #1 seeded team can make an alliance with a good robot but then may have a not so great robot for a 2nd pick because of the way alliance selections go. by the time it gets back to the first seed 23 robots are already chosen giving the #1 seed a pretty hard decision.

my point here i know the data shows the #1 seed is better but the upper seeds like 7 or 8 could actually have a better alliance since they are picking their 1st and 2nd robot at the same time.

Kims Robot
03-20-2006, 12:52 PM
I think that the serpentine draft this year is definitely evening out the alliances. Its making for much more interesting matching, and its making it almost impossible to have the #1 seed be an overpowered alliance of three amazing robots. It would be interesting to compare this data to previous years with the regular draft.

So now, being the number one seed means you get the first pick and the last pick, so unless your robot is completely unstoppable, my guess is other alliances still have a chance against you.

Now what would make it even crazier would be if they did a serpentine draft, and then set #1 to play against #5 instead of #8 (ie 1vs5, 2vs6, 3vs7, 4vs8). If FIRST is trying to even out the alliances, I think they've succeeded, and to be honest, I think I like it more than I thought I would!

dlavery
03-20-2006, 12:57 PM
So i decided to go through each regional just to see how many number one seeds went on to win the regional. well out of the 18 that have been played so far, only 7 #1 seeded teams have won the regional. Now this makes me ask does it pay to be the number 1 seeded team anymore?

As expected, there has been quite a bit of discussion about the "serpentine draft" process for selecting alliances. Now we are starting to have some actual data to support those discussions.

Historically, the traditional selection process has had the highest seeded team pick first within each round of selections. That made for a significant number of very, very strong alliances during the finals (some would argue, unbeatable). If the alliances were all of equivalent capability and skill, then you could logically conclude that each of the eight alliances in the finals would have an equal probability of winning the competition. There should be an even distribution of wins vs. seed position at the conclusion of the competition season. A #1 seeded alliance should win 12.5% of the time, as would a #2 alliance, a #3 alliance, etc. But the data showed that was not the case. The #1 alliance typically won more than half the time. Yes, there were occasionally upsets and "underdog stories." But the reality is that if you were the #1 seed team, you knew you had a 1-in-2 probability of winning the event, while the lower seeded alliances (which nominally should expect a 1-in-8 chance), had considerably lower probability of winning. Being the #1 seed should allow an alliance to come into the finals with some advantage. But the concern was that the old alliance selection process gave them TOO MUCH of an advantage and it was killing the competition.

The new selection process was, in part, an attempt to slightly even out the competitive level for the finals. No one believes that we will ever reach the point that each of the final alliances will actually have a real 1-in-8 probability of winning the competitions, and that was not the intent. But the desire was to move a little bit away from the point where the #1 alliance position was almost an assured win every time.

Based on these initial numbers, it appears that that effort may have been only partially successful. Out of 18 events, a truly even distribution would have the #1 alliance winning 2.25 times, or 12.5% of the time. But these data shows that they have won 8 times, giving a 44% probability of winning the competition. Given that the #1 alliance is still out-performing pure probability by a factor of more than three, I think that we can safely conclude that being in the #1 alliance position is still a very good place to be. The teams are obviously able to still build strong alliances, and play the game successfully. While the sample size is still a little small for an accurate small number statistics analysis, it does appear that the #1 alliance position still wins the competition a disproportionately high number of times. Picking first during the draft still provides a strategically important advantage to the #1 alliance position that is only slightly affected by, and absolutely not overshadowed by, the serpentine draft process.

-dave

Tim Delles
03-20-2006, 01:04 PM
I think that the serpentine draft this year is definitely evening out the alliances. Its making for much more interesting matching, and its making it almost impossible to have the #1 seed be an overpowered alliance of three amazing robots. It would be interesting to compare this data to previous years with the regular draft.

So now, being the number one seed means you get the first pick and the last pick, so unless your robot is completely unstoppable, my guess is other alliances still have a chance against you.

Now what would make it even crazier would be if they did a serpentine draft, and then set #1 to play against #5 instead of #8 (ie 1vs5, 2vs6, 3vs7, 4vs8). If FIRST is trying to even out the alliances, I think they've succeeded, and to be honest, I think I like it more than I thought I would!

I will see what I can do about getting some information from a few years ago competition and comparing it to this years. I just have to find the time.

billbo911
03-20-2006, 01:32 PM
The new selection process was, in part, an attempt to slightly even out the competitive level for the finals. No one believes that we will ever reach the point that each of the final alliances will actually have a real 1-in-8 probability of winning the competitions, and that was not the intent. But the desire was to move a little bit away from the point where the #1 alliance position was almost an assured win every time.

-dave
I believe the new process has achieved an excellent balance of rewarding the teams that have earned the highest ranking and not guaranteeing them that they will win the event.

These high ranking teams have earned their position, not by luck but by hard work, perseverance and solid planning and design. I don't think it would be right to force them to a position of having to choose alliance partners that would bring their chances down to the point of only having a 12.5% chance of winning. I also don't think having a 75% or greater chance of winning is right either.

Personally, I wouldn't want to send teams to Nationals that were not representative of our Regionals best! By having the #1 seed teams running at about 50% winning rate, I think we actually are seeing the Best of the Best advancing. Here's why: During qualification rounds, quite often a really good team will be allied with teams that are, well, not so strong. Ultimately, their overall ranking may not be reflective of how well that team can really perform. Then during the alliance selection process they are chosen by a team that has paid attention and seen what the different teams are capable of doing and therefore, assemble a team capable of winning the Regional.

All in all, I say leave the process how it is. It is a good balance of "Fairness" without too much "Bias".

Gene F
03-20-2006, 01:34 PM
While I do think that the alliance selection process has made a difference and made the alliances a little more even, I thnk the bigger impact is the three on three seeding rounds. It is now much more difficult for a single team to control the outcome of every match they are in. I have heard a lot of stories of teams that say they have as a team failed in almost every match and still end up being a seeded team because their alliance partners in the seeding rounds have carried them. This makes the scouting process all that much more important. I like it!!

Joe Ross
03-20-2006, 01:51 PM
Not only does the #1 seed win 50% of the time, none of the other seeds are even close.

Jeff Rodriguez
03-20-2006, 01:57 PM
Not only does the #1 seed win 50% of the time, none of the other seeds are even close.

Are you going by the teams seed before alliance selection or their seed at the time of selection?
Your chart has no 5 seeds winning, but the winning alliance at UTC was the 5th seed alliance. They were headed by 177, who seeded 8th, but due to picking within the top 8, ended up picking 5th.

Joe Ross
03-20-2006, 01:59 PM
Are you going by the teams seed before alliance selection or their seed at the time of selection?
Your chart has no 5 seeds winning, but the winning alliance at UTC was the 5th seed alliance. They were headed by 177, who seeded 8th, but due to picking within the top 8, ended up picking 5th.

That's correct, I did it by original seed, as that is what teams can control. (It's also easier to quickly gather that data ;))

Beth Sweet
03-20-2006, 02:44 PM
Here's my take.

#1 seed only matters if you know what you want. You can be #1 seed all you want, but if you have bad scouting data and thus select either an inefficient or uncomplimentry partner, there's no point.

I think that it has been made tougher this year for #1 seeds because of the new selection process. As I've voiced before, it's not my favorite change however probably does make things more fair.

MikeDubreuil
03-20-2006, 03:07 PM
Here's my 2 cents...

The #1 seed is still dominant because their first pick could be the #2 seeded robot. When 2/3 of an alliance are composed of the best robots at the regional it's guaranteed success.

I think FIRST should re-instate the rule that requires the top 8 seeded teams to pick teams outside of the top 8. Adding that rule back in and keeping the new serpentine rule should make for some exciting finals.

Tytus Gerrish
03-20-2006, 03:10 PM
in florida atleast, the #2 seed turned down the #1 seed and they both ended up getting eliminated. if they had teamed up it would have been over right there.

Tim Delles
03-20-2006, 03:14 PM
Here's my 2 cents...

The #1 seed is still dominant because their first pick could be the #2 seeded robot. When 2/3 of an alliance are composed of the best robots at the regional it's guaranteed success.

I think FIRST should re-instate the rule that requires the top 8 seeded teams to pick teams outside of the top 8. Adding that rule back in and keeping the new serpentine rule should make for some exciting finals.

So are you saying that the number 1 and number 2 seeded teams are the best at the regional/championships? If so ( I hate to say it again) then maybe you should think back to Nationals in 2003 when Miss Daisy picked Buzz who was seeded 17th, and the techno-ticks who were seeded 60th. they went on to win thier division. Also if you look at the number 1 seed from their division they picked truck town who was seeded 25th and then team 9 who was seeded 35th. The reason that in a lot of regionals the number 1 seeded team picks teams in the top 8 is because they have worked with them and they work well. But thier normally aren't 70 teams that they can choose from to work with.

Last year 175 and 33 picked team 108 who was seeded 79th. and they won thier division.

Also 56 and 254 picked 64 who was seeded 63rd and won thier division.

Jeff Rodriguez
03-20-2006, 03:18 PM
The #1 seed is still dominant because their first pick could be the #2 seeded robot. When 2/3 of an alliance are composed of the best robots at the regional it's guaranteed success.

Yes, but it takes three robots to win this game. The best two robots at a regional will never beat three good robots in this game.

MikeDubreuil
03-20-2006, 03:28 PM
So are you saying that the number 1 and number 2 seeded teams are the best at the regional/championships?
Statisticly speaking, yes.
...then maybe you should think back to Nationals in...
Statisticly speaking, that is an anomaly.

Yes, but it takes three robots to win this game. The best two robots at a regional will never beat three good robots in this game.
Perhaps not a 2 on 3, but statisticaly speaking there is a much better chance an alliance composed of the two highest seeded robots will win the regional.

Tom Bottiglieri
03-20-2006, 03:35 PM
Yes, but it takes three robots to win this game. The best two robots at a regional will never beat three good robots in this game.
Yes but two great robots and one good robot (a little hard to come by at smaller regionals, but almost a definite in Atlanta) will beat an alliance of three good robots.

Tim Delles
03-20-2006, 03:41 PM
Statisticly speaking, yes.

Statisticly speaking, that is an anomaly.

Perhaps not a 2 on 3, but statisticaly speaking there is a much better chance an alliance composed of the best two robots will win the regional.

the BEST TWO ROBOTS, not the 2 highest seeded.

NJ
25 (#1) picked 103 (#10)

FLR
1126 (#1) picked 229 (#6)

GLR
Being clarified right now

Pittsburg
395(#1) picked 1038(#6)

Chesapeake
1629(#1) picked 175(#4)

Detroit
469 (#1) picked 217 (#14)

Mid-West
1756(#1) picked 111(#3)

Peachtree
1261 (#1) picked 1414 (#2)

SVR
254(#1) picked 581 (#9)

so of all the teams that seeded number 1 and went on to win only 1 had the alliance of the number 1 and number 2 seeded teams together. statistically it looks bad for the number 1 and 2 seeded alliances to team up.

Also what statistics are you using to show that the number 1 and number 2 seeded teams are really the best at any given regional?

Rick TYler
03-20-2006, 03:51 PM
t
so of all the teams that seeded number 1 and went on to win only 1 had the alliance of the number 1 and number 2 seeded teams together. statistically it looks bad for the number 1 and 2 seeded alliances to team up.

Also what statistics are you using to show that the number 1 and number 2 seeded teams are really the best at any given regional?

First, I apologize for being flippant up there in this thread.

Second, your data show a trend, but it is a pretty small sample size. You might not really be able to draw good conclusions until all the regionals are over. You would then have a sample that might be big enough to be statistically significant. It's an interesting exercise. Thanks for taking time to do the research.

MikeDubreuil
03-20-2006, 03:53 PM
the BEST TWO ROBOTS, not the 2 highest seeded...

Also what statistics are you using to show that the number 1 and number 2 seeded teams are really the best at any given regional?

Sorry, I meant two highest seeded robots (I have edited my previous post).

The #1 and #2 seeded robots are the best robots. Otherwise, you are saying that the seeding rounds are used to place robots in random order. :confused:

lukevanoort
03-20-2006, 03:53 PM
I think FIRST should re-instate the rule that requires the top 8 seeded teams to pick teams outside of the top 8. Adding that rule back in and keeping the new serpentine rule should make for some exciting finals.
I agree, this would greatly increase fairness. Lets say there is a imaginary team 666 (I'm not satanic, I just happen to know this number is not in use) at the imaginary Pork Ridge regional. 666 is considered one of the best, if not the best team at Pork Ridge. However, they get a little bad luck in their last match and ends up seeded number 7. In alliance choosing, under last year's method the #1 seeded (very good) team 1337 could pick 666 then another strong team, 13. This alliance would be nearly unbeatable. With this year's system, 1337 could pick 666 then have to wait a while till they get to choose again. This time 13 was snatched up by another alliance, so they have to go with 990, another good team, but not as good as 13. Now they will be the favorites to win, but not unstoppable. Then, with a serpentine system and the old rule, 1337 can't pick 666, so they pick 13. 666 picks a really good team, 803, then when they get the choice again, they pick 193. 1337 then picks 990. Now the alliances are more even, and there aren't any super alliances and the winner is uncertain, assuming all the captains did their homework and chose good teams.

Alan Anderson
03-20-2006, 03:53 PM
Also what statistics are you using to show that the number 1 and number 2 seeded teams are really the best at any given regional?
The Law of Large Numbers tells us that the ranking at the end of a sufficient quantity of randomly assigned qualification matches will reflect the robot goodness with high confidence. It's too bad that an actual competition doesn't have anything near the number of matches necessary to make that happen. All we get is a very rough approximation.

Besides, an effective alliance is not just the sum of its component teams. Two or three complementary robots can do better together than two or three nominally "better" robots that don't work with each other as well.

Tim Delles
03-20-2006, 04:05 PM
Sorry, I meant two highest seeded robots (I have edited my previous post).

The #1 and #2 seeded robots are the best robots. Otherwise, you are saying that the seeding rounds are used to place robots in random order. :confused:

All I am trying to say is that as Alan as said we don't play enough matches to prove who really should be seeded number 1 or number 2, that they are just rough approxamations. Looking at FLR, some of the top seeded teams never played against each other. Also remember that the #1 seeded team may need something different than what the #2 seeded team has to offer. But yes if you want to take it as me saying the robots are placed in random order then go ahead and do so, because when thinking about ranking you don't take into account matches when you have 2 robots against 3 because someone didn't make the match (Since that is not a good judge of your robot with respect to the alliance). You don't play all the robots at a regional, and I would venture to say that you barely play half at the Championship. So the best robots could be teams that aren't seeded number 1 and number 2. But by all means if you are number 1 pick the #2 seed. Because i know that if i was the number 3 seed, i would pick a perfect partner for me, no matter thier seed.

I do plan on keeping a tab on all of the #1 alliances that win regionals, so that i have a slightly larger, but by no means perfect, sample.

GaryVoshol
03-20-2006, 04:15 PM
GLR
Being clarified right now
#1 (1503) picked #3 (1114)

Nuttyman54
03-20-2006, 04:36 PM
just to prove that the seeds do not necessarily demonstrate the best robots: We were seeded 15th (out of 40) at the end of fridays matches.

Number of matches we were scheduled for: 7
Number of matches we fielded a robot: 4
Number of matches our robot moved: 2
Number of matches our robot was able to score: 0
Number of matches our robot worked: 0
Reason we ended up 15th: Our alliance partners

The robot was fixed, we played both of our matches on saturday as a fully functional alliance and lost both. We ended up 26th seed and were not picked.
Yes, the top seeded robots will be some of the best robots, but just as we had a lucky break on Friday, other teams got the worse end when we DIDN'T field a robot, and lost because of that. They may have had the best robot at the regional, but couldn't win a 2 on 3 match against 3 good robots

Sgraff_SRHS06
03-20-2006, 04:58 PM
As expected, there has been quite a bit of discussion about the "serpentine draft" process for selecting alliances. Now we are starting to have some actual data to support those discussions.

Historically, the traditional selection process has had the highest seeded team pick first within each round of selections. That made for a significant number of very, very strong alliances during the finals (some would argue, unbeatable). If the alliances were all of equivalent capability and skill, then you could logically conclude that each of the eight alliances in the finals would have an equal probability of winning the competition. There should be an even distribution of wins vs. seed position at the conclusion of the competition season. A #1 seeded alliance should win 12.5% of the time, as would a #2 alliance, a #3 alliance, etc. But the data showed that was not the case. The #1 alliance typically won more than half the time. Yes, there were occasionally upsets and "underdog stories." But the reality is that if you were the #1 seed team, you knew you had a 1-in-2 probability of winning the event, while the lower seeded alliances (which nominally should expect a 1-in-8 chance), had considerably lower probability of winning. Being the #1 seed should allow an alliance to come into the finals with some advantage. But the concern was that the old alliance selection process gave them TOO MUCH of an advantage and it was killing the competition.

The new selection process was, in part, an attempt to slightly even out the competitive level for the finals. No one believes that we will ever reach the point that each of the final alliances will actually have a real 1-in-8 probability of winning the competitions, and that was not the intent. But the desire was to move a little bit away from the point where the #1 alliance position was almost an assured win every time.

Based on these initial numbers, it appears that that effort may have been only partially successful. Out of 18 events, a truly even distribution would have the #1 alliance winning 2.25 times, or 12.5% of the time. But these data shows that they have won 8 times, giving a 44% probability of winning the competition. Given that the #1 alliance is still out-performing pure probability by a factor of more than three, I think that we can safely conclude that being in the #1 alliance position is still a very good place to be. The teams are obviously able to still build strong alliances, and play the game successfully. While the sample size is still a little small for an accurate small number statistics analysis, it does appear that the #1 alliance position still wins the competition a disproportionately high number of times. Picking first during the draft still provides a strategically important advantage to the #1 alliance position that is only slightly affected by, and absolutely not overshadowed by, the serpentine draft process.

-dave

But that does not stop the higher seeds from doing their most vicious weapon--consolidate with other members of the Top-8 to get an advantage. At Chesapeakes, consolidation madness was everywhere. When 1629 got together with 175 (the #4-ranked team) and 181 (Birds of Prey) combined with 888 (the #5-ranked team) in the alliances, they were unbeatable (even though we (the #7 ranked team) combined with 614 to form the #4 alliance). If consolidation between the top-8 were not allowed to happen (like there would be a rule stating that the top-8 could not select each other), then the serpentine-selection rule would make the matches much more even.

MattB703
03-20-2006, 05:10 PM
But that does not stop the higher seeds from doing their most vicious weapon--consolidate with other members of the Top-8 to get an advantage. At Chesapeakes, consolidation madness was everywhere. When 1629 got together with 175 (the #4-ranked team) and 181 (Birds of Prey) combined with 888 (the #5-ranked team) in the alliances, they were unbeatable (even though we (the #7 ranked team) combined with 614 to form the #4 alliance). If consolidation between the top-8 were not allowed to happen (like there would be a rule stating that the top-8 could not select each other), then the serpentine-selection rule would make the matches much more even.

This is certainly true in the regional competitions and the smaller the regional the more true it will be. However, I believe that at big regionals like GTR and in the Divisions at the Championship Event the new system will work very well and you will see much less domination by the #1 seed teams even with the ability of #1 seed to pick with-in the top 8.

In summary; I like the current system.

Matt B.

Sgraff_SRHS06
03-20-2006, 05:18 PM
All I am trying to say is that as Alan as said we don't play enough matches to prove who really should be seeded number 1 or number 2, that they are just rough approxamations. Looking at FLR, some of the top seeded teams never played against each other. Also remember that the #1 seeded team may need something different than what the #2 seeded team has to offer. But yes if you want to take it as me saying the robots are placed in random order then go ahead and do so, because when thinking about ranking you don't take into account matches when you have 2 robots against 3 because someone didn't make the match (Since that is not a good judge of your robot with respect to the alliance). You don't play all the robots at a regional, and I would venture to say that you barely play half at the Championship. So the best robots could be teams that aren't seeded number 1 and number 2. But by all means if you are number 1 pick the #2 seed. Because i know that if i was the number 3 seed, i would pick a perfect partner for me, no matter thier seed.



Exactly agreed. That is why you sometimes pick oddball partners. They may work to help you or they may serve to hurt you. But it's all luck (we found out the hard way).

ScoutingNerd175
03-20-2006, 06:17 PM
I'm going to put some numbers out there. I suppose many (not all) would consider the best team to be the team that scored the most points. So lets look at last years nationals since our selections there have already been brought up.
175 was seeded third. Seed 1 picked Seed 2 bumping us up to the second pick. We then picked 33 (possibly seed 11?). The reason we did this was because Team 33 had the best robot in curie division. (again if you consider the highest scoring robot to be the best robot) By those same numbers (objective as always and I'm sorry if it seems like I'm tooting my own horn) we were the second best robot in the division. The third, I believe was 118. Team 108 (I had no idea they were seeded so low, our scouting system doesn't take wins and losses or alliance total scores into account) was somewhere in the 5-8 range as far as scoring goes. So, at least in our division, Seeding did not have that much to do with actual scoring ability.
At Chesapeake last year, the top scoring robot (173) was seeded first. They then picked 1027, the third highest scoring robot. They did in fact win. The third seed robot was, 007, was the sixth highest scoring robot. Seeding was a bit closer than nationals, but not much.
This year a Chesapeake the highest scoring robot, 293, was third seed, the 2nd highest scoring robot, 341, was, I think 10th seed, the fifth, 175, was 4th seed, the sixth was 1389, which was actually still around for the second pick of the third alliance, the 1st seed, only undefeated team, was 1629 the 8th highest scoring robot. I am not sure of the seeds of the 56 and 399, the third and fourth highest scoring teams. The seeding this year reflected score even less than last year. However, the seeding did accurately predict winning in every single match up of the finals. This, I think, is the exception, not the rule. Again this assumes that you consider the best robot to be the highest scoring robot. Not necessarily true. Part of the first seed's success at Chesapeake was based on the fact that all three teams got up the ramp all but one time.
So there's the endless string of data that seems to come whenever I post.
I think that it really depends on the ability of the lower seeded teams to scout. If every single team had great scouting, then this would level the playing field. As it is, some teams do not have great scouting. In Chesapeake I think 6 out of 8 first picks were out of the top 8 at that time. This brought I think seed 14 into picking position. This team would be at a disadvantage because they were not expecting to be picking. Some teams end up seeded in the top 8 without ever having done great scouting, making their picks somewhat random. If you are a number one seed and the earlier seeds leave good robots (much more likely at Nationals) you will have a much better chance than if you are a first seed and no good robots have been left behind (much more likely at smaller regionals). Let me know if any of that made sense.

Tom Bottiglieri
03-20-2006, 08:02 PM
The Law of Large Numbers tells us that the ranking at the end of a sufficient quantity of randomly assigned qualification matches will reflect the robot goodness with high confidence. It's too bad that an actual competition doesn't have anything near the number of matches necessary to make that happen. All we get is a very rough approximation.

Besides, an effective alliance is not just the sum of its component teams. Two or three complementary robots can do better together than two or three nominally "better" robots that don't work with each other as well.
I took some time to look at that problem from a statistical point of view. If a team does well in one match because they are with complimenting robots, they have a lesser chance of repeating the same scenario. Therefore, "how good teams actually are" can be found by looking at their individual score for a match compared to the teams they are allianced with's average scores over the event.

I have started an excel spreadsheet program that will help factor luck out of the equation for rankings.

www.team195.com/scouting/aimhighstats.xls

Please take the time to look at this spreadsheet. If you have any questions, please post or PM me. The format is quite crude seeing as I made this in a short amount of time, but I plan to further the development of this tool to help aid in team selections at nationals.

Tim Delles
03-20-2006, 08:16 PM
Just a side note. I checked into, but i haven't wrote anything down. But for those that think that the #1 and #2 seeded teams are the best think of this. of the regionals i have looked through. only 1 #1 #2 alliance has gone on to win the regional this year. so what does that say about #1 and #2 seeded teams being the best?

Just something to talk about.

Nuttyman54
03-20-2006, 08:32 PM
It all depends on what is considered the "Best" robot. Highest scoring is not necessarily the number 1 seed. The #1 seed SHOULD be the robot that can do well no matter who it plays with. If you were to continue playing until all possible alliances pairings had played, that is who would come out on top. In this case, it would be interesting if the #1 and #2 seeds would be the best alliance partners.

However, this is not the case. At the regionals, it will end up that the robots which are able to score more despite heavy defense, incompetent alliance partners, etc. end up on top. And so, the #1 and #2 seeds may not have complementary strategies. In fact, they may have the SAME strategy. In this game, that could mean that they get in eachother's way, or become easier to defend against.

The Lucas
03-20-2006, 09:28 PM
The Law of Large Numbers tells us that the ranking at the end of a sufficient quantity of randomly assigned qualification matches will reflect the robot goodness with high confidence. It's too bad that an actual competition doesn't have anything near the number of matches necessary to make that happen. All we get is a very rough approximation.

Considering that matches are 3v3 that number would have to be HUGE. When we went to 3 teams per alliance we sacrificed a good deal of accuracy in our seeding. Your robot is nothing more than 1/3 of youR "randomly" paired alliance. You are against 3 other bots, 2 or 3 of which might be powerhouses. If you are by far the best bot on your "randomly" paired alliance, you will draw all the defense and your partners may not be able to take advantage of this. I like the new 3 bot alliances because of the complexities it add to elims, but it does make seeding more difficult. It could be worse. At least seeding is based on wins and losses instead of entirely on loser's score (like 2003).

Scott358
03-20-2006, 11:39 PM
I would suggest significant factors are:

1. The quality of scouting (as stated previously)

2. The depth of the field

3. How different game play is between the qualifying rounds and the playoffs(specifically defense).

lukevanoort
03-21-2006, 09:01 PM
I took some time to look at that problem from a statistical point of view. If a team does well in one match because they are with complimenting robots, they have a lesser chance of repeating the same scenario. Therefore, "how good teams actually are" can be found by looking at their individual score for a match compared to the teams they are allianced with's average scores over the event.

I have started an excel spreadsheet program that will help factor luck out of the equation for rankings.

www.team195.com/scouting/aimhighstats.xls

Please take the time to look at this spreadsheet. If you have any questions, please post or PM me. The format is quite crude seeing as I made this in a short amount of time, but I plan to further the development of this tool to help aid in team selections at nationals.
Remember you can't really quantify what team is the best off of just a few factors. A team that doesn't have a good score can be very good, or a team with a high score can be low-quality (for example they might have unreliable electricals). The best robot isn't the best shooter, or the best defender, or the best picker upper, it's a combination of good components that work together well. In my experience, the only thing that would be able to take all the factors and spit out good, reliable, accurate rankings is an experienced human with a photographic memory.

Peter Matteson
03-21-2006, 09:24 PM
Remember you can't really quantify what team is the best off of just a few factors. A team that doesn't have a good score can be very good, or a team with a high score can be low-quality (for example they might have unreliable electricals). The best robot isn't the best shooter, or the best defender, or the best picker upper, it's a combination of good components that work together well. In my experience, the only thing that would be able to take all the factors and spit out good, reliable, accurate rankings is an experienced human with a photographic memory.

Along those line the best elimination round robot is not necessarily the best qualifying robot. Remember the qualifiers and elims never quite play the same.

MikeDubreuil
03-22-2006, 06:35 AM
First, thank's Tom for posting that spreadsheet. It's an interesting tool.

I would agree that the seeding rounds give a rough estimate of the best robots at a competition. Suprisingly, in Tom's spreadsheet the top 8 teams in the real rankings stayed ranked within the top 8 in the scaled rankings. This means that the seeding system is working fairly well.

I would agree that top 8 teams can pick "oddball" robots and win a regional. This is because the current ranking system rewards high scoring offensive rounds. Which is unfortunate because the skills needed during the finals are a little different. Offense wins #1 seed, defense wins championships.

My original point was that next year there should be a rule that two teams from within the top 8 should not be allowed to allaince together during the finals. They are the best robots at a competition. This isn't communism. FIRST would rather have the final rounds be exciting for all teams rather that just the dominant alliance. Every alliance should have the same chance winning.

Tim Delles
03-22-2006, 09:03 AM
My original point was that next year there should be a rule that two teams from within the top 8 should not be allowed to allaince together during the finals. They are the best robots at a competition. This isn't communism. FIRST would rather have the final rounds be exciting for all teams rather that just the dominant alliance. Every alliance should have the same chance winning.


Yes but then you would be taking away from the qualifcation matches because if you know the number one seed will pick you if your not in the top 8 then all you would have to do to make sure you are picked by them is get out of the top 8.

My opinion on making the top 8 pick outside the top 8 will just make the qualifing rounds useless because if a team wants to be with another team bad enough they will just lose. And then were is the fun and enthusiasm in that for the kids and for the people watching?

MattB703
03-22-2006, 11:24 AM
...

Offense wins #1 seed, defense wins championships.

...


I don't know about that. It wasn't true last year and I don't think it will be true this year. We will see I guess.

StephLee
03-25-2006, 01:06 PM
I would agree that top 8 teams can pick "oddball" robots and win a regional. This is because the current ranking system rewards high scoring offensive rounds. Which is unfortunate because the skills needed during the finals are a little different. Offense wins #1 seed, defense wins championships.


I agree with everything you said (especially the last thing), except I have an example of something contrary to high scoring offensive bots being rewarded. At Pittsburgh (granted, the smallest regional, and therefor statistically less reliable), the top three seeds could not shoot for the high goal at all. All three, however, were pretty good dumpers, but their top quality IMHO was their strong defense.

Another point I think has already been brought up: the #1 and #2 seeds may have too much in common for them to do well by picking each other. The alliance that wins will be the alliance whose members best complement each other's strengths and weaknesses, as well as work together cohesively. From Chesapeake, the only loss for our #1 alliance (1629, 175, and 1184) was due to an offsides penalty that resulted from miscommunication, from not working together. We had three very different bots, two of which could shoot, dump, and play at least some defense. One could dump and play excelent defense. The thing I love about this year's game is the number of different strategies that can win. You have to adapt your strategy for every single game, more so than last year...and that's why I carried a giant bottle of Tylenol all weekend!

TheOtherTaylor
03-26-2006, 12:13 AM
From Sacramento/Davis Regional today, we (1351) were the #1 seed.

Our first pick was seed #18 (it may have been #13, those who know me know my memory for that sort of thing is a little fuzzy) which a -lot- of people were surprised by, but the pick was subject to huge ammounts of discussion and a lot of scouting by our best scouters. We had a way to get our own bot onto the ramp and win the round. The only rounds they could keep us from just dumping balls into the high goal was the time we were triple-teamed, and 649 worked perfectly as a high shooter unless they had someone on them...

But not to bore with strategy. A victor fried several mins before our first match, requiring us to use our timeout and a lot of frantic people to replace it and plug it in. I'm assuming because of the general panic, a pwm didn't get plugged in all the way (just to clarify, I don't blame anyone for this, it took hours to figure out where the problem was afterwards and multiple tries just to plug the dang thing in because it was sticking) and we never could figure out the problem on the field. Not having the left hand side of our drivetrain pretty much lost us two matches we would have easily won.

All it takes for the #1 seed to lose is a stupid system failure with bad timing to render their entire bot useless for twenty minutes.


[edited as per request and to correct a couple things, I typed this late]

EricH
03-26-2006, 02:13 AM
From LA, 2nd seed won, 1st seed was finalist.
Actually, first seed picking second seed may have the potential to go all the way. The record for the first and second seeds at LA: 9-0-0. By the end of the finals, it was 13-2-1, and they were simply outstrategized.

It's not the alliance that wins the finals, it's the strategy that is tailored to the alliance faced that wins.