Ladle 42 Voting Discussion

[INW]

[quote="Concord"]

Code: Select all

Ladle X

If
Team A - future Seed 1
def.
Team B - future Seed 3

Team C - future seed 2
def
Team D - future Seed 4

Then Ladle XI

Top Bracket
Team A
Team D

Bottom Bracket
Team B
Team C

Another thought as I read this post. I thought it was Seed 1 v. Seed 4 on the top bracket and Seed 2 v. Seed 3 on the bottom. If it isn't, then we have been doing it incorrectly all along.

[dariv]

If there is no play-off it should be randomised (and they should both be called "3rd seed"), if there is a play off the winner ("3rd seed") should enter the side of the bracket containing the 2nd place seed. Simple?

[Z-Man]

Another thought as I read this post. I thought it was Seed 1 v. Seed 4 on the top bracket and Seed 2 v. Seed 3 on the bottom.

I may be totally dumb, but that's also what the code posted by Concord says.

First, let me tell you why Concord is wrong. Then, I'll tell you why he's right. Then, why he's wrong again.

The purposes of seeding can be:
A. Ensure an interesting final.
B. Ensure interesting semifinals.
C. Seeds should fall to the best teams and stay with the best teams.
D. Rank the best teams.
If team performances are stable across ladles, our system accomplishes A to C. Not D. As has been pointed out, there is no known relation between Seed 3 and Seed 2 and 4. And I say there doesn't need to be. The point of giving Seed 3 and 4 the positions they are (or should be) given, the swapped positions where the team only losing to the final winner is set up to the loser of the final, is to give that team a good shot at making it to the final the next time. Assuming constant team performance, the system will stabilize to giving Seed 1 to the best team, Seed 2 to the second best team and Seed 3 and 4 to the third and fourth best team in alternating order. It will never establish a comparison between the two, but there is nothing wrong with that. It's also not unfair: the only difference between the slots Seed 3 and 4 get is the order in which they need to beat Seed 1 and 2 to win the final. Yes, the bracket structure of the next ladle is independent of the final outcome, but what is wrong with that? Demanding that every match influences the next ladle is a bit silly. A match for "third place" (it isn't, really; the winner may as well be the second best performer) would not hurt, but I doubt players will participate enough to make it meaningful.

Now, above, I assumed stable team performance. That is something obviously a bit detached from reality. How does the system perform theoretically under disturbances? Not so well, it turns out. If, for some reason, the fifth best team snags a seeded spot, they can only lose it if they are in the same 'quarter bracket' as the non-seeded team that's better than them, they are already protected from meeting the other three better teams. So it takes four ladles on average to kick them out. Likewise, it takes two ladles on average to remove the unjustly seeded sixth best team. You only have to assume a small number of flukes during each ladle, a small number of team quality changes, a small probability of newly formed 'all star' teams to topple the conclusion that the seeded spots go to the best teams.

But here's the thing: there's no fixing that. A 'third place' match will not change who gets seeded, neither will randomization. No matter what you do, quarter finalists will fluctuate, seeds will fluctuate. Property C is not a property that can be achieved under practical circumstances. I'd say the right way to go is to drop it. I propose this new system:

Seed 1 goes to the final winner.
Seed 2 goes to the final loser.
Seed 3 goes to the team that lost to the final winner in the semifinal. (So far, no change.)
Seed 4 goes to the team that lost to the final winner in the quarterfinal.
Seed 5 goes to the team that lost to the final winner in the octofinal (opening round with 16 teams).
So seeds go to those teams that were only beaten by the final winner. In a ranking, they would all be contenders for second place.

Seeded teams play in the following ladle in the bracket as follows:
Seed 1 goes to one halfbracket, everyone else to the other.
Seed 2 goes to one quarterbracket, everyone left to the other.
Seed 3 goes to one eights bracket, everyone left to the other.
Seed 4 meets Seed 5 in the opening round.
All other spots are randomized.

This definitely drops C and D. Seeds can no longer be considered rewards or achievements (look at Seed 5, all they needed to do was be in the right spot of the bracket randomly). Seeds are no longer stable; in fact, it's impossible for Seeds 3 and beyond to get seeded again next time. Still, seeded positions are advantageous: if a team were to drop its seeded position, chances are > 50% they'd land in the halfbracket of the winner, reducing their chance of making it to the final. Property A is superbly fulfilled: the second best team definitely gets a seed and will definitely meet the best team in the final of the next ladle if those two perform consistently. And we have two new, desirable properties fulfilled (and combined into one, 'it mixes things up'):
E: Seeds go to teams that didn't get a chance to compare directly in the last ladle, and they will compare directly in the next ladle.
F: The winner of the last ladle doesn't have to face the same teams again until the final.
(That also means the winner will get what can be considered a 'free pass' to the final, which you may not find desireable.)
Of course, the process can be cut off at any point, for an arbitrary number of seeded teams.

Or you could switch to just seeding two teams. Solves the problems and still keeps a high probability of an interesting final.

[Concord]

Another thought as I read this post. I thought it was Seed 1 v. Seed 4 on the top bracket and Seed 2 v. Seed 3 on the bottom. If it isn't, then we have been doing it incorrectly all along.

It is, I'll illustrate it again, in hopefully a clearer way.

Ladle 42 Semifinals (For now, we don't care about the current seeds, only the future ones)
Team A defeats Team B
Team C defeats Team D
Team A vs. Team B in the final

In the following Ladle one halfbracket will have Team B & Team C and the other Team A & Team D, regardless of the final. This means that the brackets are not necessarily equal (with the true 1&4 and the true 2&3) and as things continue of multiple Ladles become unequal.

To Z-Man's points. A third place game ensures that the bottom bracket has both teams that might be considered 2nd best. Together they have defeated or beat an opponent who defeated 3/4 of the previous bracket. This means that each bracket is at least initially equal in seeded strength. Z-Man is correct in his sentiment that the lowest number seed will always be meaningless and pose a trouble spot, as will seeds determined by which game you lost.

The real alternative is to go back to how we used to set the challenge board, and have captains self-select their spots.

[Z-Man]

A third place game ensures that the bottom bracket has both teams that might be considered 2nd best.

That's already the case for the simple swap. Well, *two of* the teams that might be considered second best. Seed 4 has lost against a team that lost against the best; they can be third best at most. If you let team B and D in a third place game, the best outcome would be that original Seed 3 wins against original Seed 4, acknowledging its place as a second place contender; if original Seed 4 wins, however, all that is achieved is that neither can legitimately claim to be second best, since both lost to someone who lost another game. Overall, you don't increase the chance that the second place in the weaker bracket half goes to another second place contender, you just gain information.
(Now, of course, you do increase the chance that the better of the two candidates gets the spot there, which, yes, improves the bracket balance or leaves it unchanged.)

Tossing a coin, by the way, shouldn't even be an option in a poll on that matter. It's pretty much guaranteed to make matters worse on average. To be precise, chances are 2 in 3 that Seed 3 would win a 'third place' match against Seed 4. Go math! Pick all 24 possible orderings of half finalists, and the third best team will already be in Seed Spot 3 in 16 cases. So only in one in six cases will randomization improve the bracket balance, in two in six cases, it will worsen it. (and in three of six cases, change nothing.)

[sinewav]

Excellent. As soon as you two agree on something we can edit the wiki and forget about adding a measure to the voting thread that will surely confuse the piss out of everyone.

[Lord Pein]

I skimmed most of this topic.

IMO there should be only two seeds; but the reason everyone wanted seeds so badly in the first place to make sure the brackets didn't end up like that one time when I did them and all the good teams were on one side, right?

[syllabear]

Team ability changes much more than the difference between 3 and 4th ranked seeds for the most part.

Zman, your alternate idea will eventually cycle through until all (or most of) the worst teams are on one side of the bracket in one ladle (the reason we had seeding in the first place).

Leave it the way it is for sure. 1v"4" and 2v"3". After a few ladles it will even out. Eventually if there are two teams that are stronger than the rest, they will meet in the finals always. It will also ensure that if there are 4 teams stronger than the rest, we will get intersting semifinals.

However, due to oracles rise and fall, it will take a few ladles to resettle itself (with R, SP, CT and unk in the top 4 IMO right now, depending on how YIDG performs next time, we might remove one of these for them)

[INW]

Another thought as I read this post. I thought it was Seed 1 v. Seed 4 on the top bracket and Seed 2 v. Seed 3 on the bottom. If it isn't, then we have been doing it incorrectly all along.

It is, I'll illustrate it again, in hopefully a clearer way.

Yes. I over read it.

I propose this new system:

Seed 1 goes to the final winner.
Seed 2 goes to the final loser.
Seed 3 goes to the team that lost to the final winner in the quarterfinal. (So far, no change.)
Seed 4 goes to the team that lost to the final winner in the octofinal.
Seed 5 goes to the team that lost to the final winner in the... opening round, with 16 teams.
So seeds go to those teams that were only beaten by the final winner. In a ranking, they would all be contenders for second place.

That would not work at all. So if a horrible team (lost 100-0 twice) to the winner of the ladle in the first round, they should get the 5th seed?
Anyway, it would be much easier to just keep it like we have it.

[Jip]

Seed 3 goes to the team that lost to the final winner in the quarterfinal. (So far, no change.)

Now seed 3 goes to the team that lost to the final winner in the semifinal.

[Z-Man]

Seed 3 goes to the team that lost to the final winner in the quarterfinal. (So far, no change.)

Now seed 3 goes to the team that lost to the final winner in the semifinal.

Oh right, wrote that down wrong. Will edit. It's supposed to be the same way as it is now up to that point.

Zman, your alternate idea will eventually cycle through until all (or most of) the worst teams are on one side of the bracket in one ladle (the reason we had seeding in the first place).

No. Only Seed 1 and one of the other seeds will keep their seed, the rest will be returned to the randomized population. Even assuming constand performance, only the top two seeds will be constant, the rest of them will be random (with higher seeds going to more skilled teams with higher probability). It's practically a complete shuffling every time, I'm pretty sure it is stagnation proof.

So if a horrible team (lost 100-0 twice) to the winner of the ladle in the first round, they should get the 5th seed?

Yep. Look what they get for it. Sure, they won't meet the last ladle's winner until the final, but to get there, they have to battle through all the other seeds.

[owned]

I think the biggest problem with z-man's solution is that effectively, it is a two-seed system. It has a lot of benefits, but two seed systems have always been shot down in the past. It is possible to maybe extend this system so it more represents a four-seed system.

It would look like this:

Seed 1 is winner of final
Seed 2 is loser of final
Seed 3 is loser to seed 1 in semis
Seed 4 is loser to seed 2 in semis
(what it is now)
Seed 5 is loser to seed 1 in quarters
Seed 6 is loser to seed 2 in quarters
Seed 7 is loser to seed 1 in round of 16
Seed 8 is loser to seed 2 in round of 16

Seed 1-4 are arranged like they are now.
Seeds 6 and 8 are on the same quarter and other eighth of seed 4, while seeds 7 and 5 are on the other quarter and other eighth of seed 3.

This is basically z-man's plan, but extended so it more represents a four-seeded system.

There are a few downsides to this:
-Seeds could lose their meaning in the minds of arma players as 5-8 are very easy to get.
-This is a little complicated.

But I think the upsides outweigh the downsides in this case.

[Concord]

We're confusing the point again. The idea was not to make great brackets, it was to make somewhat even brackets.

The rather small issue with the current system is that a upstart team is that when an upstart team comes out of the 4 or 3 seed quarter bracket, they often win a seed. They are able to maintain this seed and usually improve it until they fall apart, which weakens whatever bracket they're in. Upstart teams tend to enjoy only short periods of success.

Seeds are complicated. I believe there are more direct ways of getting our desired result.

This is a suggestion. A single person or a couple people simply make brackets with the design of making them even, fair and compelling. By signing a team up you acknowledge that person x will place you on the bracket. That person or group of people, the bracket-makers, are of course accountable to the community. This should work, if the community acknowledges the bracketmakers authority. The bracket-maker has no incentive to make bad brackets and has the obligation to write up a brief summary of each choice he made.

Another suggestion. It's essentially a stock market. Before the brackets are made, each player can wager in units of 1 on any team up to for 4 teams. They can only bet 1 on each team and no less than 1. Teams are valued as the sum of those bets on them. Teams are then seeded based on their stock price. These are reset after each ladle.

[INW]

I don't like that idea at all. Brackets are made to have upsets. Remember #16 (Illinois) upset over #1 (Kansas)? NCAA Basketball? Brackets are made to be "fair" but have upsets. Look at an Major League playoff. This year for example. The Seahawks upset the defending champion Saints. The Jets upset the Colts, then the Patriots. Brackets are full of upsets. One year, or ladle, is independent of the other. A team may be number 1 one ladle and number 32 the next. That is how tron rolls. Lets get over that and keep it the way we have it...I just don't understand what is so bad with the current system...

[Concord]

nothing about any system prevents upsets. What did you think you meant?