9.3.4 Match Seeding Points

[kirtar]

Originally Posted by leafy

Everyone being your alliance, or everyone being people in the competition? In the former, yes. In the latter, no; The cooperition bonus is only deterministic of the seeding rank with people whom you tie seeding points. If you don't score any points, the maximum points the other team can get (will get) is increased (even doubled or more) because you aren't going after any balls.

Everyone in this case == everyone (and I repeat... everyone), so it would be the latter

Originally Posted by EricH

Actually, if you had read the thread, there are two points of view. You have one of them.

The other is that the coopertition bonus is added to your seeding points (hence the "bonus" part), to produce a much higher seeding point total. Instead of the winner getting their own score as seeding points, the winner gets their own score plus their coopertition bonus as seeding points.

Also note:
9.3.7
Now, see this:
9.3.5
In other words, the Coopertition Bonus is in fact seeding points, which in fact are added to the seeding score, which is your ranking.

Winners have a distinct advantage here, especially in close games.

I agree. How the manual is worded at this specific point in time, it is abundantly clear that the Coopertition Bonus is a part of your total seeding score as well as the first tiebreaker. It is a bonus not a separate score, otherwise it wouldn't say that it was a bonus of additional seeding points.

[Joe Johnson]

The Seeding of teams is a tricky thing, even when a you have many many comparisons between teams (think of MLB with a 160ish game season). It is even more difficult when you have 7 or 8 matches for each team and you have multiple teams compared each match.

FIRST has the added problem of trying to encourage GP while still having a game worth playing.

The proposed system for 2010 is not that much different from some of the systems that we have had in the past. I think there are going to be anomalies but there always are anomalies.

There are always going to be teams in the top 8 slots that are "undeserving" in someone's eyes.

The drafting process is the great equalizer... ...which is a topic for another thread...

Joe J.

[Richard Wallace]

This seeding system appears to be an embodiment of the invention protected under US Patent 7507169.

The inventor? Dean Kamen.

[mathking]

This a fascinating exercise in applied game theory. I actually think collusion (that is the game theoretic term, not meant to imply anything negative) between the two alliances, where one scores 0 points and they all try to score as many as possible for the winning side so that all six teams get the same number of seeding points is not a bad strategy. There is definitely a problem in that there is an incentive for the "winning" side to cheat at the end of a match. But the fact that elimination competition is based solely on wins and losses means that if you always use this strategy you will likely not be as prepared for elimination rounds as a team that does not use this strategy. Nonetheless, there are a lot of chances for such collusion/coopertition.

[kirtar]

The seeding system has always encouraged the winning alliance to ensure that its opponents had a somewhat high score, but to a lesser degree. Previously the opponent's score was used as the first tiebreaker. Now it's used as a part of the initial round as well as the first tiebreaker.

Originally Posted by mathking

This a fascinating exercise in applied game theory. I actually think collusion (that is the game theoretic term, not meant to imply anything negative) between the two alliances, where one scores 0 points and they all try to score as many as possible for the winning side so that all six teams get the same number of seeding points is not a bad strategy. There is definitely a problem in that there is an incentive for the "winning" side to cheat at the end of a match. But the fact that elimination competition is based solely on wins and losses means that if you always use this strategy you will likely not be as prepared for elimination rounds as a team that does not use this strategy. Nonetheless, there are a lot of chances for such collusion/coopertition.

If you're working with the other alliance, it's much better for both alliances to rig up a tie.

[AllenGregoryIV]

This seems like an interesting system to me and I am not sure if I like it yet.

In a situation where the match is close it will always be better to score for your own team to try and win the match thus getting your penalized score and double the losing teams score.

However once the game gets lopsided say 10-0. The losing team, if they believe they cannot win, should attempt to score on the winning team’s goal because those goals will be given to their seeding score. The winning team should also try to score for their opponents because those scores will be worth double in their seeding score.

In the past it has been advantageous for the team that is winning to score for the losing team but never, to my knowledge, has it been better for a losing team to try to make the score differential higher. In past systems the losing team would get their score as their seeding score and would always continue to score for themselves.

[ScottOliveira]

Originally Posted by kirtar

If you're working with the other alliance, it's much better for both alliances to rig up a tie.

It's actually much better for both alliances to say this, and then one to cheat at the end (the cheating team will receive more points). However if a team cheats, it's less likely to be able to use this strategy on later matches, as other teams won't trust them. However, in the last match an alliance plays it is more likely to cheat, because the trust matters little after that point. Using game theory again, specifically backwards induction, it's possible to show that everyone will cheat every single game (game theory doesn't account for honor, only an effort to maximize points).

[kirtar]

Originally Posted by ScottOliveira

It's actually much better for both alliances to say this, and then one to cheat at the end (the cheating team will receive more points). However if a team cheats, it's less likely to be able to use this strategy on later matches, as other teams won't trust them. However, in the last match an alliance plays it is more likely to cheat, because the trust matters little after that point. Using game theory again, specifically backwards induction, it's possible to show that everyone will cheat every single game (game theory doesn't account for honor, only an effort to maximize points).

Maybe I should rephrase that. It would be most beneficial to both alliances to do so, but that is an ideal scenario which will likely not happen due to competition. It is the situation in which the gain of one team is not necessarily at the expense of the other per say (or at least not as large of an expense both in overall seeding and the tiebreaker).

[Matthew2c4u]

Originally Posted by EricH

Actually, if you had read the thread, there are two points of view. You have one of them.

The other is that the coopertition bonus is added to your seeding points (hence the "bonus" part), to produce a much higher seeding point total. Instead of the winner getting their own score as seeding points, the winner gets their own score plus their coopertition bonus as seeding points.

Also note:

Now, see this:

In other words, the Coopertition Bonus is in fact seeding points, which in fact are added to the seeding score, which is your ranking.

Winners have a distinct advantage here, especially in close games.

Basically If you win you get coopertiton plus seeding, Lose = seeding
Secondary ranking factor = Highest coopertiton from any one match.
3rd is a point value based on hanging robot points for your alliance from every match that you played in combined.

[mathking]

If you're working with the other alliance, it's much better for both alliances to rig up a tie.

There is an incentive to work for one side getting all the points so that both alliances get the same (hopefully high) score, but there is an incentive for the side with all the points to try to score on the side with none at the end because then they will get more points than, and move up in the rankings relative to, the teams on the losing alliance. On the other hand, you can only "cheat" like this once, because other teams will no longer trust you.

If everyone were to try for the n to 0 score strategy, it could end up being quite arbitrary as to who ended up as the top 8 teams for alliance selection. And it might well not be clear who those top 8 teams should select as partners. Still, I think we will see some examples of coopertitive collusion.

[Matthew2c4u]

Coopertition bonus is always worth the win unless you have a very low chance then collusion is valued.

[Jimmy K]

For any fixed total number of points (S) that are scored in a match, if you score x points, then your seeding score (if the game is penalty free) is:
S - x (if x < S/2)
2S - x (if x >= S/2)

Within each case, your seeding score decreases with the number of points you scored. The result is that the "optimal" options are to get shutout, or to win by the smallest possible margin (or even tie).

If the game becomes lopsided to the point where one alliance is guaranteed to win, then:
The winning alliance scoring 1 pt gives both the winning alliance and the losing alliance 1 seeding pt.
The losing alliance scoring 1 pt gives the winning alliance 2 seeding pts and the losing alliance 0 seeding pts.

This makes for a balance of competition when the game is close enough to be "competitive" and cooperation (scoring for the other team) when the game is decided.

Also, defensive robots are punished more in this game. A team that gets shut out 0 - 20 in every game beats a team that wins every game 19 - 0.

IMO, this is worse than last year's <G14>, but at least this doesn't apply for elimination matches.

[Lil' Lavery]

Originally Posted by mathking

This a fascinating exercise in applied game theory. I actually think collusion (that is the game theoretic term, not meant to imply anything negative) between the two alliances, where one scores 0 points and they all try to score as many as possible for the winning side so that all six teams get the same number of seeding points is not a bad strategy. There is definitely a problem in that there is an incentive for the "winning" side to cheat at the end of a match. But the fact that elimination competition is based solely on wins and losses means that if you always use this strategy you will likely not be as prepared for elimination rounds as a team that does not use this strategy. Nonetheless, there are a lot of chances for such collusion/coopertition.

This is essentially what I was going to post. 6 robots working together to shuttle the balls from the midfield and score them on the same goals is has a high probability of achieving a higher score than a 3v3 match with defense, ball possession fights, traffic jams, etc.

Obviously the strategy won't be applicable to all robot designs (defensive, and even potentially end-game specialists). It may well also run into issues dealing with <G29>, depending on the implementation and robot design. And obviously not every alliance will be willing to do it. And the tie-based scenario is much higher-reward (though also higher risk, both in terms of cheating and just honest mistakes and miscues).

However, I can guarantee you we will see it attempted at least once per regional event.

[Matthew2c4u]

So if your going up a set of excellent teams and your alliance partners aren't @ the same level. You simply dont score @ all and protect your goals so they cant score and your seeding points come out to
SS(your seeding points)
SSO(your opponents seeding points)
WP(Winning teams points)
LP (Losing teams points)
PP(penalty points)
SS=WP
SSO= WP+ (2xLP) - PP
So basically if you help your opponent score and hold down any scoring on your goals to 1 or 2 you will get more seeding score than your opponent if they get double as many penaltys as you let goals by, Which is totally feasible if they arent prepared for your strategy.
Also if you help them score you will definatly get more points than you would have if you were normally facing them. So it helps your alliance and it slightly hurts they're alliance but they still get more points for winning. Its like acting symbiotically with a slight drawback.
My guess is it will be team updated to be fixed quite fast.
- my 2 cents.
This is all theorectial i doubt it will every be employed.

[StevenB]

I agree with MathKing and Lil' Lavery that a 6-team cooperative effort could be very fruitful. Let's try some numbers, ignoring the tower bonuses for the moment.
1) Suppose that in a "normal" round, the Red Alliance can score 8 points against the Blue defense, while the Blue Alliance can only score 3. Red gets 8 + 2*3 = 14 seeding points; Blue gets 8 seeding points.

2) In a second scenario, Red scores 8 times, 6 for Red, and twice for Blue. Blue scores 3, making the final score 6 to 5. Red gets 6 + 2*5 = 16; Blue gets 6.

3) Now suppose the alliances cooperate - this makes both sides far more productive. I'll suggest that Red can now score 12 balls, while Blue scores 4 (in Red's goal). Red gets 16 + 2*0 = 16 seeding points; Blue also gets 16 points.

I think playing this way is highly advantageous for both sides. Seeding-points wise, option 2 is the best win for Red. However, option 3 doesn't come with the risk that Blue could make a sudden comeback and win with the help of Red's two goals. This is where the tower points might really make a difference. If Blue had a stronger defense, the balance tips further in favor of option 3. Even if you can beat an alliance handily, having their help might still put you ahead.