Re: Am I the only one who LOVED the seeding system this year?
 

Are win-loss records stored anywere? Someone mentioned that more matches are played now so i was wondering if there is someway to compare standings using the old system and new system. Of course i am not that smart and just throwing that idea out there for someone to fill thier off-season time.:)

Re: Am I the only one who LOVED the seeding system this year?
 

FIRST used point based seeding prior to 2003, with W-L in eliminations. They went through a few different formulas, probably some of which were posted in here (I don't remember). People complained no matter what the forumla was (both about the formula and about the difference in play of elims and quals), so in 2003, FIRST used point based ranking in both the seeding and eliminations rounds. In the eliminations, you would play 2 matches and which ever alliance had the higher seeding score combined would move on. This had the consequence of the alliance that won the first match would not play to win the second match, but play to keep the scores low. If the second matches scores were low, they would advance based on the combined score. If they played all out in the 2nd match and lost, they might not advance.

After that, FIRST changed to W-L for both the seeding and elimination rounds, until this year.

Re: Am I the only one who LOVED the seeding system this year?
 

In 2007 and 2006, it was illegal to score for your opponents. It would not have been that difficult for referees to spot obvious attempts at scoring for the other side in Breakaway, but the GDC chose to leave this mechanic in the game.

I've been thinking about it, and here is my stab at a "Unified Theory of FRC Seeding Algorithms"...

The variables
WPoints = unpenalized points scored by the winning alliance
LPoints = unpenalized points scored by the losing alliance
WPenalties = winner's penalty points
LPenalties = loser's penalty points
WBonus = bonus seeding points for winning
TBonus = bonus seeding points for a tie
WSS = winner's seeding score
LSS = loser's seeding score
TSS = seeding score for both alliances in case of a tie (WP = LP)
C1, C2, C3, C4, C5, C6, PW, PL, PT = Constants

The formulas
Winner is defined as the alliance with the most points minus penalties.
WSS = C1*WPoints + C2*LPoints + WBonus - PW*WPenalties
LSS = C3*WPoints + C4*LPoints - PL*LPenalties
TSS = C5*WPoints + C6*LPoints + TBonus - PT*WPenalties

(for the last example, let WPenalties be YOUR penalties only)

Examples
Looking at the above equations, let:
C1 = 1
C2 = 2
C3 = 1
C4 = 0
C5 = 1
C6 = 2
PW = 1
PL = 0
PT = 1
WBonus = 5
TBonus = 0
This is the Breakaway scoring system.

Another example, W-L-T:
C1 = 0
C2 = 0
C3 = 0
C4 = 0
C5 = 0
C6 = 0
PW = 0
PL = 0
PT = 0
WBonus = 2
LBonus = 1

So clearly we can play with these constants to generate anything from score-doesn't-matter-as-long-as-you-win to full on coopertition and anything in between.

What is the optimal choice of the above constants? Most importantly, which should be zero and which should definitely be non-zero (regardless of game)?

C1 = 2
C2 = 1
C3 = 1
C4 = 2
C5 = 1
C6 = 2
PW = 1
PL = 1
PT = 1
WBonus = Dependent on score distribution; could be either 0 or non-zero
TBonus = 0

What would peoples' objections to the above consist of? Each team always benefits from additional scoring for itself AND for their opponent scoring regardless of the current score (both winner and loser get a contribution from both alliances' scores). There is still some incentive for defense, as given a finite number of game elements, each team is always better off scoring for its own alliance (so to parlay into Breakaway terms, punting the balls from your opponent's offensive zone down to yours would help your overall score). Each team is penalized for the infractions they commit (a flaw in Breakaway IMO, as the loser penalties disappear). High scoring and close matches are necessarily going to generate more seeding points than blowouts and low scoring affairs.

Re: Am I the only one who LOVED the seeding system this year?
 

A number of locations have Win/Loss, data but the easiest and most complete data for this type of calculation is probably Ed Law's Scouting Database. Another good option (although it doesn't have Champs data) would be the 1114 Scouting Database

Re: Am I the only one who LOVED the seeding system this year?
 

I would think that rewarding exceptional performance would be a good thing, but there seems to be a cultural stigma against having lopsided outcomes. I believe that encouraging someone to either do less than their best or otherwise artificially create a more aesthetically pleasing final score is wrong and has no place in a seeding system. Sure it can be humiliating to lose by a wide margin (been there), but it is also an exceptional motivator.

And while I'm on my soapbox... I really dislike using pre-penalty scores for anything. No one should benefit from a penalty. Then again a penalty should be reserved for behavior that you really want to inhibit and even then should be proportional to the damage incurred. I'm not much for calling penalties on robots that happened to be dragging chains as I saw multiple times this year. Sure it is against the letter of the rules but where is the advantage? Besides it seems to me that the team is already being penalized.

Re: Am I the only one who LOVED the seeding system this year?
 

If we're looking at a new formula, or open to new ways of ranking teams, why not go to a tiered system.

It could work... roughly... something like this.

All teams start out in tier 3, with randomly assigned partners and opponents.

The teams that win move up to tier 2. The teams that lose move down to tier 3. Teams that tie or don't play in the first round (due to the total number of teams not being divisible by 6) will have to be dealt with, possibly by coin toss, or by remaining in tier 3.

In following rounds teams move up... or down... depending on how they do, but always play against teams from within the same tier.

Ranking could take place in a number of ways... you could sort by tier, then subsort by w-l-t or some other measure, or you could make a win in a higher tier worth more points than a win in a lower tier and sort by points.

So yes, it would take a bit of thinking to work through the bugs and come up with an algorithm that works. I'm not really worried about the fine details at the moment, as they can - likely - be worked out.

The general idea, however, is that stronger teams will play with and against stronger teams, while weaker teams will play with and against weaker teams. After two or three rounds the top tier matches will begin to look like elimination round matches. In lower tier matches teams that wouldn't normally have a chance to effect the outcome of a match would get to have some control of their own destiny. Almost no one would go home undefeated, and almost no one would go home without a win. (Even more so than now.)

So take this as a brainstorming suggestion... not a concrete proposal... and consider the concept rather than the mechanics.

Does the concept have appeal?

Jason

Re: Am I the only one who LOVED the seeding system this year?
 

Jason,

That is very similar to a Swiss system of match pairing.

The biggest issue I see in implementing this type of system for FRC are that teams won't know when their next match is until at most an hour or so before it occurs (if you use rankings after the previous round to pair the next round, ignoring the current round). Another issue is that the system becomes much more complex when you add in the various desirable factors of FRC pairing such as reducing repeat pairing, balancing alliance colors, minimum match spacing, etc.

It's a cool idea, and perhaps one worth looking into, but it's certainly not without drawbacks.

Re: Am I the only one who LOVED the seeding system this year?
 

Thanks for the link to the Swiss System... I knew I'd seen tournament structures like this before, but didn't know the name.

I agree with the complexities, though...

Jason

Re: Am I the only one who LOVED the seeding system this year?
 

I think that you should get points for the points your robot scored. This is to make sure the good teams are at the top. That is what was bad in Atlanta for us. We had 2 bad matches where the other robot shut down on us and we were 8th.

Here is my plan:

W= W+2L+3P
L=W+3P

P is the points your robot scored.
W is winner's points.
L is loser's points.

Re: Am I the only one who LOVED the seeding system this year?
 

Overdrive is the only game in recent memory that was scored in a manner that I would consider it feasible for field personnel to accurately track this and even then it's questionable. You would also need extra volunteers to fill these roles.

Verdict: Probably not going to happen

Re: Am I the only one who LOVED the seeding system this year?
 

love the idea, but one big problem/potential complaint is what about defenders? By this defenders would receive less points. Now this could be a gamble teams would have to consider (play strong defense and hope to get picked for elims). But that's not entirely bad.

I liked this years system in that the top teams were at the top. In the past I didn't care about rank cause it didn't matter until late Friday. This year it was cool competing with other teams trying to get the top seed.

Re: Am I the only one who LOVED the seeding system this year?
 

That's pretty stupid for most teams, since that makes the assumption that any action taken by a robot that doesn't directly result in a scored point doesn't contribute to the alliance. Feeders were kinda pretty big this year.

Re: Am I the only one who LOVED the seeding system this year?
 

By your metric, we would have been much lower in Archimedes. Our specific strategy was to feed balls to the home zone. Feed until surplus, then snipe from the midfield. If surplus is gone, re-stock home zone. If we would have used your metric, we would have been fighting with our own alliance members so that we could be the ones to score the balls. Instead of 3v3 or 6v0, it would have turned into 1v1v1v1v1v1.:confused:

Re: Am I the only one who LOVED the seeding system this year?
 

Basically, it comes down to doing a regression fit on which of a set of score scenarios should be most heavily rewarded.

Consider:
A. Winning 20-0
B. Winning 15-5
C. Winning 11-9
D. Winning 10-0
E. Losing 0-20
F. Losing 5-15
G. Losing 9-11
H. Losing 0-10
I. Tie 10-10
J. Tie 5-5

Which of the above should be worth more than others?

Should winning 20-0 be worth more, less, or the same as winning 11-9? 15-5?

Should winning 10-0 be worth more, less, or the same as losing 9-11 or 0-20?

If I lose 0-10 or 0-20, should my score change at all?

If you can come up with a concrete answer to the above, you can generate a linear scoring function that will behave the way you want it to.

Re: Am I the only one who LOVED the seeding system this year?
 

Very interesting post Jared; really gets the brain churning...

When I go for an "offense is best" approach I end up with something like:

A=B=C>I>G>D>F=J>E=H=zero

but I'm finding that I have an innate bias to favor wins over ties over losses regardless of the score, so relationships like G>D are troubling. Also, I find that I could easily change my mind if the scoring mechanism were different than in Breakaway.

Alternatively, we could just seed everyone based on their least squares contribution like 1114 does in their database. Of course, that might be a bit difficult to explain to casual observers and there are linear algebra issues with small numbers of matches.