NationStates

I suggest that the Baptism of Iron gets held every World Bowl regardless of the amount of new nations.

Oh we also have a shirt new that says, We've been to the World Cup 3 Times, thats 3 more times then the Rockets.

Dancougar wrote:I'm actually not sure how you would drop off the world... Elos aren't weighted over time, so if you're at 1408 and you disappear, that rating isn't going to degrade until they play someone again or they're manually dropped

(emphasis mine) That's what happens. I only track ratings in a sliding window of the last three World Bowl cycles, so if you haven't played an individual game (that I know of) within the last three cycles, you are no longer included in the ratings--otherwise I'd be keeping track of ratings for hundreds of countries that don't exist and never will show their face again, which would make the list unmanageably large.

Here's what I've been doing, in a nutshell: Everyone starts with a rating of 1500. Come next World Bowl, those with an established rating carry it over, while newcomers get a new 1500 rating. Come time for WB4, everything from WB1 is sliced off the spreadsheet, and ratings are now calculated with everyone in WB2 starting with a 1500 rating, including those who had ratings from WB1, while those who participated in WB1 but not WB2 or WB3 or WB4 no longer have any games being included in the calculation and so are dropped from the list. When WB5 rolls around, WB2 is dropped from the front end and only games in WB3, WB4, and WB5 are included, and so on.

Incidentally, it IS possible to degrade ratings over time, by using a higher K-factor for more recent World Bowls than for more distant ones--but it would mean that not only are rating increases smaller, but rating decreases aren't as big either. This is also how you would weight different tournaments, or individual rounds within tournaments (something I happen to be generally opposed to with a couple of exceptions, but that's a different discussion, and of course any Elo proposal adopted by the WB community wouldn't necessarily have to match my own personal preferences).

Bluth Corporation wrote:Everyone starts with a rating of 1500. Come next World Bowl, those with an established rating carry it over, while newcomers get a new 1500 rating. Come time for WB4, everything from WB1 is sliced off the spreadsheet, and ratings are now calculated with everyone in WB2 starting with a 1500 rating, including those who had ratings from WB1, while those who participated in WB1 but not WB2 or WB3 or WB4 no longer have any games being included in the calculation and so are dropped from the list. When WB5 rolls around, WB2 is dropped from the front end and only games in WB3, WB4, and WB5 are included, and so on.

Hmm, makes sense, although then you end up in a situation where newcomers get a jump over teams which have performed poorly over the past three Bowls, and we have similar nastiness to what happened on MD1.

Bluth Corporation wrote:Incidentally, it IS possible to degrade ratings over time, by using a higher K-factor for more recent World Bowls than for more distant ones--but it would mean that not only are rating increases smaller, but rating decreases aren't as big either. This is also how you would weight different tournaments, or individual rounds within tournaments (something I happen to be generally opposed to with a couple of exceptions, but that's a different discussion, and of course any Elo proposal adopted by the WB community wouldn't necessarily have to match my own personal preferences).

I think if we're going to use any kind of Elo formula, it is going to need some smarter (or dumber, depending on how you look at it!) way of handling the "revolving door" of competition that we have. The existing system is better-designed to handle that, while Elo ratings are applied with more permanence, say, to a chess player who is going to be active for probably twenty or thirty years, or to a national football team which will exist as long as the nation does, and that's pretty stable. We don't have that luxury.

How about this? Currently, the existing formula awards 0.25 points per victory and degrades the score over time, as desired. Why don't we apply Elo-like weights to that number based on the relative strength of the teams? For example, it turns a Dancougar win over Steroga into maybe 0.20 instead, while Steroga gets 0.30 if it gets an upset. (This admittedly does not have the same convergence properties as Elo in that only the winning team gets ranking points, but does have the advantage of being simpler to calculate.)

Dancougar wrote:
Hmm, makes sense, although then you end up in a situation where newcomers get a jump over teams which have performed poorly over the past three Bowls, and we have similar nastiness to what happened on MD1.

In an earlier post, I outlined a means of dealing with that: basically, it involves calculating the LOWEST POSSIBLE rating attainable given the number of games possible over the last three World Bowls, which is just 1500-(K*number of preliminary-round games). Or you could just give all newcomers a rating X points lower than the lowest currently-held rating, etc.

The World Bowl VII wiki page is now complete, at http://wikistates.outwardhosting.com/wiki/World_Bowl_VII. Thanks to Qazox for keeping the playoff bracket updated when I couldn't!

I'll be finishing the page on Dancougar's host bid (unless you want to do it yourself) and doing the page for the BoI (VII or II) over the next couple of days.

Bluth Corporation wrote:I'll be finishing the page on Dancougar's host bid (unless you want to do it yourself)

Go ahead and do whatever you want with it; I've decided that any future work I do is going to be over at NSwiki, and in any case, so far I haven't been bothered to write more than one article that isn't a stub.

And you could write stuff like "Dancougar is a huge jerk" and I wouldn't ever know!

Bluth Corporation wrote:In an earlier post, I outlined a means of dealing with that: basically, it involves calculating the LOWEST POSSIBLE rating attainable given the number of games possible over the last three World Bowls, which is just 1500-(K*number of preliminary-round games). Or you could just give all newcomers a rating X points lower than the lowest currently-held rating, etc.

I just want to check something to make sure I'm understanding the system correctly.

Let us assume that everyone started with a rating of 1500 at the start of WB1 (which I think you said was actually the case, so that's not really much of an assumption).

Let us then also assume that the lowest possible rating a team could end up at after the first three World Bowls is 1000. I doubt that's the actual value, but let's go with that just for the sake of illustration.

Your proposal, then, would have a brand-new team coming in for WB4 start with a rating of 1000.

Based on that, this would then make (all other factors being equal) the lowest possible rating by the time WB7 rolls around equal to 500. Thus, a new team for WB7 would start at 500.

If I'm understanding this correctly, wouldn't this have the effect of putting new teams at an ever-increasing disadvantage over time? Or (probably more likely), have I just missed an important detail somewhere?

I am absolutely not a fan of using elo's as our official ranking system. They are biased towards defensive teams.

From Bluth on jolt thread:
For those of you who understand the Elo system, what I did was, instead of assigning a "score" of 1 for a win and 0 for a loss, the "score" is now equal to the fraction of the points in the entire game scored by a given team. So in a game that ends with a score 15-5, the winner's "score" for Elo purposes will be 15/20=0.75, and the loser's will be 5/20=0.75.

Given this logic, a team winning 7-3 receives .7. A team winning 28-21 will receive whatever 28/49, which in quick math in head is about .56. Much lower, and if these results were consistent with teams, the defensive team will really begin to have an advantage. In every case, a winner in a low scoring game will benefit.

I will never support this.

Margin of victory doesn't mean squat. Ultimately, you win, you lose. (Or you tie in an environment that permits it.) I would be very much against any rating system that factors in victory margins.

Newmanistan wrote:From Bluth on jolt thread:
For those of you who understand the Elo system, what I did was, instead of assigning a "score" of 1 for a win and 0 for a loss, the "score" is now equal to the fraction of the points in the entire game scored by a given team. So in a game that ends with a score 15-5, the winner's "score" for Elo purposes will be 15/20=0.75, and the loser's will be 5/20=0.75.

That's interesting, because that's not how the world soccer Elo rankings work at all. The result of the match is still 1 or 0 (or 1/2 for a draw); the way that team strength is taken into account is that the stronger team has an "expected result" closer to 1. The difference in the actual result and the expected result is multiplied by two weights, one for margin of victory and one for importance of the match, to calculate the value of a match.

But wouldn't the margin of victory only make a difference in blowouts? Ultimately, when RL rankings are done, isn't MoV taken into account as well (especially in college FB rankings, for instance)?

The Fanboyists wrote:But wouldn't the margin of victory only make a difference in blowouts? Ultimately, when RL rankings are done, isn't MoV taken into account as well (especially in college FB rankings, for instance)?

In terms of having a real effect on the match value, yes, but in theory all wins could be worth something depending on how the bonus weight is calculated. In football, for instance, a one-goal win doesn't add any extra weight and a two-goal win provides a 1.5x boost. And for a three-goal win or better, they calculate a weight according to (11+N)/8, where N is the margin of victory.

So what's wrong with using the MoV in determining ratings, then?

I never said I had a problem with it... since I call "not me" on "who's going to calculate all this stuff", I'm fine with the consensus.

Please don't make this like college football. Margins of victory shouldn't mean a thing except for point differential during a qualifying go around.

Why shouldn't a heavy underdog holding coming up short but nonetheless making a valiant effort in a 28-26 loss to a powerhouse get more credit than the same underdog just rolling over and being trounced 49-3?

Bluth Corporation wrote:Why shouldn't a heavy underdog holding coming up short but nonetheless making a valiant effort in a 28-26 loss to a powerhouse get more credit than the same underdog just rolling over and being trounced 49-3?

In each case, they lost. That's why. I played three sports in high school, there was never a such thing as a moral victory. Either you win, or you lose. (or draw)

Dancougar, I didn't mean you necessarily, I just knew that others had raised an objection to it;

And yeah, there isn't a such thing as a moral victory (also speaking from experience as well), but when rankings are done, even in high school where I am, how well a team plays in any given game is given regard; if nothing else, how hard other teams will try against another team is based in a lot of cases on it.

So?

The point of a rating system isn't to measure wins and losses--that's what won-loss records are for. The purpose of a rating system is to measure strength, and a team that can run neck-and-neck with a powerhouse is clearly a stronger team than a team that gets routed, even if they both lose.

Cassadaigua wrote:

Bluth Corporation wrote:Why shouldn't a heavy underdog holding coming up short but nonetheless making a valiant effort in a 28-26 loss to a powerhouse get more credit than the same underdog just rolling over and being trounced 49-3?

In each case, they lost. That's why. I played three sports in high school, there was never a such thing as a moral victory. Either you win, or you lose. (or draw)

I don't know Cass. Every year when Garden Spot Almost beats Manheim Central(in football) and we dont get slaughtered when we have a good team i consider it a moral victory.

I didn't feel any different after losing 13-0 to Qazox then I did in losing 52-0 to them. Elo's are nice for fun comparisons not for actual rankings.

You may not have, but that doesn't change the fact that, all else being equal, a smaller loss indicates a stronger performance than a bigger loss--and it is STRENGTH that ratings are created to measure.

Although it is true that measuring in absolute terms, where a win is a win and a loss is a loss, also calculates the same thing, just not with the kind of precision you're advocating.

Please keep in mind that it's quite possible to use Elo with just a straight win/loss/draw rather than fraction of total points scored, which totally eliminates the objections of Cassadaigua and Newmanistan while still maintaining the other advantages Elo possesses over the current system.

That said, I don't understand the objection against margin/degree of victory since it is very clearly a major indicator of how strong a team is. If all you care about is wins and losses, then just use won-loss record. But ratings measure strength, and how big the wins (or losses) are is a very important part of that.

Newmanistan's objection, on the other hand, is quite valid and one I've struggled to deal with for some time, but I haven't figured out a way yet. But again, it's quite possible to implement Elo just using straight up wins and losses. It's less than ideal, but because of the other advantages of Elo I think it'd still be better than what we have now.