SquintGnome wrote:

itsnotatumor wrote:

As it stands I see it as a disincentive to play certain maps or players. Every time I get in the upper teens/low 20's in the rankings I can get point swings of 300+ a day.  I need to win 2-4 games just to make up for 1 loss. I noticed that I started looking carefully at who's in a game and skip it if there's too many low ranked players.  It also seems that high ranked players tend to play less games. Is this the reason, or do people just cut back overtime?

 
+1

-1

Can someone explaind to me why rating systems that are not 'zero-sum' are bad.  I realize that ratings overall would go up, but not sure if that is a bad thing.

If we did not have that constraint.  We could have a system where a 1000 player who beats a 2000 player gets 40 points while the 2000 player loses 10. 

SquintGnome wrote:

Can someone explaind to me why rating systems that are not 'zero-sum' are bad.  I realize that ratings overall would go up, but not sure if that is a bad thing.

If we did not have that constraint.  We could have a system where a 1000 player who beats a 2000 player gets 40 points while the 2000 player loses 10. 

AttilaTheHun wrote:

Any rating system will have difficulty because most of these games are luck based.  So even if you create a system that's ideal, the tendencies of top rated players will still be to avoid luck-based games against lower-ranked players.  Chess relies less on luck and so methinks would be a bit easier to design an ideal ranking system.

From the point of view of the rating system, non-luck based games are treated as probabilistic events the exact same way luck-based games are. The rating system has no way to get inside a player's head. It can only use results. If one player wins 60% of the time versus a lesser skilled opponent who wins 60% of the time against an even lesser skilled opponent, a good rating system will reflect that skill difference whether the game is luck-based or not. (A skilled player is perfectly happy playing the luck-based game because the skill will be reflected in the rating even though some losses will be taken due to luck.

As btilly mentioned, rating systems have been applied successfully in luck-based games already. We can easily find luck-based games that involve more skill separation than some non-luck-based game. (Backgammon versus Five - Backgammon is harder!)

But I think you have something like global ratings in mind. Doing something meaningful there is probably hard. But within a single game type, a single board, a good rating system will reflect the skill differences very well, whether it is luck-based or not.

SquintGnome wrote:

Can someone explaind to me why rating systems that are not 'zero-sum' are bad.  I realize that ratings overall would go up, but not sure if that is a bad thing.

If we did not have that constraint.  We could have a system where a 1000 player who beats a 2000 player gets 40 points while the 2000 player loses 10. 

The reason for it is to avoid ratings inflation. In an inflationary system, a mediocre player who plays enough can keep up with the rising average rating, while a skilled player who plays, but not enough, will barely keep up. That's the usual reasoning, but modern systems have abandoned this strict requirement!

Glicko is not zero sum unless the two players' rating have the same reliability. It doesn't lead to serious inflation because the players who play a lot will engage in zero-sum events. With Glicko, when an unknown 1000 player beats a well-known 2000 player, more points are gained for the 1000 player than are lost for the stable 2000 player. If, however, the 1000 player is also a stable player, the same amount of points are gained/lost for both players. (AND, I know you will like this part - there is a cap to what you can lose in a single game based on how reliable your rating is.)

Actually Glicko over time is deflationary.

The reason is because unknown 1000 players come in, lose, then wind up with a reliable rating below 1000.  Then they start to improve and suck rating points out of the system.  This known tendency is particularly noticeable with clumps of people who play each other a lot while improving.  (In the real world, consider college towns to see this happen.)

I also agree with Hugh that it is easier to fit a rating system to a board than globally.  But we tend to have more global games, and I don't think it worthwhile to have per board rating systems.  The moral is that there is simply a limit to how good a rating system can be.  (But we are not currently at that limit.)

btilly wrote:

But we tend to have more global games, and I don't think it worthwhile to have per board rating systems. 

What does this mean?

Yertle wrote:

btilly wrote:

But we tend to have more global games, and I don't think it worthwhile to have per board rating systems. 

What does this mean?

It means that it does not make sense to have a different rating system for each board, with rules custom-tailored to that board.

The alternative is to apply a generic rating system to all boards, and accept that it will work better for some than others.

btilly, I was also confused by your meaning. You mean, Glicko, Trueskill, or whatever, with a fixed set of parameters is applied to all boards. There is a global calculation.

My meaning of global is in the sense of the data. A game result on a Hex board only applies to the Hex board rating. A result on Wargear Warfare only to that rating, etc. The diversity of boards is great enough to have a need to separate these. (Admittedly, it would be nice to have categorical ratings, but that is complicated in its own way.)

smoke wrote:

SquintGnome wrote:

itsnotatumor wrote:

As it stands I see it as a disincentive to play certain maps or players. Every time I get in the upper teens/low 20's in the rankings I can get point swings of 300+ a day.  I need to win 2-4 games just to make up for 1 loss. I noticed that I started looking carefully at who's in a game and skip it if there's too many low ranked players.  It also seems that high ranked players tend to play less games. Is this the reason, or do people just cut back overtime?

 
+1

-1

Why -1?   Please explain. You have only 6 Rankable games running, most of which seem to have  experienced players. That seem pretty selective...

itsnotatumor wrote:

smoke wrote:

SquintGnome wrote:

itsnotatumor wrote:

As it stands I see it as a disincentive to play certain maps or players. Every time I get in the upper teens/low 20's in the rankings I can get point swings of 300+ a day.  I need to win 2-4 games just to make up for 1 loss. I noticed that I started looking carefully at who's in a game and skip it if there's too many low ranked players.  It also seems that high ranked players tend to play less games. Is this the reason, or do people just cut back overtime?

 
+1

-1

Why -1?   Please explain. You have only 6 Rankable games running, most of which seem to have  experienced players. That seem pretty selective...

I am selective, but not on the basis of the number of low ranking players. More low rankings just means greater chance of winning, with commensurately greater risk of high point loss, of course, should I lose. But I rarely play maps with fewer than 5 players--a frequent dueler, like SquintGnome, faces a much smaller reward to offset the risk of losing to a low ranker. So I can probably afford 2 bad losses (to a low rank) for every win to stay even. 

To your other point, I don't know if other high ranking players cut back, but I don't. I think Cona Chris and Mad Bomber play a lot of games. I just happen to be playing fewer right now, and more team or tournie games. If I play a lot more than 15-20 ranked games (typical for me) , like you do, I'd just play worse and lose more.

/End thread hijack.

I was thinking it made sense to calculate your the points you can gain from someone once they are eliminated, or calculate the number of points you have lost when eliminated.  That way if a game or some games go on for 6 months or more, and you've been eliminated already, you have a truer reflection of your score. 

It's unlikely but say you were in 100 games, all ongoing, all of them you've been eliminated, you have both a currently inflated rating that is both harder to push up, and will also fall further when those games end than it would have.

I don't really see a downside to this tweak.  But I don't feel passionately about it one way or the other, just was fueling some discussion that appears to have turned into a discussion about different ranking systems (which is fine).

Edit - on quick reflection, I guess this still allows one to drag out your wins until eliminated in other games anyways and may provide no benefit at all except meaning people only then drag out wins until they are eliminated in another game as opposed to dragging it out until the other game is over.  I don't drag my wins out anyways because when I see a turn I'm compulsive about taking it (unless it's team simulgear or a dueling simulgear map)  I just like to forget about a game once I'm out of it, even if it drags on for a long time :)  

smoke and itsnotatumor hijacked our hijack. smoke makes the good point that having lower rated opponents is usually the _better_ situation for someone who is good that cares about a rating, especially at the level of individual boards!

I can't answer for other top rankers, but I just plain slowed down. There are one or two boards where I get vain and try to hold the top spot through the obvious tactic of not playing until someone gets close. But overall, I don't find ratings worth getting in the way of a fun game. Can't have fun without playing.

I slowed down a little bit because I have a wedding in a couple of weeks (my own) and there's no point in starting a bunch of games.  I do get vain in a couple of boards in wanting the top rank, but I just generally start games without caring who my opponents are.  I'd rather have a good game than just a win.

Hugh wrote:

smoke and itsnotatumor hijacked our hijack. smoke makes the good point that having lower rated opponents is usually the _better_ situation for someone who is good that cares about a rating, especially at the level of individual boards!

I can't answer for other top rankers, but I just plain slowed down. There are one or two boards where I get vain and try to hold the top spot through the obvious tactic of not playing until someone gets close. But overall, I don't find ratings worth getting in the way of a fun game. Can't have fun without playing.

I've slowed down a lot, too...been a bit burned out lately.

I'll admit I will avoid some two-player games if I think there is a good player whose ranking hasn't really caught up with their true skill level.

Kind of related - having some sort of filtering for players could help with some of these issues.  I'd love to have some way of limiting boards I am learning to other newbs, and conversely filtering newbs from games where I am an expert.

The best example of where this would be helpful is a game like 'hex'.  Basically if I want to try a public game of this and I start a game up, it is likely to be joined by someone like Hugh or some other top player.  I am almost 100% certain to lose this game, as well as the next game, and the game after that, and probably every game ever.   So now, even though I kind of like Hex, i'll probably never play a public game on it.  

I too suck terribly at hex. I'll play ya anytime you feel like winning! 

Ozyman wrote:

Kind of related - having some sort of filtering for players could help with some of these issues.  I'd love to have some way of limiting boards I am learning to other newbs, and conversely filtering newbs from games where I am an expert.

It was only a matter of time before we resurrected the ratings filter option idea!

For what it's worth, when I know someone would rather not play me, I don't join their games unless a top spot is at stake. The ideal for those kinds of games is to get evenly matched players. There is probably a way to make this happen artificially. You can also play unranked, but it probably feels like poker with no money at stake.

Since we resurrected the 'filter option' idea, I will resuscitate my contribution - a filter that allows people to join your game based on rating, for example +/- 200. 

Or percentage +/- 25% (or 50%).  Or even based upon # of games played on the board, if someone has a philosophical objection to a score based limiter.  I don't think it would be as good, but it would be better than nothing.  You wouldn't need to change the max points lost/won.