I think I'v finally understood it at last. Yesterday I beat nordic's team in ladder and i lost 6 points, i think i know why now.

Okay, so lets have 3 teams. Pulsey's, Nordic's and Falker's.

A while ago...
Nordic's beats Falker.
Pulsey's beats Falker.

Ratings
Pulsey's 2066
Falker's 1800+
Nordic's 1600+

Yesterday...
Pulsey's beats Nordic. This gives a slight decrease to Nordic's rataing, which also gives a decrease to Falker's rating, since Nordic beat Falker earlier. This makes Pulsey's win over Falker's less valuable now. The win over a lowly rated team like Nordic's is not enough to cover the decrease in value of my win against a high rated team like Falker's. So YES!, beating a team can cause your rating to decrease.

No games happened during that period apart from this one. So i'm pretty certain there werent outside factors.
Did you guys know this already?

I've seen my 1v1 ladder rating go up when a game from a >1500 ranked player expired, so I suspected it, but I couldn't prove it because ranks from other players I had faced were changing too.

Hauptmann, I think Pulsey's point was that beating somebody can decrease your rating. It seems counterintuitive that winning a game can actually negatively affect your rating. You're just saying that someone dropping off the ladder can increase your rating, which seems logical.

We've been over this Pulsey, it's much more complicated than that. A beats B B goes lower, A goes higher, but because B is lower, so is C D E F G because they beat B too, A beat C D E F G so now A loses more points than they gained. It's not because A beat B that A lost points, it's because A's victory over B shifted the ladder. Take into account how many games B has played vs how many games A has played, and you'll see that B has a bigger pull on the ladder than A does, so therefore any changes to B will change more on the ladder than changes to A.

Slammy, the logic of my statement was;

defeating a player with low rank -> decreases rank, as opposed to increasing

game you won expires -> rank goes up, as opposed to down

but Lolowut makes a point, shifting values of other players due to victories/defeats. That doesn't explain how a victory results in point loss when one defeat shouldn't have a significant enough impact for a decrease of the winner's rating, if A -> B -> C than the impact would reverse from C -> B -> A causing an infinite influx of ratings that wouldn't be able to stabilize, theoretically speaking. I'm no mathematician, I use Riya for such things.

Edited 3/8/2014 06:18:07

Falker and Longhouse have 37 total games valid in the ladder right now, Pulsey and Fridge only have 14, so it's pretty obvious to tell that there's a bigger impact on the ladder when Falker/Longhouse's ratings change as opposed to Pulsey/Fridge's. Just because you don't understand mathematics well enough to figure out why something is happening doesn't mean it's broken. Go through the warlight wiki, the ELO system is thoroughly explained and you can find the formulas. If you don't understand how defeating a team that has more than double your games can have a bigger negative impact on the ladder than positive, at least keep that to yourself.

Lolowut - why all the hate? Has Pulsey stated the obvious? Yes. However - has Fizzer told it's not happening? i might be wrong but i think it happened a couple of times. Of course beating a team doesn't directly make you lose points, but indirectly it does.

On the other hand this only tends to happen when people's games are low so I guess it's an ineffective way for the ladder to handle people who try to run the ladder quickly.

Personally I'd prefer a soft limit on rating based on the number of games played but that's just my opinion.

yes, we knew it already. anyone who read the rules of the ladder properly knew it already.

Yes, we covered it.

http://warlight.net/Forum/Thread?ThreadID=20712

No, propagating wins and losses through the rest of the field is not the cause.

The loss of points is due to the fact that 2 things happen at the same time, 1)win % goes up, and 2)average opponent strength goes down, and in ELO formula(*) sometimes loss due to 2) overcomes gain due to 1). The formula is flawed.

* ELO rating = average opp rating - log(1/pwin-1) * 400

Although it's technically correct I'd not think of it that way. Since if you by winning over someone loses rating you'll lose rating simply by playing against them. Thus the points are gone(not there to being lost) the second you're paired against them. However should you lose against them you'd lose points, while winning you stay even.

Now if the opposite happened, going up in rating bc of a loss that would be something.
However it's worth mentioning that is excatly what can happen with the famous chess-ratings, you can go down in rating even though you win, and go up if you lose, but that's only if you've few games played.

Hennns: you do go up when getting a loss aswell. Just lose to very higher ranking players and you'll see it happening.