Guild Wars 2

I know this has been discussed and explained, but sorry, when you see something like this (attached), it makes zero sense.

How in the Mists can I have almost the same number of losses as my friend here, with more than 50% more wins, and yet have LOWER rating?

This, more than anything, demoralizes me from playing. I don’t mind being in perma silver 2, but this… it just feels insulting.

alccode.1297:

I know this has been discussed and explained, but sorry, when you see something like this (attached), it makes zero sense.

How in the Mists can I have almost the same number of losses as my friend here, with more than 50% more wins, and yet have LOWER rating?

This, more than anything, demoralizes me from playing. I don’t mind being in perma silver 2, but this… it just feels insulting.

Maybe he has faced harder (more higher rated opponents) than you have thus gained more rating from it. That’s what I can draw from your picture. I might be in the wrong though lol

Rating is derived from who you win against, not your raw win/loss. Winning against higher rating opponents will get you more rating than winning against lower rated opponents.

It’s actually possible to advance your rating with a negative win/loss, but that is uncommon.

he plays w/ or against better players and you dont. for example you can have a 40w-20L in bronze and a buddy of yours 40w-20L in plat.3

Thanks for the replies! I understand, but it seems a bit circular let alone flimsy as a basis of a combat rating system. If I play against bronze and someone plays against plat, how do we know the plat is any good? His rating? But ahhh, his rating is also dependent on who he plays against, not his W/L ratio. So it’s basically whoever happens to get in plat, stays in plat, and vice versa.

Also, it still doesn’t make sense. If both of us are in silver 2, we should on average end up facing the same skilled opponents. BUT if over 100 games I have a much larger win % than my friend, whose win % is quite poor, how is it still possible to have a similar rating? Certainly they should be more divergent at that point.

And finally, if combat rating is a measure of your effectiveness or skill as a player, it should be based on how well you do objectively speaking, not relative to anyone else. Why isn’t it based on K/D ratio for instance?

alccode.1297:

Why isn’t it based on K/D ratio for instance?

Not everyone focuses in killing

Sivannie.2670:

Not everyone focuses in killing

Yeah you’re right, it’s not even about killing. It’s more about strategy. Heh, sorry, that was a silly suggestion on my part.

In any case, my second last point stands, how it is illogical for two people who face on average similarly skilled opponents – by virtue of both being in the same division – to still have the same rating even though one has a much higher win %.

One other thing that can make huge differences in rating is if you win/lose while in decay state. just throwing this out here, but this too could have affected your rating if you know you had decay at some point and lost.

alccode.1297:

Thanks for the replies! I understand, but it seems a bit circular let alone flimsy as a basis of a combat rating system. If I play against bronze and someone plays against plat, how do we know the plat is any good? His rating? But ahhh, his rating is also dependent on who he plays against, not his W/L ratio. So it’s basically whoever happens to get in plat, stays in plat, and vice versa.

Also, it still doesn’t make sense. If both of us are in silver 2, we should on average end up facing the same skilled opponents. BUT if over 100 games I have a much larger win % than my friend, whose win % is quite poor, how is it still possible to have a similar rating? Certainly they should be more divergent at that point.

And finally, if combat rating is a measure of your effectiveness or skill as a player, it should be based on how well you do objectively speaking, not relative to anyone else. Why isn’t it based on K/D ratio for instance?

just simply NO, just because you or your buddy started in plat or w/e division if you dont belong there or not at par w/ player skills you will experience massive lose streak and drop you to the division you belong. the W/L ratio of that will be very negative like 40% 20W-40L

Everyone here forgets placement games, which are the root of all problems.

What can happen, as an extreme example, is that you had 1-9 in your placement and your friend had 9-1. You started at 1100 and your friend at 1400. Across the season, you both have the same amount of loses ( 50) and you have approx 30 more wins than him. If you assume one gets 10 points per win, then it makes the 300 points difference and you both end up having the same amount of points.

I’m taking an extreme case as an example, but that’s what happens generally

Abazigal.3679:

What can happen, as an extreme example, is that you had 1-9 in your placement and your friend had 9-1. You started at 1100 and your friend at 1400. Across the season, you both have the same amount of loses ( 50) and you have approx 30 more wins than him. If you assume one gets 10 points per win, then it makes the 300 points difference and you both end up having the same amount of points.

That’s a good explanation. If this is true, then the system is truly terrible, in that the first 10 placement games have more of an impact on rating than the next 90 games. It’s incredibly disproportionate.

There should be an “oops” factor in the rating adjustment if, after the 10 placement games, the system notices that you are performing very well for your placed division. And not this +10, +11, +12 increase in points only. In other words, there should be a “placement routine” always active, always ready to jump in and make larger adjustments as required.

Either that, or the placement shouldn’t affect your starting division that much at all, if it’s so much based on RNG (in terms of who you’re matched up during placement). Let people start out in a similar mid-range tier, and then they can float up or down as appropriate.

In any case, the system as it stands is massively broken – either in implementation or in DESIGN - to allow such discrepancies as I showed in OP.

It’s true that these systems are not designed to have these placement games. Real rating systems start everyone at exactly the same rating and they play, with the same volatility calculation.

However, all of these systems are also not designed to find matches base on “average team rating” which is just trash for actually knowing what’s going on.

Example
Team A: 2000 + 4 × 1000, = 6000 rating
Team B: 5 × 1200 = 6000 rating
It’s very likely that the rating 2000 player can’t carry vs all 5 people alone. And all of the 1200 rating players should easily kill all the rank 1000 players.

The game will compare his rating 2000 to their avg rating 1200 and say “why did you lose bro?”, boom -30 rating.

and if the 2000 wins that terrible match
g’job bro, here is +4 points for carrying that team like Atlas

phantom.1675:

It’s true that these systems are not designed to have these placement games. Real rating systems start everyone at exactly the same rating and they play, with the same volatility calculation.

You do realize that the “placement” games are just UI and have no bearing on the algorithm itself.

“placement” games only exist so that players aren’t freaking out at the high rating changes that occur when a player is at extreme volatility. (all players start at high volatility)

phantom.1675:

It’s true that these systems are not designed to have these placement games. Real rating systems start everyone at exactly the same rating and they play, with the same volatility calculation.

However, all of these systems are also not designed to find matches base on “average team rating” which is just trash for actually knowing what’s going on.

Example
Team A: 2000 + 4 × 1000, = 6000 rating
Team B: 5 × 1200 = 6000 rating
It’s very likely that the rating 2000 player can’t carry vs all 5 people alone. And all of the 1200 rating players should easily kill all the rank 1000 players.

The game will compare his rating 2000 to their avg rating 1200 and say “why did you lose bro?”, boom -30 rating.

terrible example:

mm tries to get players of the same interval a such extreme match only will ocur in a very low ppulated our and if the 2000 player is duoing with one of the 1000, and the “punishment” in this case is so low for forcing the system to put them in “easy” matches
and losing
2000+1000 / 2 =1500
the system will try to find players arround 1500 for the duo and for the other team

“I”?

“My friend”?

Hahahaha… Since when was ranked ratings about individuals?

Last I checked it was the RNG of ending up in a decent team (or the RNG of having a kittenty enemy team, depending on how you want to look at it).

Crinn.7864:

phantom.1675:

It’s true that these systems are not designed to have these placement games. Real rating systems start everyone at exactly the same rating and they play, with the same volatility calculation.

You do realize that the “placement” games are just UI and have no bearing on the algorithm itself.

“placement” games only exist so that players aren’t freaking out at the high rating changes that occur when a player is at extreme volatility. (all players start at high volatility)

So I guess my point is if you reset volatility at the start of each season (which does happen), they should start everyone at the same MMR. If they want to carry MMR from season to season, volatility should never reset. In addition, they need a better way of attributing what the expected outcome is as only 1/5 players in a team.

Take an extreme example like Sindrenerr/Misha who is winning at like 90% rate. If you are playing against those two and dont have a duo queue of pro league players, you are going to lose. In that situation you shouldnt be losing 10-20 rating, you should lose something like 0-2 and that just does not happen.

And yes I agree placement is just UI. But it’s UI that masks the fact that everyone’s starting MMR is not the same.

megilandil.7506:

phantom.1675:

It’s true that these systems are not designed to have these placement games. Real rating systems start everyone at exactly the same rating and they play, with the same volatility calculation.

However, all of these systems are also not designed to find matches base on “average team rating” which is just trash for actually knowing what’s going on.

Example
Team A: 2000 + 4 × 1000, = 6000 rating
Team B: 5 × 1200 = 6000 rating
It’s very likely that the rating 2000 player can’t carry vs all 5 people alone. And all of the 1200 rating players should easily kill all the rank 1000 players.

The game will compare his rating 2000 to their avg rating 1200 and say “why did you lose bro?”, boom -30 rating.

terrible example:

mm tries to get players of the same interval a such extreme match only will ocur in a very low ppulated our and if the 2000 player is duoing with one of the 1000, and the “punishment” in this case is so low for forcing the system to put them in “easy” matches
and losing
2000+1000 / 2 =1500
the system will try to find players arround 1500 for the duo and for the other team

I am aware they aren’t designed to match players with that different of a rating. However, I have been matched with ppl at both +400 and -400. So it happens. (These are in all solo queue games as well so duo queue does not matter). My example is just for ease of numbers and I realize that large of an extreme does not happen; point is the system does not prevent it from happening.