Jump to content
  • Sign Up

4 Top Scores Shouldn't Lose Rank


shion.2084

Recommended Posts

I know the way the current rating system works is that the amount of points you gain/lose are relative to the average rating of the game.So lets say my rating is 1500 and my friends rating is 1400. If we are duo queued we should expected the average rating of the game to be 1450 and if we win:

  • I will get "x" amount of points
  • He will get more than "x" amount of pointsThis is because he is below the average rating of the game and I am above the average rating of the game hence if we win, he will get more points and I will get less.

I think this system is currently fine where it stands.Almost every other ranked competitive game gives you a strict +/- points depending on whether you win or lose.

To put things into perspective: lets look at two games:

  1. Dota 2
  2. Overwatch

Dota 2 generally gives you +/- 25 points for every game whether you win or lose. There are no exceptions to this rule and this causes people to do everything they can to win. Supports will sacrifice themselves to ensure that their carries don't die and as a result this promotes a promising team dynamic. If supports were to lose more points if they died more... this would just break the entire matchmaking system and way the game functions.

Overwatch, back in the earlier seasons when I played it, gave you +/- points based on your performance in the game. While this sounds good in theory, supports were ALWAYS lower rating than their dps counterparts. A support will be focused on healing and hence have piss poor damage compared to their dps counterparts. This caused dps players to gain about 25 points per win, whereas support players would only gain about 15 points per win. This system made absolutely no sense and would decentivize support players to play support. This was one out of many reasons why you would generally have 6 dps players on your team rather than the optimal 2 tank, 2 dps, and 2 support setup.

Link to comment
Share on other sites

  • Replies 59
  • Created
  • Last Reply

Top Posters In This Topic

@Vagrant.7206 said:

@Crab Fear.1624 said:They should consider closer scores and lowering the rating loss a little.

I mean does a game that goes 500-496 deserve the same loss in rating as a loss of 500-15?

I definitely agree with this. I think close matches should reduce rating loss, precisely because it shows that both teams played well, one just happened to play
slightly
better. Or got lucky.

500-499 should have minimal rating loss for the losers, while 500-0 should have a MASSIVE rating loss for the losers. This incentivizes working with your team and playing until the end. Win gains shouldn't change though IMO, because a win is a win.

Also, it prevents people for afking when losing 100-300 for example ; Even if match seems lost, everyone will do its best to lose a minimal amount of points.

Link to comment
Share on other sites

@Ben Phongluangtham.1065 said:We're not likely to every alter the skill rating change formulas based on top stats. As others have noted, it would motivate bad behavior. People would chase top stats at all costs, even sacrificing the win to do so. Top stats are not always indicative on how much you helped your team and many important strategies may not net you any top stats at all.

We do reward top stats by giving an extra pip.

Why not remove top stats then? They still promote that behavior to some extent, albeit less than if it affected skill rating.

Even removing top stats from being displayed at the end of the game would be a beneficial since it contributes to toxicity.

Link to comment
Share on other sites

@Crab Fear.1624 said:They should consider closer scores and lowering the rating loss a little.

I mean does a game that goes 500-496 deserve the same loss in rating as a loss of 500-15?

If you do this, you have to reduce the rating gain for the winners by the same amount, otherwise you will have run-away rating inflation (not to mention opportunities for match manipulation).

I'm not sure people would be entirely happy with playing a long, hard-fought game that goes right down to the wire, only to gain +3 rating.

Link to comment
Share on other sites

The matchmaker is far from perfect but basing things on top stats..... I swear I am terrible at pvp yet I have top stats (damage/kills/offense) almost every single match I play (even with the community declared dead build called chronomancer). Keep top stats tho, I do chase them for daily purposes (only reason I am in pvp to start with). Oh and multiple ppl can have same top stats in end as well.

Link to comment
Share on other sites

@Ragnar.4257 said:

@Crab Fear.1624 said:They should consider closer scores and lowering the rating loss a little.

I mean does a game that goes 500-496 deserve the same loss in rating as a loss of 500-15?

If you do this, you have to reduce the rating gain for the winners by the same amount, otherwise you will have run-away rating inflation (not to mention opportunities for match manipulation).

If you can, explain where the inflation would happen.

I'm not sure I understand what you are saying, and I don't want to make any assumptions before I debate the point.

Link to comment
Share on other sites

@Crab Fear.1624 said:

@Crab Fear.1624 said:They should consider closer scores and lowering the rating loss a little.

I mean does a game that goes 500-496 deserve the same loss in rating as a loss of 500-15?

If you do this, you have to reduce the rating gain for the winners by the same amount, otherwise you will have run-away rating inflation (not to mention opportunities for match manipulation).

If you can, explain where the inflation would happen.

I'm not sure I understand what you are saying, and I don't want to make any assumptions before I debate the point.

If you gain +15 for a win, and only lose -5 for a loss, then everyone's rating will keep on increasing and increasing.

Assuming you have reached your "correct" rating, your rating should be roughly static, not forever increasing.

Not to mention, players will very quickly learn that they can cheat the system by agreeing with the enemy team to let them catch-up to 450-500, or whatever the threshold is. So that everyone only gains rating, and never loses it. It's a win-trader's wet-dream.

If you do this, then rating becomes meaningless, and might as well be replaced with a "number of games played" counter.

The rating system has to be very finely tuned to create "equilibrium". If gains exceed losses, then you end up with everyone in Legendary-3. If losses exceed gains, then you end up with everyone in Bronze-1. As much as people complain about the current system, it is actually balanced correctly, the evidence for this is the distribution of ratings. If it were imbalanced, then you would see 1 of the 2 latter scenarios.

It is very important to think about all the implications of a suggestion. The good and the bad. Because players will abuse it for the bad.

Link to comment
Share on other sites

@Ragnar.4257 said:

@Crab Fear.1624 said:They should consider closer scores and lowering the rating loss a little.

I mean does a game that goes 500-496 deserve the same loss in rating as a loss of 500-15?

If you do this, you have to reduce the rating gain for the winners by the same amount, otherwise you will have run-away rating inflation (not to mention opportunities for match manipulation).

If you can, explain where the inflation would happen.

I'm not sure I understand what you are saying, and I don't want to make any assumptions before I debate the point.

If you gain +15 for a win, and only lose -5 for a loss, then everyone's rating will keep on increasing and increasing.

You would actually need to be closer in score, 496-500 is a match that should not recieve the same rewards as a match of 500-100. The closer players get to matchmaker's estimation should be rewarded more accuratley and punished less.

If the match maker assumes that either team in the match could win, and it is very close, then the punishment should be less for the losing team.

If the team that should have won, or at least put up the good fight actually loses drastically, there would be a reason to take away more points.

Real sports teams have a similar system, and that is why a team can beat another team and yet still see each other again in the championship.

You don't tumble down the leaderboard just because someone really really good barely beat you.

That doesn't make any sense.

Assuming you have reached your "correct" rating, your rating should be roughly static, not forever increasing.

Not to mention, players will very quickly learn that they can cheat the system by agreeing with the enemy team to let them catch-up to 450-500, or whatever the threshold is. So that everyone only gains rating, and never loses it. It's a win-trader's wet-dream.

We only have duos. This would be EXTREMELY unlikely to happen. To get a whole team to agree to let them catch up. Also matches have a time limit, how long would this catch up take?

This is also assuming that everyone is out to cheat.

If you do this, then rating becomes meaningless, and might as well be replaced with a "number of games played" counter.

The rating system has to be very finely tuned to create "equilibrium". If gains exceed losses, then you end up with everyone in Legendary-3. If losses exceed gains, then you end up with everyone in Bronze-1. As much as people complain about the current system, it is actually balanced correctly, the evidence for this is the distribution of ratings. If it were imbalanced, then you would see 1 of the 2 latter scenarios.

You actually have to win or "perform well". Top stats are not an implication of performance in many cases when it comes to "team play ability".

Closer scores are.

It is very important to think about all the implications of a suggestion. The good and the bad. Because players will abuse it for the bad.

Since you bring up this point. Is duo q still healthy for the game?

Link to comment
Share on other sites

@Ben Phongluangtham.1065 said:We're not likely to every alter the skill rating change formulas based on top stats. As others have noted, it would motivate bad behavior. People would chase top stats at all costs, even sacrificing the win to do so. Top stats are not always indicative on how much you helped your team and many important strategies may not net you any top stats at all.

We do reward top stats by giving an extra pip.

What about removing it then? it doesn't really give anything positive ... if you still want to see you can check stats either way ! :)

plus daily top stats is pretty dum dum in this direction too. More dailies would be highly appreciated :D

thanks for your participation to this thread !

Link to comment
Share on other sites

@Crab Fear.1624 said:You would actually need to be closer in score, 496-500 is a match that should not recieve the same rewards as a match of 500-100. The closer players get to matchmaker's estimation should be rewarded more accuratley and punished less.

If the match maker assumes that either team in the match could win, and it is very close, then the punishment should be less for the losing team.

By the same logic, the gain should also be less for the winning team.

@Crab Fear.1624 said:If the team that should have won, or at least put up the good fight actually loses drastically, there would be a reason to take away more points.

Sure, and if you do that, then the winners should gain correspondingly more points.

I don't think you're understanding the concept. If you increase/decrease the losses for losing, then you must increase/decrease the gains for winning by the corresponding amount, otherwise you create imbalance.

@Crab Fear.1624 said:Real sports teams have a similar system, and that is why a team can beat another team and yet still see each other again in the championship.

You don't tumble down the leaderboard just because someone really really good barely beat you.

Except, real sports teams play a fixed number of games. You don't have 1 team who only plays 3 games a season, and another team who plays 300. Otherwise, the team that plays 300 games would obviously have much more points than the team that plays 3.

Rating is not the same thing as points in a league-table.

@Crab Fear.1624 said:That doesn't make any sense.

No, it doesn't, because you're not comparing like-for-like. If you insist on comparing cars to toasters, then you'll end up with nonsense conclusions.

@Crab Fear.1624 said:We only have duos. This would be EXTREMELY unlikely to happen. To get a whole team to agree to let them catch up. Also matches have a time limit, how long would this catch up take?

It only takes 1 or 2 players cheating to completely throw off the balance of a match.

@Crab Fear.1624 said:This is also assuming that everyone is out to cheat.

If you make it so easy to do so, they definitely will.

@Crab Fear.1624 said:You actually have to win or "perform well". Top stats are not an implication of performance in many cases when it comes to "team play ability".

?? Did I say anything about Top Stats?

@Crab Fear.1624 said:Since you bring up this point. Is duo q still healthy for the game?

The developers have to make a compromise. On the one-hand, there are players who want a pure solo ladder. On the other-hand, there are players who want to play with their friends, and who don't want to have to deal with random stranger's BS. Duo-Q is a reasonable compromise between these 2 positions, and given that there is not enough population to support separate Solo and Team queues, its probably the best we can have.

Link to comment
Share on other sites

@"Ragnar.4257" said:

By the same logic, the gain should also be less for the winning team.

Sure, and if you do that, then the winners should gain correspondingly more points.

I don't think you're understanding the concept. If you increase/decrease the losses for losing, then you must increase/decrease the gains for winning by the corresponding amount, otherwise you create imbalance.

??? So gain more or not gain more? Performance and expectations based ratings are something we obviously both understand here.

Rating is not the same thing as points in a league-table.

No, it doesn't, because you're not comparing like-for-like. If you insist on comparing cars to toasters, then you'll end up with nonsense conclusions.

These things are actually very similar.

https://en.wikipedia.org/wiki/Sports_rating_system#Rating_methods

You will see TrueSkill and Glicko in the same group

Doesn't matter how many games are played.

The system is looking at the ratings of the two team with rewards/punishments for outcomes.

It only takes 1 or 2 players cheating to completely throw off the balance of a match.

@Crab Fear.1624 said:You actually have to win or "perform well". Top stats are not an implication of performance in many cases when it comes to "team play ability".

?? Did I say anything about Top Stats?

I didn't say you did, I was building on the team play point, which you left out.

@Crab Fear.1624 said:Since you bring up this point. Is duo q still healthy for the game?

The developers have to make a compromise. On the one-hand, there are players who want a pure solo ladder. On the other-hand, there are players who want to play with their friends, and who don't want to have to deal with random stranger's BS. Duo-Q is a reasonable compromise between these 2 positions, and given that there is not enough population to support separate Solo and Team queues, its probably the best we can have.

@Crab Fear.1624 said:This is also assuming that everyone is out to cheat.

If you make it so easy to do so, they definitely will.It only takes 1 or 2 players cheating to completely throw off the balance of a match.

Link to comment
Share on other sites

@Crab Fear.1624 said:

By the same logic, the gain should also be less for the winning team.

Sure, and if you do that, then the winners should gain correspondingly more points.

I don't think you're understanding the concept. If you increase/decrease the losses for losing, then you must increase/decrease the gains for winning by the corresponding amount, otherwise you create imbalance.

??? So gain more or not gain more? Performance and expectations based ratings are something we obviously both understand here.

sigh

I didn't say you can't reduce the amount you lose for a close loss.

What I said was that if you do that, then you must also make the corresponding reduction to the gains for the winners.

Likewise, if you increase the amount you lose for a major loss, then you must also make the corresponding increase to the gains for the winners.

-5 for losers, +5 for winners, in a close game.Or-30 for losers, +30 for winners, in a steam-roll.

NOT-5 for losers, +15 for winners, in a close game

I don't know how I can explain this any more clearly. Its a very basic concept. You must scale both the gains and the losses. Not just one or the other.

(All of this assuming evenly rated teams with a 50/50 expected outcome. The subject of differences in rating / expected results is entirely separate and should not be conflated).

Link to comment
Share on other sites

@Ragnar.4257 said:

sigh

I didn't say you can't reduce the amount you lose for a close loss.

What I said was that if you do that, then you must also make the corresponding reduction to the gains for the winners.

Likewise, if you increase the amount you lose for a major loss, then you must also make the corresponding increase to the gains for the winners.

-5 for losers, +5 for winners, in a close game.Or-30 for losers, +30 for winners, in a steam-roll.

NOT-5 for losers, +15 for winners, in a close game

I don't know how I can explain this any more clearly. Its a very basic concept. You must scale both the gains and the losses. Not just one or the other.

(All of this assuming evenly rated teams with a 50/50 expected outcome. The subject of differences in rating / expected results is entirely separate and should not be conflated).

It is good you edited your post to clarify you meant 50/50 chances.

You say scale gains and losses (both). Ok, I understand that. But it must be done accordingly against performance expectations.

As it stands when the match starts and finishes, you get the same amount of points based on outcome and not performance.

We can not just consider the 50/50 matches when making these conclusions.

Sometimes the team balance is 20/80.

But, all of this is pointless if we don't address the original concern I had with your argument.

We can come back to this tangent.

First, I wanted to know where the inflation came from.

You responded. I saw it.

But, you imagined only one scenario. -5, and +15.

Performance in games and difference in team ratings is really going to prevent this from being the case every time.

If you consistently beat better teams, your overall rating will go up until it reaches your true rating.

At this point, you should regularly face more challenging teams and players.

If the matchmaker can make closer matches with appropriate win and loss awards you can't go up forever.

I don't think 50/50 win loss rewards should be the same either. There must be some punishment minimum. (win loss rewards should not be equal)

I also feel the worse you do considering expectations, the more you should lose.

Why should the system reward giving up and punish trying hard?

Link to comment
Share on other sites

@Ragnar.4257 said:

@Crab Fear.1624 said:They should consider closer scores and lowering the rating loss a little.

I mean does a game that goes 500-496 deserve the same loss in rating as a loss of 500-15?

If you do this, you have to reduce the rating gain for the winners by the same amount, otherwise you will have run-away rating inflation (not to mention opportunities for match manipulation).

If you can, explain where the inflation would happen.

I'm not sure I understand what you are saying, and I don't want to make any assumptions before I debate the point.

The rating system has to be very finely tuned to create "equilibrium". If gains exceed losses, then you end up with everyone in Legendary-3. If losses exceed gains, then you end up with everyone in Bronze-1. As much as people complain about the current system, it is actually balanced correctly, the evidence for this is the distribution of ratings. If it were imbalanced, then you would see 1 of the 2 latter scenarios.

Actually, the global average is gradually decreasing as players leave the game permanently. The bell curve is shrinking from one side only.

Link to comment
Share on other sites

@"Crab Fear.1624"

You're conflating 2 entirely separate parameters.

Parameter A = scaling based on how close a match is.

Parameter B = scaling based on the expected outcome, based on the relative ratings of the 2 teams.

Total Rating Change = A x B

These are 2 separate parameters. One has no bearing on the other.

Parameter B is already in the game, it already exists, so I don't know why you're even bringing it up as a factor to consider. If you are expected to win a game, you gain less for a win and lose more for a loss. And vice versa.

You are talking about introducing a new Parameter A. This does not in any way impact Parameter B. We can hold Parameter B constant (for the sake of argument, at 50/50). It doesn't in any way affect how Parameter A should be treated. We could also hold Parameter B at 80/20 or 20/80. It still doesn't affect Parameter A.

For the purposes of discussing what affect the closeness of a match should have on rating change, we can completely ignore the subject of expected outcome. It. Is. A. Different. Parameter.

Please look up "Separation of Concerns", or basic algebra.

Your last few lines are just.... nonsensical. In what sense does the system reward giving up and punish try hard? And how can a system possibly be balanced if (all other factors aside) wins are not rewarded equally to losses? Again. Separation. Of. Concerns.


Acceleration = Force / Mass

I'm saying "If you increase Force, then you increase Acceleration", and you're going "but what if the Mass is different".......... it is not relevant to the point!

I'm saying "The sky is blue", and you're going "yeah, but I don't like ice cream"......... ?!? And this shows that the sky is not blue how?

Link to comment
Share on other sites

@Vagrant.7206 said:

@Crab Fear.1624 said:They should consider closer scores and lowering the rating loss a little.

I mean does a game that goes 500-496 deserve the same loss in rating as a loss of 500-15?

If you do this, you have to reduce the rating gain for the winners by the same amount, otherwise you will have run-away rating inflation (not to mention opportunities for match manipulation).

If you can, explain where the inflation would happen.

I'm not sure I understand what you are saying, and I don't want to make any assumptions before I debate the point.

The rating system has to be very finely tuned to create "equilibrium". If gains exceed losses, then you end up with everyone in Legendary-3. If losses exceed gains, then you end up with everyone in Bronze-1. As much as people complain about the current system, it is actually balanced correctly, the evidence for this is the distribution of ratings. If it were imbalanced, then you would see 1 of the 2 latter scenarios.

Actually, the global average is gradually decreasing as players leave the game permanently. The bell curve is shrinking from one side only.

Yes, but it is still maintaining a bell-curve, normalised I think on 1200 rating. How do you know it is shrinking from one side only?

Link to comment
Share on other sites

@Ragnar.4257 said:@"Crab Fear.1624"

You're conflating 2 entirely separate parameters.

Parameter A = scaling based on how close a match is.

Parameter B = scaling based on the expected outcome, based on the relative ratings of the 2 teams.

Total Rating Change = A x B

These are 2 separate parameters. One has no bearing on the other.

Parameter B is already in the game, it already exists, so I don't know why you're even bringing it up as a factor to consider. If you are expected to win a game, you gain less for a win and lose more for a loss. And vice versa.

You are talking about introducing a new Parameter A. This does not in any way impact Parameter B. We can hold Parameter B constant (for the sake of argument, at 50/50). It doesn't in any way affect how Parameter A should be treated. We could also hold Parameter B at 80/20 or 20/80. It still doesn't affect Parameter A.

For the purposes of discussing what affect the closeness of a match should have on rating change, we can completely ignore the subject of expected outcome. It. Is. A. Different. Parameter.

Please look up "Separation of Concerns", or basic algebra.

Your last few lines are just.... nonsensical. In what sense does the system reward giving up and punish try hard? And how can a system possibly be balanced if (all other factors aside) wins are not rewarded equally to losses? Again. Separation. Of. Concerns.


Acceleration = Force / Mass

I'm saying "If you increase Force, then you increase Acceleration", and you're going "but what if the Mass is different".......... it is not relevant to the point!

I'm saying "The sky is blue", and you're going "yeah, but I don't like ice cream"......... ?!? And this shows that the sky is not blue how?

Nope.

Here is my original idea:

They should consider closer scores and lowering the rating loss a little.

Here is your response:

If you do this, you have to reduce the rating gain for the winners by the same amount, otherwise you will have run-away rating inflation.

Here is me asking what you mean:

If you can, explain where the inflation would happen.

Here is your response:

If you gain +15 for a win, and only lose -5 for a loss, then everyone's rating will keep on increasing and increasing.

I answered back later:

But, you imagined only one scenario. -5, and +15.

Performance in games and difference in team ratings is really going to prevent this from being the case every time.

Here is your response:

You're conflating 2 entirely separate parameters.

Here is my response:

I am not mixing them, blending them, combining them, or confused.

I see two things: (A)Evenness of the match, (B)outcome of the match.

(A) affects the amount of points lost or gained (this will be set in stone once the match begins as the base outcome)

(B) determines if you gain or lose points

I must be asking for a ©closeness of final score

© modifies the points lost or gained (this swings the base outcome left or right)

Link to comment
Share on other sites

I’d like to see a ratings system that incentivizes everyone to play until the end of the game.

I’d like to see a rating system where, even if you lose, your team can celebrate not losing as badly as they might have.

I’d like to see a rating system with reverse elasticity. The more games you play the more elastic your rating is (but not as big as current start of season elasticity. At the start of the season you can gain a little rating or lose a little rating as people play lots of games. Then, as you approach the number of games needed to place on the leaderboards (for entire season) your elasticity increases to force top players to choose between staying on the leaderboard by continuing to play and winning or not playing and letting the lower ranked players increase their ranks and kick “top rated” players out of the leaderboard.

I’d like to see a leaderboard for the Professions based on players most played profession that season.

Most of all I’d like to see some monetization aimed at PvP players so Anet can justify spending money on developing these features and more. It’s easy to see why Anet makes the content it does, because people buy gems for it. If only pvp had something optional (non-competitive) to help Anet fund PvP development we might actually have more PvP game modes, maps and features.

Link to comment
Share on other sites

@"Crab Fear.1624" said:

Nope.

Here is my original idea:

They should consider closer scores and lowering the rating loss a little.

Here is your response:

If you do this, you have to reduce the rating gain for the winners by the same amount, otherwise you will have run-away rating inflation.

Here is me asking what you mean:

If you can, explain where the inflation would happen.

Here is your response:

If you gain +15 for a win, and only lose -5 for a loss, then everyone's rating will keep on increasing and increasing.

I answered back later:

But, you imagined only one scenario. -5, and +15.

It could be -10/+15, it could be -3/+10, it could be -10/+10, it could be -12/+8, it doesn't matter. The exact values are not relevant.

You're asking for lowering the rating loss.

You literally just said that. It's in the quote.

But you did not ask for a corresponding lowering of rating gain for the winners.

If we assume that, currently, the system is balanced, as in, on average across all games, the total gains and losses even out. Then currently, taken as a whole across all matches, losses and gains are equal.

By reducing the rating loss, on average, across all matches, gains will now exceed losses. So rating will inflate. Again, -5/+15, -10/+10, it really really doesn't matter. You are changing one side of the equation, you must therefore change the other side, or else create an imbalance.

Currently L = G

You are asking for L-1 = G

I am saying you must make it L-1 = G-1

How will you balance out that gains are now greater than losses? Well, you suggest it here:

Performance in games and difference in team ratings is really going to prevent this from being the case every time.

But this is already in the game. The system is already balanced around differences in team ratings. You are introducing a new parameter on one side. You must therefore also introduce it on the other side, or create an imbalance.

You CANNOT cite an already-accounted-for variable as a counter-balance to a new variable.

Here is your response:

You're conflating 2 entirely separate parameters.

Here is my response:

I am not mixing them, blending them, combining them, or confused.

I see two things: (A)Evenness of the match, (B)outcome of the match.

(A) affects the amount of points lost or gained (this will be set in stone once the match begins as the base outcome)

(B) determines if you gain or lose points

I must be asking for a ©closeness of final score

© modifies the points lost or gained (this swings the base outcome left or right)

This is exactly what I have been saying the whole time! In your example, "C" modifies the points lost or gained. Lost or gained. Lost or gained. Not just lost! That's all I've been trying to convey!

"© modifies the points lost or gained"

not

"© modifies the points lost"

Just those 2 words "or gained" on the end make all the difference.

If it is "© modifies the points lost or gained" then that is totally fine! That works!

But if it ONLY affects losses, and not gains, then it is imbalanced! I can't believe you need an entire thread of explanation to reach this point.

Link to comment
Share on other sites

@Ragnar.4257 said:

@"Crab Fear.1624" said:

Nope.

Here is my original idea:

They should consider closer scores and lowering the rating loss a little.

Here is your response:

If you do this, you have to reduce the rating gain for the winners by the same amount, otherwise you will have run-away rating inflation.

Here is me asking what you mean:

If you can, explain where the inflation would happen.

Here is your response:

If you gain +15 for a win, and only lose -5 for a loss, then everyone's rating will keep on increasing and increasing.

I answered back later:

But, you imagined only one scenario. -5, and +15.

It could be -10/+15, it could be -3/+10, it could be -10/+10, it could be -12/+8, it doesn't matter. The exact values are not relevant.

You're asking for lowering the rating loss.

You literally just said that. It's in the quote.

But you did not ask for a corresponding lowering of rating gain for the winners.

If we assume that, currently, the system is balanced, as in, on average across all games, the total gains and losses even out. Then currently, taken as a whole across all matches, losses and gains are equal.

By reducing the rating loss, on average, across all matches, gains will now exceed losses. So rating will inflate. Again, -5/+15, -10/+10, it really really doesn't matter. You are changing one side of the equation, you must therefore change the other side, or else create an imbalance.

Currently L = G

You are asking for L-1 = G

I am saying you must make it L-1 = G-1

How will you balance out that gains are now greater than losses? Well, you suggest it here:

Performance in games and difference in team ratings is really going to prevent this from being the case every time.

But this is already in the game. The system is
already
balanced around differences in team ratings. You are introducing a new parameter on one side. You must therefore also introduce it on the other side, or create an imbalance.

You CANNOT cite an already-accounted-for variable as a counter-balance to a new variable.

Here is your response:

You're conflating 2 entirely separate parameters.

Here is my response:

I am not mixing them, blending them, combining them, or confused.

I see two things: (A)Evenness of the match, (B)outcome of the match.

(A) affects the amount of points lost or gained (this will be set in stone once the match begins as the base outcome)

(B) determines if you gain or lose points

I must be asking for a ©closeness of final score

© modifies the points lost or gained (this swings the base outcome left or right)

This is exactly what I have been saying the whole time! In your example, "C" modifies the points lost or gained. Lost or gained. Lost or gained. Not just lost! That's all I've been trying to convey!

"© modifies the points lost or gained"

not

"© modifies the points lost"

Just those 2 words "or gained" on the end make all the difference.

But, I never said it was not the case, or they shouldn't do balance the reward.

Quite the opposite.

You are explain a point that was already acknowledged and understood many times over.

I asked about inflation.

¯_(ツ)_/¯

Link to comment
Share on other sites

@Ragnar.4257 said:

@"Crab Fear.1624" said:They should consider closer scores and lowering the rating loss a little.

I mean does a game that goes 500-496 deserve the same loss in rating as a loss of 500-15?

If
you do this, you have to reduce the rating gain for the winners by the same amount,
otherwise
you will have run-away rating inflation (not to mention opportunities for match manipulation).

I'm not sure people would be entirely happy with playing a long, hard-fought game that goes right down to the wire, only to gain +3 rating.

Here is my original post.

Read it again.

I said: "You have to reduce the rating gain for the winners by the same amount, otherwise you will have run-away rating inflation".

I specifically stated that the rating inflation would ONLY occur if your "©" only affected losses and not gains.

If losses and gains are both scaled equally, no problem. If ONLY losses are scaled, and not gains, THAT is the scenario where inflation occurs.

Jesus Christ.

Lets say that, based on expected outcome for a game, you would get -16 for a loss, and +10 for a win.

Now lets say you lose the game, but the score is 490-500, you are suggesting that you shouldn't lose -16, but instead only lose -8 (so, in this example C=0.5)

All I'm saying is that, by the same logic, if you win by 500-490, you should only gain +5 and not +10 (using the same factor C = 0.5).

The inflation would come in IF and ONLY IF, you reduced the -16 to -8, but left the +10 and did not reduce to +5.

Link to comment
Share on other sites

@Ragnar.4257 said:

@"Crab Fear.1624" said:They should consider closer scores and lowering the rating loss a little.

I mean does a game that goes 500-496 deserve the same loss in rating as a loss of 500-15?

If you do this, you have to reduce the rating gain for the winners by the same amount, otherwise you will have run-away rating inflation (not to mention opportunities for match manipulation).

I'm not sure people would be entirely happy with playing a long, hard-fought game that goes right down to the wire, only to gain +3 rating.

Here is my original post.

Read it again.

I said: "You have to reduce the rating gain for the winners by the same amount, otherwise you will have run-away rating inflation".

I specifically stated that the rating inflation would ONLY occur if your "©" only affected losses and not gains.

If losses and gains are both scaled equally, no problem. If ONLY losses are scaled, and not gains, THAT is the scenario where inflation occurs.

Jesus Christ.

For an even match

Currently you win +12 and you lose -16

if the score is closer (and you save 3 points)

You want it so be +9 and -13?

(that is a 4 unit gap regardless) edit: between the balance I mean, not the score

"I don't think 50/50 win loss rewards should be the same either. There must be some punishment minimum. (win loss rewards should not be equal)"

It can never be win +12 or lose -12, or win +11 lose -10.

Losses should never be (-)lower than the winning reward.

So, this scenario of yours doesn't exist.

Flying Spaghetti Monster.

Link to comment
Share on other sites

And now you've totally moved the goal-posts, on to an entirely separate point about whether wins and losses should be rewarded equally.

Which they should be IF everyone is getting 50/50 expected matchups and getting 50% win-rates.

Do you think your rating should increase even if you're winning 50% of 50/50 expected matchups? What?! That would mean that the more games you play, the higher your rating goes. That means that rating is simply equivalent to number of games played. Which is meaningless PvE achievement drivel.

Link to comment
Share on other sites

@"Crab Fear.1624" said:

"I don't think 50/50 win loss rewards should be the same either. There must be some punishment minimum. (win loss rewards should not be equal)"

It can never be win +12 or lose -12, or win +11 lose -10.

Losses should never be (-)lower than the winning reward.

So, this scenario of yours doesn't exist.

Flying Spaghetti Monster.

What does this even mean?

What scenario?

Losses should never be lower than the winning reward? What? Since when?

Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...