[PC] Kill Ratio to Replace Percent Won

[Metsfanmax]

I read that. I specifically addressed that with my last sentence: "someone who has lost only a few games will have a displayed win rate of 100% or greater even though their actual is below that." So you would have to turn it off well before 100% if you want someone who has lost any games to be guaranteed not to have a 100% win ratio. (Which should be a requirement of a sensible 'win ratio' metric.) And besides, the discontinuity there is just ugly.

You're going to have to do a lot more work before this makes sense. If I play 10 1v1 games and lose all of them, and then play 10 12 player games and win all of them, my win ratio according to the nominal formula would be 300%. Deciding to just call that 100% is not providing a subtle fix to an otherwise reasonable formula, it is exposing how inappropriate it is for measuring what we normally think of as win ratio.

This is one of the reasons I prefer kill ratio to this weighted win percentage. The kill ratio is always a sensible number between 0 and 100% -- you can't kill more opponents than all of them. In my example, the kill ratio would be 92% -- I played 120 opponents and defeated 110 of them. You can even call this win ratio if you want, it's not untrue and it's what you really want anyway.

[betiko]

I read that. I specifically addressed that with my last sentence: "someone who has lost only a few games will have a displayed win rate of 100% or greater even though their actual is below that." So you would have to turn it off well before 100% if you want someone who has lost any games to be guaranteed not to have a 100% win ratio. (Which should be a requirement of a sensible 'win ratio' metric.) And besides, the discontinuity there is just ugly.

You're going to have to do a lot more work before this makes sense.

the objective of this "expected win rate aggregate" is to give a clearer view on how well a person does compared to the game type he plays. If someone ends up with 100% win rate with my formula, well he is a complete noob farmer or he has played very few games.
If you think the actual win rate displayed gives a better idea of a player's skill well I would totally disagree. The win rates as they are right now are totally irrelevant, mostly with the bot games counted in them.
If you don't think it needs a fix and that I owe you pages of demonstrations and answer your typical "suggestions forum attitude", well, I'm not going to bother.

[Metsfanmax]

If you think the actual win rate displayed gives a better idea of a player's skill well I would totally disagree. The win rates as they are right now are totally irrelevant, mostly with the bot games counted in them.

I don't know why you think I prefer the status quo. I already said you should prefer kill ratio as being better than both the current thing and your proposed thing. Your solution is actually worse than the status quo, straight out, even though it is well intentioned.

If you don't think it needs a fix and that I owe you pages of demonstrations and answer your typical "suggestions forum attitude", well, I'm not going to bother.

See my above edit:

If I play 10 1v1 games and lose all of them, and then play 10 12 player games and win all of them, my win ratio according to the nominal formula would be 300%. Deciding to just call that 100% is not providing a subtle fix to an otherwise reasonable formula, it is exposing how inappropriate it is for measuring what we normally think of as win ratio.

This is one of the reasons I prefer kill ratio to this weighted win percentage. The kill ratio is always a sensible number between 0 and 100% -- you can't kill more opponents than all of them. In my example, the kill ratio would be 92% -- I played 120 opponents and defeated 110 of them. You can even call this win ratio if you want, it's not untrue and it's what you really want anyway.

Anyway no, I don't expect you owe me pages of demonstrations. I already showed why your approach is flawed. I expect you to simply accept that and come up with a better fix. You could accept mine, or do something else, just please abandon any formula that could result in something more than 100%.

[Metsfanmax]

The fundamental problem with the weighted win rate (as expressed by betiko) is that it is not properly normalized. You're multiplying a bunch of things in the numerator to give them additional weighting, but you're not dividing by a denominator that brings the total back under 100%.

We can rescue this by dividing everything by the proper denominator. The denominator could be equal to the weighted average of the multipliers. So in my example, we would take (0/20 * 1 + 10/20 * 6) == 3.0 and turn it into (0/20 * 1 + 10/20 * 6) / (10/20 * 1 + 10/20 * 6) == 0.857. This isn't a unique way to do it but is far more agreeable with our intuitions.

[betiko]

If you think the actual win rate displayed gives a better idea of a player's skill well I would totally disagree. The win rates as they are right now are totally irrelevant, mostly with the bot games counted in them.

I don't know why you think I prefer the status quo. I already said you should prefer kill ratio as being better than both the current thing and your proposed thing. Your solution is actually worse than the status quo, straight out, even though it is well intentioned.

If you don't think it needs a fix and that I owe you pages of demonstrations and answer your typical "suggestions forum attitude", well, I'm not going to bother.

See my above edit:

This is one of the reasons I prefer kill ratio to this weighted win percentage. The kill ratio is always a sensible number between 0 and 100% -- you can't kill more opponents than all of them. In my example, the kill ratio would be 92% -- I played 120 opponents and defeated 110 of them. You can even call this win ratio if you want, it's not untrue and it's what you really want anyway.

But no, I don't expect you owe me pages of demonstrations. I already showed why your approach is flawed. I am content with you simply accepting that and coming up with a better fix.

how is about kill ratio vs this? it's something completely different. both threads have been merged and have absolutely nothing to do with one another...
Risk is about being the last one/team standing. With your bullshit idea people would start going for suicide kills in lost games just to increase their "kill ratio" you are so found of. Your "suggestion" would very likely change people's behavior for the worse.
The only consequence on gameplay my suggestion could have is that a guy really good in let's say 4 player games only plays 4 player games and would have a killer ratio... but he would also have a killer score.
If that goes too close to 100% (and I highly doubt anyone would be above 80% on a full game history sample without being a farmer); well who cares. Everybody would be rated on that same index.

Go play candy crush metsfan, you have never been of any relevance to CC games, yet you always talk as the wisest and most listened player on the site.

[Metsfanmax]

how is about kill ratio vs this? it's something completely different.

They're not that different. Both are attempts to deal with the fact that winning a game with more players is more impressive than winning a game with fewer players, and thus attempts to replace the current profile win ratio metric. They differ only in the details of how they are calculated; as you say, good players will have a killer score on either index.

Risk is about being the last one/team standing. With your bullshit idea people would start going for suicide kills in lost games just to increase their "kill ratio" you are so found of. Your "suggestion" would very likely change people's behavior for the worse.

In my proposal, you don't generally 'kill' anyone if you eliminate them in a game. The winner of any non-terminator games gets the credit for all eliminations in that game. It is only in terminator that this would vary, to give credit for terminator kills (since eliminating people -- even at the risk of losing the game -- is explicitly the goal of Terminator). This should have been obvious from my numerical example in which I automatically gave full credit to the winner of the 12 player games, or could have been obvious if you had asked what I meant instead of going off on a rant.

(Although even your rant is dubious. Most people care about their score way more than their profile's win ratio and it is hard to imagine them changing their behavior much to sacrifice the former for the latter.)

The only consequence on gameplay my suggestion could have is that a guy really good in let's say 4 player games only plays 4 player games and would have a killer ratio... but he would also have a killer score.

If I play only 4 player games, then by your suggestion if I win half of them my 'aggregrate win ratio' is already 100%. If I win 60% of them, my 'win ratio' is 120%. This is true in general: as long as I do twice as good as the expected win rate, my 'aggregrate win ratio' will be 100%. Now I'll be the first to admit that doing twice as well as expectations is pretty damn hard, and an impressive achievement. But it's not so good that I can say I won 100% of the time.

Go play candy crush metsfan, you have never been of any relevance to CC games, yet you always talk as the wisest and most listened player on the site.

If only we lived in a world where people accepted their mistakes and fixed them instead of insulting the person who pointed them out.

[Metsfanmax]

The fundamental problem with the weighted win rate (as expressed by betiko) is that it is not properly normalized. You're multiplying a bunch of things in the numerator to give them additional weighting, but you're not dividing by a denominator that brings the total back under 100%.

We can rescue this by dividing everything by the proper denominator. The denominator could be equal to the weighted average of the multipliers. So in my example, we would take (0/20 * 1 + 10/20 * 6) == 3.0 and turn it into (0/20 * 1 + 10/20 * 6) / (10/20 * 1 + 10/20 * 6) == 0.857. This isn't a unique way to do it but is far more agreeable with our intuitions.

Actually, I don't think this works either. Because in the example where I only play one type of game, then the multipliers go away. I'll think more carefully about this -- I think there's an objectively good way to normalize it.

[betiko]

Mets, you have edited your post AFTER I commented on it; then I got fastposted by your second post and just went to bed. What I was commenting on had NO example and I have quoted exactly what I was replying to.
Let's just try to continue brainstorming on a better system and not turn this into a pissing contest.
What I replyied to was basically "no that sucks, try harder" and you weren't even bouncing back with a new solution. I see you did in the aftermath.
Of course my proposition can be improved, I actually tried to improve kaskavel and of course other people can improve it. Now let me adress your points and let's try to be a bit more civilised please.

You're going to have to do a lot more work before this makes sense. If I play 10 1v1 games and lose all of them, and then play 10 12 player games and win all of them, my win ratio according to the nominal formula would be 300%. Deciding to just call that 100% is not providing a subtle fix to an otherwise reasonable formula, it is exposing how inappropriate it is for measuring what we normally think of as win ratio.

No that is wrong. if you have played thse 20 games the way you said it (by the way, no way someone could get those numbers; no one would ever win 10 12 player games straight (unless they are noob farming in freestyle). If you have read what I said, and you said you did, you acknowledged that the win rate would be capped at 100% per game mode. In your example, the win rate with my method would be 50% (half of the games at 0%, half of the games at 100%)... so the same as CC. My method is ALWAYS between 0% and 100% as you would have a capping! Don't make me repeat it a fourth time.

In my proposal, you don't generally 'kill' anyone if you eliminate them in a game. The winner of any non-terminator games gets the credit for all eliminations in that game. It is only in terminator that this would vary, to give credit for terminator kills (since eliminating people -- even at the risk of losing the game -- is explicitly the goal of Terminator). This should have been obvious from my numerical example in which I automatically gave full credit to the winner of the 12 player games, or could have been obvious if you had asked what I meant instead of going off on a rant.

(Although even your rant is dubious. Most people care about their score way more than their profile's win ratio and it is hard to imagine them changing their behavior much to sacrifice the former for the latter.)

I might be wrong, but I have not seen you specify that if you don't win the game you don't get the kills. In fact, what you were trying to do, from what I understood was to reward kills to those who don't win games (and again; there was NO numerical example from you when I replyied, editing your old posts to insert numerical examples when the conversation is happening live is not the best thing to do.

I think you should want to. The current and proposed metric hurts people who play a lot of Terminator games and eliminate lots of opponents but end up losing overall.


If I play only 4 player games, then by your suggestion if I win half of them my 'aggregrate win ratio' is already 100%. If I win 60% of them, my 'win ratio' is 120%. This is true in general: as long as I do twice as good as the expected win rate, my 'aggregrate win ratio' will be 100%. Now I'll be the first to admit that doing twice as well as expectations is pretty damn hard, and an impressive achievement. But it's not so good that I can say I won 100% of the time.

I have mentioned it earlier. This is supposed to be an "aggregated expected win rate index". Just find me one player who can achieve 60% win rate on a game type where the expected win rate is 25% . And again: CAPPED AT 100%. But the more I think about it, the more I think taht it's dumb to cap... and I have no problem to see this going above 100%

If only we lived in a world where people accepted their mistakes and fixed them instead of insulting the person who pointed them out.

I have no problem recognising my mistakes, but I was answering to someone not helping and just provocating (before you edited your message...)

[Metsfanmax]

Mets, you have edited your post AFTER I commented on it; then I got fastposted by your second post and just went to bed. What I was commenting on had NO example and I have quoted exactly what I was replying to.

The post I objected to (the one talking about my "bullshit idea") had the numerical example in it, so I'm not sure where you're going with this.

You're going to have to do a lot more work before this makes sense. If I play 10 1v1 games and lose all of them, and then play 10 12 player games and win all of them, my win ratio according to the nominal formula would be 300%. Deciding to just call that 100% is not providing a subtle fix to an otherwise reasonable formula, it is exposing how inappropriate it is for measuring what we normally think of as win ratio.

No that is wrong. if you have played thse 20 games the way you said it (by the way, no way someone could get those numbers; no one would ever win 10 12 player games straight (unless they are noob farming in freestyle).

So basically your argument is that no one could ever get those numbers... except when they could.

I understand that actually getting above 100% in practice would be difficult. That is not the point. The point is the fact that you could demonstrates why this is a flawed system. Since it is a flawed system, it will still give flawed results even when those results are below 100%. The flaw is, as I stated in my above post, that this is not properly normalized. It is an inherent problem with the method that can be stated precisely mathematically, and that means you can't get around it by appealing to the fact that in practice no one will be able to tell that it is flawed.

If you have read what I said, and you said you did, you acknowledged that the win rate would be capped at 100% per game mode. In your example, the win rate with my method would be 50% (half of the games at 0%, half of the games at 100%)... so the same as CC. My method is ALWAYS between 0% and 100% as you would have a capping! Don't make me repeat it a fourth time.

My point was to explain that the capping you propose is an ad hoc bandage to fix a broken system; my numerical example is to show how badly you would need the capping to fix it. You are correctly recognizing the symptoms of an improper normalization, but you are fixing it the wrong way. The problem of course here is that you're trying to reward people more for winning games with more players, but then you start taking away that reward if they win too much and get above 100% on any game type. Essentially you are penalizing people for winning.

I might be wrong, but I have not seen you specify that if you don't win the game you don't get the kills. In fact, what you were trying to do, from what I understood was to reward kills to those who don't win games (and again; there was NO numerical example from you when I replyied, editing your old posts to insert numerical examples when the conversation is happening live is not the best thing to do.

So now that you understand what I am proposing, are you going to respond to it?

I have mentioned it earlier. This is supposed to be an "aggregated expected win rate index". Just find me one player who can achieve 60% win rate on a game type where the expected win rate is 25%.

Your argument is "my system may be broken but that's OK because people will never realize it." It is not a strong one.

And again: CAPPED AT 100%. But the more I think about it, the more I think taht it's dumb to cap... and I have no problem to see this going above 100%

Then it is clearly not a win ratio (which was my original point).

[betiko]

Honestly mets, you have a way of adressing issues that is so unconstructive and stubborn, that you managed to make me lose any interest to brainstorm further.
Pretty much like in any suggestion thread where you show up that you kill like a cancer. Congrats and good luck to anyone who would try to discuss the suggestion with this guy, not going to bother.

[Metsfanmax]

Honestly mets, you have a way of adressing issues that is so unconstructive and stubborn, that you managed to make me lose any interest to brainstorm further.
Pretty much like in any suggestion thread where you show up that you kill like a cancer. Congrats and good luck to anyone who would try to discuss the suggestion with this guy, not going to bother.

I don't believe this has been unconstructive: thanks to your start I came up with the precise idea we need to save this suggestion and turn it into something that is clearly better than the status quo. I don't have yet the exact details of that implementation worked out but we know what needs to be done and that is most of the battle. I would appreciate it if you helped with that, especially since you wanted this to happen.

(At any rate, I caution people to think carefully about what they're saying. You called me unconstructive, yet a few posts ago for no good reason you were saying that I have absolutely no relevance to CC. I'm ignoring this because the suggestion is more important than the egos, and I hope you will join me in this perspective.)