Forums / Community / Matchmaking Feedback & Discussion

CSR INCREASE WHEN BEATING A TEAM OF 3 IN Slayer

OP MRGEE3K

  1. 1
  2. ...
  3. 2
  4. 3
  5. ...
  6. 4
eLantern wrote:
ZaedynFel wrote:
Flip a coin, you will get 3 heads out of 4, doesn't mean it's not a fair coin.
Perfectly stated.
You also would describe MMR being correct one out of four matches as, having done well in predicting his match outcomes? What about zero out of four? That also can happen when a coin is flipped four times.

Those weren't coin flips. Expected win% in those three losses were 90,75, and 64%. Do you still think the use of a coin flip analogy was the perfect statement to justify describing one out of four correct predictions as, MMR doing well in predicting those four matches?
I was very hopeful that things would change with the new update but clearly things haven't.

I've played two games today where some scum bag on the other team quit when his team was getting killed and thus all our work for me and my teammates was for absolutely nothing.

Here's one of those games: https://www.halowaypoint.com/en-us/games/halo-5-guardians/xbox-one/mode/arena/matches/d9aa61e3-9fd5-4a80-b7b0-2fdedd26611f/players/cicconmr00?gameHistoryMatchIndex=0&gameHistoryGameModeFilter=Arena

Maybe this is only a problem for Golds/Platinums that don't represent their frustrations on here but we are probably the bulk of people who play Halo 5 so someone should care about how much quitting has affected this game.

I feel strongly I should have gone up here, but I'm sure you'll come up with some excuse how because my teams MMR was so much higher that I shouldn't....well why am I being paired in games where my teams MMR is that much higher, maybe that is why we have so many damn quitters.

I just don't get how I'm the only one who cares enough to complain about this....think about how negatively this rule impacts non-quitters....there's massive negative selection here, these scum bags who quit will artificially inflate the CSR of the losing team and artificially deflate the CSR of the winning team, it makes no sense. We go up 29 to 11 and we are penalized for it and the people on the losing team are rewarded?

When is this going to change?

If someone doesn't quit with in the time it would take to idle out from the start of the game then there should be no CSR impact on quitting. If you lose 3 on 4 after going down by 10-15 kills you deserve to drop rank.
ZaedynFel wrote:
It's pro-rated by how long the players are in the match. If they leave a the last second, you still get near full credit.
Idk.... I still think it's crap. If people rage-quit, we should deserve the full CSR increase. I mean look at this game from last night.They outclassed us by a little bit and I did the best in gamw.....but we made their whole team quit....yet my CSR went up literally like 1 or 2 points as a fresh Diamond 3 (haven't played ranked in multiple seasons now). I feel like the ranking structure is seriously flawed when it comes to gains and losses from winning and losing. It's always felt like losing drops you almost 30 points no matter what even if their MMR/ranks are higher or equal to your team, but winning only ever gives 1-5 CSR even at equal ranks for you.

It overcompensates severely and makes Ranked not even fun to play because you can spend 5 hours winning all of your games and then lose 1 match and all that progress is gone, even if a teammate lagged out and forced you down 3v4.
ZaedynFel wrote:
It's pro-rated by how long the players are in the match. If they leave a the last second, you still get near full credit.
Idk.... I still think it's crap. If people rage-quit, we should deserve the full CSR increase. I mean look at this game from last night.They outclassed us by a little bit and I did the best in gamw.....but we made their whole team quit....yet my CSR went up literally like 1 or 2 points as a fresh Diamond 3 (haven't played ranked in multiple seasons now). I feel like the ranking structure is seriously flawed when it comes to gains and losses from winning and losing. It's always felt like losing drops you almost 30 points no matter what even if their MMR/ranks are higher or equal to your team, but winning only ever gives 1-5 CSR even at equal ranks for you.

It overcompensates severely and makes Ranked not even fun to play because you can spend 5 hours winning all of your games and then lose 1 match and all that progress is gone, even if a teammate lagged out and forced you down 3v4.
At the end of the day, your CSR is accurate. If you are getting little for a win and less for a loss, and your predictions are still accurate, that means you are inflated or usually playing with people better than you.

But as long as the ranks are accurate, we're not going to hand out extra points that aren't deserved by your long term established performance.
ZaedynFel wrote:
cicconmr00 wrote:
Regarding the two above threads, I have an actual game example for you ZaedynFelPlease take a look at this game, I think we were down probably close to 10 when their 4th player quit:

https://www.halowaypoint.com/en-us/games/halo-5-guardians/xbox-one/mode/arena/matches/3171ce9e-d6c3-417a-baa4-fb5e1b925a51/players/cicconmr00?gameHistoryMatchIndex=0&gameHistoryGameModeFilter=All

I've been on an awful break passed 2 weeks since all my stuff. I guess is just bad luck but it just seems like every game I lose is 3 on 4 and every game I win is 4 on 3.

I feel like I'm playing well and I just can't move rank anywhere but down...I get to Plat 1 but my wins don't gain me any where near the CSR I need because I have so many wins in 4 on 3 matches.

This rule just sucks, it's totally ruined this game for me and I absolutely can't stand it. I want to play this game and enjoy it and it sucks that quitters just make it impossible to rank up unless they are on your team.

Just so demoralized right now....I need a new game, if anyone has any recommendations?
It treated your opponents as having 3.74 players, whereas you had 4.

You were 875 heading in to the match, and played against a 798 team.

The 798 get treated as a 746 because of the quitter.

So 875 losing to 746 (with your own 799 team) results in -26.

If they hadn't quit out, and assuming you still lost, it would have been -21 or so, I believe. So it made a difference, but not a huge one.
I have a question:

So what happens to the quitters? Do they lose max (30) for quitting/disconnecting right after it happens or does it wait until the match is over?

Asking because I was playing a match and got disconnected mid game after we were up. We ended up winning, but I think I ended up losing max, maybe. I couldn't tell. All I remember is that it ended up counting as a loss, which was a bummer.

Also, if that happens and your teammates decide to quit thereafter, what happens to them? Do you lose the same amount of CSR as the first quitter or the second quitter? I'm asking because there are times when someone quits and then a second person quits a bit later, then me and my teammate are just basically playing for our lives because we probably don't know if it's safe to quit because 1 of us is waiting for the other to quit, thinking the last 1 to quit will lose the least CSR!
D M4N8 wrote:
ZaedynFel wrote:
cicconmr00 wrote:
Regarding the two above threads, I have an actual game example for you ZaedynFelPlease take a look at this game, I think we were down probably close to 10 when their 4th player quit:

https://www.halowaypoint.com/en-us/games/halo-5-guardians/xbox-one/mode/arena/matches/3171ce9e-d6c3-417a-baa4-fb5e1b925a51/players/cicconmr00?gameHistoryMatchIndex=0&gameHistoryGameModeFilter=All

I've been on an awful break passed 2 weeks since all my stuff. I guess is just bad luck but it just seems like every game I lose is 3 on 4 and every game I win is 4 on 3.

I feel like I'm playing well and I just can't move rank anywhere but down...I get to Plat 1 but my wins don't gain me any where near the CSR I need because I have so many wins in 4 on 3 matches.

This rule just sucks, it's totally ruined this game for me and I absolutely can't stand it. I want to play this game and enjoy it and it sucks that quitters just make it impossible to rank up unless they are on your team.

Just so demoralized right now....I need a new game, if anyone has any recommendations?
It treated your opponents as having 3.74 players, whereas you had 4.

You were 875 heading in to the match, and played against a 798 team.

The 798 get treated as a 746 because of the quitter.

So 875 losing to 746 (with your own 799 team) results in -26.

If they hadn't quit out, and assuming you still lost, it would have been -21 or so, I believe. So it made a difference, but not a huge one.
I have a question:

So what happens to the quitters? Do they lose max (30) for quitting/disconnecting right after it happens or does it wait until the match is over?

Asking because I was playing a match and got disconnected mid game after we were up. We ended up winning, but I think I ended up losing max, maybe. I couldn't tell. All I remember is that it ended up counting as a loss, which was a bummer.

Also, if that happens and your teammates decide to quit thereafter, what happens to them? Do you lose the same amount of CSR as the first quitter or the second quitter? I'm asking because there are times when someone quits and then a second person quits a bit later, then me and my teammate are just basically playing for our lives because we probably don't know if it's safe to quit because 1 of us is waiting for the other to quit, thinking the last 1 to quit will lose the least CSR!
The first person to leave a game always loses 30 and gets a ban flag. If it's their first quit recently, they won't get banned. If not, they will get banned / escalated ban.

Anyone who quits after the first quitter loses what you would lose if you lost the match normally and they also don't get banned.
ZaedynFel wrote:
D M4N8 wrote:
ZaedynFel wrote:
cicconmr00 wrote:
Regarding the two above threads, I have an actual game example for you ZaedynFelPlease take a look at this game, I think we were down probably close to 10 when their 4th player quit:

https://www.halowaypoint.com/en-us/games/halo-5-guardians/xbox-one/mode/arena/matches/3171ce9e-d6c3-417a-baa4-fb5e1b925a51/players/cicconmr00?gameHistoryMatchIndex=0&gameHistoryGameModeFilter=All

I've been on an awful break passed 2 weeks since all my stuff. I guess is just bad luck but it just seems like every game I lose is 3 on 4 and every game I win is 4 on 3.

I feel like I'm playing well and I just can't move rank anywhere but down...I get to Plat 1 but my wins don't gain me any where near the CSR I need because I have so many wins in 4 on 3 matches.

This rule just sucks, it's totally ruined this game for me and I absolutely can't stand it. I want to play this game and enjoy it and it sucks that quitters just make it impossible to rank up unless they are on your team.

Just so demoralized right now....I need a new game, if anyone has any recommendations?
It treated your opponents as having 3.74 players, whereas you had 4.

You were 875 heading in to the match, and played against a 798 team.

The 798 get treated as a 746 because of the quitter.

So 875 losing to 746 (with your own 799 team) results in -26.

If they hadn't quit out, and assuming you still lost, it would have been -21 or so, I believe. So it made a difference, but not a huge one.
I have a question:

So what happens to the quitters? Do they lose max (30) for quitting/disconnecting right after it happens or does it wait until the match is over?

Asking because I was playing a match and got disconnected mid game after we were up. We ended up winning, but I think I ended up losing max, maybe. I couldn't tell. All I remember is that it ended up counting as a loss, which was a bummer.

Also, if that happens and your teammates decide to quit thereafter, what happens to them? Do you lose the same amount of CSR as the first quitter or the second quitter? I'm asking because there are times when someone quits and then a second person quits a bit later, then me and my teammate are just basically playing for our lives because we probably don't know if it's safe to quit because 1 of us is waiting for the other to quit, thinking the last 1 to quit will lose the least CSR!
The first person to leave a game always loses 30 and gets a ban flag. If it's their first quit recently, they won't get banned. If not, they will get banned / escalated ban.

Anyone who quits after the first quitter loses what you would lose if you lost the match normally and they also don't get banned.
What was the prediction success rate of Trueskill 2 for ranked matches last season overall and Team Arena, in particular?

I imagine that the matchmaker attempts to achieve matches as close to 50/50 as possible. What was the average that the matchmaker obtained last season in ranked matches in general and for Team Arena?
ZaedynFel wrote:
D M4N8 wrote:
ZaedynFel wrote:
cicconmr00 wrote:
Regarding the two above threads, I have an actual game example for you ZaedynFelPlease take a look at this game, I think we were down probably close to 10 when their 4th player quit:

https://www.halowaypoint.com/en-us/games/halo-5-guardians/xbox-one/mode/arena/matches/3171ce9e-d6c3-417a-baa4-fb5e1b925a51/players/cicconmr00?gameHistoryMatchIndex=0&gameHistoryGameModeFilter=All

I've been on an awful break passed 2 weeks since all my stuff. I guess is just bad luck but it just seems like every game I lose is 3 on 4 and every game I win is 4 on 3.

I feel like I'm playing well and I just can't move rank anywhere but down...I get to Plat 1 but my wins don't gain me any where near the CSR I need because I have so many wins in 4 on 3 matches.

This rule just sucks, it's totally ruined this game for me and I absolutely can't stand it. I want to play this game and enjoy it and it sucks that quitters just make it impossible to rank up unless they are on your team.

Just so demoralized right now....I need a new game, if anyone has any recommendations?
It treated your opponents as having 3.74 players, whereas you had 4.

You were 875 heading in to the match, and played against a 798 team.

The 798 get treated as a 746 because of the quitter.

So 875 losing to 746 (with your own 799 team) results in -26.

If they hadn't quit out, and assuming you still lost, it would have been -21 or so, I believe. So it made a difference, but not a huge one.
I have a question:

So what happens to the quitters? Do they lose max (30) for quitting/disconnecting right after it happens or does it wait until the match is over?

Asking because I was playing a match and got disconnected mid game after we were up. We ended up winning, but I think I ended up losing max, maybe. I couldn't tell. All I remember is that it ended up counting as a loss, which was a bummer.

Also, if that happens and your teammates decide to quit thereafter, what happens to them? Do you lose the same amount of CSR as the first quitter or the second quitter? I'm asking because there are times when someone quits and then a second person quits a bit later, then me and my teammate are just basically playing for our lives because we probably don't know if it's safe to quit because 1 of us is waiting for the other to quit, thinking the last 1 to quit will lose the least CSR!
The first person to leave a game always loses 30 and gets a ban flag. If it's their first quit recently, they won't get banned. If not, they will get banned / escalated ban.

Anyone who quits after the first quitter loses what you would lose if you lost the match normally and they also don't get banned.
What was the prediction success rate of Trueskill 2 for ranked matches last season overall and Team Arena, in particular?

I imagine that the matchmaker attempts to achieve matches as close to 50/50 as possible. What was the average that the matchmaker obtained last season in ranked matches in general and for Team Arena?
Around 62%, which is because we tune for about 60/40 at worst. The whole system is around 60% since we matchmake on it.
I have seen many responses each time breaking down why the ranking system correctly adjusted the csr of any given player complaining about it. Obviously since the system is not actually bugging, the response is always going to be "it worked like it was supposed to" "you are ranked where you should be". The real problem is that the model that has been created might not be actually good. I am aware that creating a good ranking system is not easy but I think it would be better to take people's comments and criticisms into consideration rather than just brushing them off and saying that system worked as intended.

Here's an example of a game in which I played where I believe a better ranking system would not have ranked me down so harshly.

https://www.halowaypoint.com/en-us/games/halo-5-guardians/xbox-one/mode/arena/matches/6d810719-f02a-42a3-9cbe-4015ea42529f/players/phat%20pat0101?gameHistoryMatchIndex=2&gameHistoryGameModeFilter=Arena

The other team had a player quit relatively early and we went on to lose the 4 v 3. We lost almost entirely because a player on my team went 3 and 16 (another went 8 and 16). I played very well going 20 and 6 and I wasn't trying to nurse a KD I was actively trying to help my teammates and win the game. I lost more than half a rank from this loss. Now I don't want an explanation about how the system did what it was supposed to in that situation because I'm sure it did. What I am saying is that I believe the model currently being used is lacking information and accuracy because it really makes no sense for someone to rank down in that situation. There's nothing I can do to stop my teammates just running in and getting themselves killed.
I have seen many responses each time breaking down why the ranking system correctly adjusted the csr of any given player complaining about it. Obviously since the system is not actually bugging, the response is always going to be "it worked like it was supposed to" "you are ranked where you should be". The real problem is that the model that has been created might not be actually good. I am aware that creating a good ranking system is not easy but I think it would be better to take people's comments and criticisms into consideration rather than just brushing them off and saying that system worked as intended.

Here's an example of a game in which I played where I believe a better ranking system would not have ranked me down so harshly.

https://www.halowaypoint.com/en-us/games/halo-5-guardians/xbox-one/mode/arena/matches/6d810719-f02a-42a3-9cbe-4015ea42529f/players/phat%20pat0101?gameHistoryMatchIndex=2&gameHistoryGameModeFilter=Arena

The other team had a player quit relatively early and we went on to lose the 4 v 3. We lost almost entirely because a player on my team went 3 and 16 (another went 8 and 16). I played very well going 20 and 6 and I wasn't trying to nurse a KD I was actively trying to help my teammates and win the game. I lost more than half a rank from this loss. Now I don't want an explanation about how the system did what it was supposed to in that situation because I'm sure it did. What I am saying is that I believe the model currently being used is lacking information and accuracy because it really makes no sense for someone to rank down in that situation. There's nothing I can do to stop my teammates just running in and getting themselves killed.
Losing a 4v3 will always have a slightly weird outcome because it's extraordinarily rare. But it would overcomplicate the model if we added every possible exception to it, and make it worse overall. We don't add things to the model that almost never happen because it makes things worse overall.

99.99999999999% of the time, players exactly like those on your team win that exact same type of matchup, so when you lose it, the system makes the logical assumption that people on your team aren't as good as it thought they were.

That said, you need to separate out what the visible system can do (CSR) and what the backend system models about you (MMR).

The visible system has no idea how well you played that match, so can only guess based on the win / loss outcome what to do, and 99.9999999% of the time, the correct thing to do is exactly what it did. So you lost 30 CSR.

The backend model (MMR), on the other hand, saw how well you did (kills) and did not dock your MMR as much as it did the other players on your team. For example, the 3-16 player lost an equivalent of 87 MMR from that match.

You, on the other hand, finished that match with an MMR still higher than your CSR, so, yes, when stuff like that happens, yes, you will lose a lot of CSR because the CSR system doesn't look at kills. But you will gain it back quickly in future wins because the system overall goes towards your MMR.

Also, as far as accuracy goes, the best judgement of a skill system is how often the higher MMR team wins the match, which with this system, is the highest I've ever seen. Bending over backward to add exceptions like this to the model would lower that accuracy. So better to deal with the occasional exception and move on.
So better to deal with the occasional exception and move on.

Spoken like a true developer. It's the exceptions that drive consumer and brand perception. People don't notice when they have a great experience because they expect it. They do notice when they have a bad one though...
CSevenZer0 wrote:
So better to deal with the occasional exception and move on.Spoken like a true developer. It's the exceptions that drive consumer and brand perception. People don't notice when they have a great experience because they expect it. They do notice when they have a bad one though...
Right, and adding an exception like this would result in a net increase in bad experiences.
ZaedynFel wrote:
CSevenZer0 wrote:
So better to deal with the occasional exception and move on.Spoken like a true developer. It's the exceptions that drive consumer and brand perception. People don't notice when they have a great experience because they expect it. They do notice when they have a bad one though...
Right, and adding an exception like this would result in a net increase in bad experiences.
You said the exception is extremely rare and to include it in the code would 'overcomplicate' it.

That's developer speak for not worth the effort (ie expense).
CSevenZer0 wrote:
ZaedynFel wrote:
CSevenZer0 wrote:
So better to deal with the occasional exception and move on.Spoken like a true developer. It's the exceptions that drive consumer and brand perception. People don't notice when they have a great experience because they expect it. They do notice when they have a bad one though...
Right, and adding an exception like this would result in a net increase in bad experiences.
You said the exception is extremely rare and to include it in the code would 'overcomplicate' it.

That's developer speak for not worth the effort (ie expense).
Uh, how long have you been a developer? That's not at all what I meant.

I meant overcomplicate from a modeling / data science / machine learning / neural networks point of view (what my PhD is in), which means if you add too much complexity to overfit to outliers and exceptions, it makes accuracy on everything else go down, hence the end to that sentence that says, "and make it worse overall"
ZaedynFel wrote:
CSevenZer0 wrote:
ZaedynFel wrote:
CSevenZer0 wrote:
So better to deal with the occasional exception and move on.Spoken like a true developer. It's the exceptions that drive consumer and brand perception. People don't notice when they have a great experience because they expect it. They do notice when they have a bad one though...
Right, and adding an exception like this would result in a net increase in bad experiences.
You said the exception is extremely rare and to include it in the code would 'overcomplicate' it.

That's developer speak for not worth the effort (ie expense).
Uh, how long have you been a developer? That's not at all what I meant.

I meant overcomplicate from a modeling / data science / machine learning / neural networks point of view (what my PhD is in), which means if you add too much complexity to overfit to outliers and exceptions, it makes accuracy on everything else go down, hence the end to that sentence that says, "and make it worse overall"
This notion that it is "exceedingly rare" for a team of 4 to lose to a team of 3 just seems like a lie to me.

What are the stats? I know from my personal experience it has to be close to 70/30 and the CSR hit is as if it is 99/1, granted it seems to have gotten a little better this winter.

But guys at the end of the day there's only one way to fix this problem GET RID OF WIN/LOSS IMPACT ON RANK.

Win/Loss dependance is what makes this game awful...getting rid of it eliminates any issues with quitting, cheating, de-ranking, bad teammates, and on and on.....we grew up playing slayer that's what we want to play....we want a ranking system that doesn't predict which "team" will win, we want a ranking system that reflects our individual contributions and I can say this system unequivocally does not do that. It's incredibly unfair to single players which are the vast majority of Halo players in my opinion.
This post has been edited by a moderator. Please do not bump.
*Original post. Click at your own discretion.
Spoiler:
Show
cicconmr00 wrote:
ZaedynFel wrote:
CSevenZer0 wrote:
ZaedynFel wrote:
CSevenZer0 wrote:
So better to deal with the occasional exception and move on.Spoken like a true developer. It's the exceptions that drive consumer and brand perception. People don't notice when they have a great experience because they expect it. They do notice when they have a bad one though...
Right, and adding an exception like this would result in a net increase in bad experiences.
You said the exception is extremely rare and to include it in the code would 'overcomplicate' it.

That's developer speak for not worth the effort (ie expense).
Uh, how long have you been a developer? That's not at all what I meant.

I meant overcomplicate from a modeling / data science / machine learning / neural networks point of view (what my PhD is in), which means if you add too much complexity to overfit to outliers and exceptions, it makes accuracy on everything else go down, hence the end to that sentence that says, "and make it worse overall"
This notion that it is "exceedingly rare" for a team of 4 to lose to a team of 3 just seems like a lie to me.

What are the stats? I know from my personal experience it has to be close to 70/30 and the CSR hit is as if it is 99/1, granted it seems to have gotten a little better this winter.

But guys at the end of the day there's only one way to fix this problem GET RID OF WIN/LOSS IMPACT ON RANK.

Win/Loss dependance is what makes this game awful...getting rid of it eliminates any issues with quitting, cheating, de-ranking, bad teammates, and on and on.....we grew up playing slayer that's what we want to play....we want a ranking system that doesn't predict which "team" will win, we want a ranking system that reflects our individual contributions and I can say this system unequivocally does not do that. It's incredibly unfair to single players which are the vast majority of Halo players in my opinion.

  • % of time 3 wins vs. 4 in Slayer: 0.6%
  • Win / Loss has by far been preferred by the majority of players both in feedback to us, and in what keeps them from quitting and what motivates them to keep playing. This is overwhelmingly true in the data
  • The current CSR gives more individual credit than any one we've ever had, while staying accurate to predicting your outcomes. Meaning, the current system is also super accurate at predicting your PERSONAL peformance as well, down to avg # of kills you will have in a match.
So, yeah, wrong.

In the data as well, solo players are fairing better than ever, with CSRs often higher than their counterparts that are in parties.
ZaedynFel wrote:
cicconmr00 wrote:
ZaedynFel wrote:
CSevenZer0 wrote:
ZaedynFel wrote:
CSevenZer0 wrote:
So better to deal with the occasional exception and move on.Spoken like a true developer. It's the exceptions that drive consumer and brand perception. People don't notice when they have a great experience because they expect it. They do notice when they have a bad one though...
Right, and adding an exception like this would result in a net increase in bad experiences.
You said the exception is extremely rare and to include it in the code would 'overcomplicate' it.

That's developer speak for not worth the effort (ie expense).
Uh, how long have you been a developer? That's not at all what I meant.

I meant overcomplicate from a modeling / data science / machine learning / neural networks point of view (what my PhD is in), which means if you add too much complexity to overfit to outliers and exceptions, it makes accuracy on everything else go down, hence the end to that sentence that says, "and make it worse overall"
This notion that it is "exceedingly rare" for a team of 4 to lose to a team of 3 just seems like a lie to me.

What are the stats? I know from my personal experience it has to be close to 70/30 and the CSR hit is as if it is 99/1, granted it seems to have gotten a little better this winter.

But guys at the end of the day there's only one way to fix this problem GET RID OF WIN/LOSS IMPACT ON RANK.

Win/Loss dependance is what makes this game awful...getting rid of it eliminates any issues with quitting, cheating, de-ranking, bad teammates, and on and on.....we grew up playing slayer that's what we want to play....we want a ranking system that doesn't predict which "team" will win, we want a ranking system that reflects our individual contributions and I can say this system unequivocally does not do that. It's incredibly unfair to single players which are the vast majority of Halo players in my opinion.
  • % of time 3 wins vs. 4 in Slayer: 0.6%
  • Win / Loss has by far been preferred by the majority of players both in feedback to us, and in what keeps them from quitting and what motivates them to keep playing. This is overwhelmingly true in the data
  • The current CSR gives more individual credit than any one we've ever had, while staying accurate to predicting your outcomes. Meaning, the current system is also super accurate at predicting your PERSONAL peformance as well, down to avg # of kills you will have in a match.
So, yeah, wrong.

In the data as well, solo players are fairing better than ever, with CSRs often higher than their counterparts that are in parties.
6 out of 1000? Is this for all games that are 3 v 4 at some point during the match or only for games that start 3 v 4?
ZaedynFel wrote:
cicconmr00 wrote:
ZaedynFel wrote:
CSevenZer0 wrote:
ZaedynFel wrote:
CSevenZer0 wrote:
So better to deal with the occasional exception and move on.Spoken like a true developer. It's the exceptions that drive consumer and brand perception. People don't notice when they have a great experience because they expect it. They do notice when they have a bad one though...
Right, and adding an exception like this would result in a net increase in bad experiences.
You said the exception is extremely rare and to include it in the code would 'overcomplicate' it.

That's developer speak for not worth the effort (ie expense).
Uh, how long have you been a developer? That's not at all what I meant.

I meant overcomplicate from a modeling / data science / machine learning / neural networks point of view (what my PhD is in), which means if you add too much complexity to overfit to outliers and exceptions, it makes accuracy on everything else go down, hence the end to that sentence that says, "and make it worse overall"
This notion that it is "exceedingly rare" for a team of 4 to lose to a team of 3 just seems like a lie to me.

What are the stats? I know from my personal experience it has to be close to 70/30 and the CSR hit is as if it is 99/1, granted it seems to have gotten a little better this winter.

But guys at the end of the day there's only one way to fix this problem GET RID OF WIN/LOSS IMPACT ON RANK.

Win/Loss dependance is what makes this game awful...getting rid of it eliminates any issues with quitting, cheating, de-ranking, bad teammates, and on and on.....we grew up playing slayer that's what we want to play....we want a ranking system that doesn't predict which "team" will win, we want a ranking system that reflects our individual contributions and I can say this system unequivocally does not do that. It's incredibly unfair to single players which are the vast majority of Halo players in my opinion.
  • % of time 3 wins vs. 4 in Slayer: 0.6%
  • Win / Loss has by far been preferred by the majority of players both in feedback to us, and in what keeps them from quitting and what motivates them to keep playing. This is overwhelmingly true in the data
  • The current CSR gives more individual credit than any one we've ever had, while staying accurate to predicting your outcomes. Meaning, the current system is also super accurate at predicting your PERSONAL peformance as well, down to avg # of kills you will have in a match.
So, yeah, wrong.

In the data as well, solo players are fairing better than ever, with CSRs often higher than their counterparts that are in parties.
6 out of 1000? Is this for all games that are 3 v 4 at some point during the match or only for games that start 3 v 4?
It includes both. Anything less than having 4 players the entire match.
ZaedynFel wrote:
ZaedynFel wrote:
cicconmr00 wrote:
ZaedynFel wrote:
CSevenZer0 wrote:
ZaedynFel wrote:
CSevenZer0 wrote:
So better to deal with the occasional exception and move on.Spoken like a true developer. It's the exceptions that drive consumer and brand perception. People don't notice when they have a great experience because they expect it. They do notice when they have a bad one though...
Right, and adding an exception like this would result in a net increase in bad experiences.
You said the exception is extremely rare and to include it in the code would 'overcomplicate' it.

That's developer speak for not worth the effort (ie expense).
Uh, how long have you been a developer? That's not at all what I meant.

I meant overcomplicate from a modeling / data science / machine learning / neural networks point of view (what my PhD is in), which means if you add too much complexity to overfit to outliers and exceptions, it makes accuracy on everything else go down, hence the end to that sentence that says, "and make it worse overall"
This notion that it is "exceedingly rare" for a team of 4 to lose to a team of 3 just seems like a lie to me.

What are the stats? I know from my personal experience it has to be close to 70/30 and the CSR hit is as if it is 99/1, granted it seems to have gotten a little better this winter.

But guys at the end of the day there's only one way to fix this problem GET RID OF WIN/LOSS IMPACT ON RANK.

Win/Loss dependance is what makes this game awful...getting rid of it eliminates any issues with quitting, cheating, de-ranking, bad teammates, and on and on.....we grew up playing slayer that's what we want to play....we want a ranking system that doesn't predict which "team" will win, we want a ranking system that reflects our individual contributions and I can say this system unequivocally does not do that. It's incredibly unfair to single players which are the vast majority of Halo players in my opinion.
  • % of time 3 wins vs. 4 in Slayer: 0.6%
  • Win / Loss has by far been preferred by the majority of players both in feedback to us, and in what keeps them from quitting and what motivates them to keep playing. This is overwhelmingly true in the data
  • The current CSR gives more individual credit than any one we've ever had, while staying accurate to predicting your outcomes. Meaning, the current system is also super accurate at predicting your PERSONAL peformance as well, down to avg # of kills you will have in a match.
So, yeah, wrong.

In the data as well, solo players are fairing better than ever, with CSRs often higher than their counterparts that are in parties.
6 out of 1000? Is this for all games that are 3 v 4 at some point during the match or only for games that start 3 v 4?
It includes both. Anything less than having 4 players the entire match.
Actually, no, let's break it down better

2.75 - 3 players is 0.1% (one player leaves at the beginning, another halfway through)
3 to 3.25 players is 0.4% (one player leaves somewhere between the beginning and 25% of the way through)
3.25 to 3.5 players is 1.3% (one player leaves somewhere between 25% to 50% of the way through)
3.5 to 3.75 players is 1.4% (one player 50%-75% of the way through)
3.9 to 4 (but not including 4) is ironically 0.5% (one player leaves 90% of the way through)

The reason they are so low is because people quit when they know they're going to lose.

~99% of matches with even one quitter are lost by the quitting side.
  1. 1
  2. ...
  3. 2
  4. 3
  5. ...
  6. 4