Forums / Community / Matchmaking Feedback & Discussion

[Locked] Matchmaking Feedback Update – May 21

OP ZaedynFel

  1. 1
  2. ...
  3. 8
  4. 9
  5. 10
  6. 11
  7. 12
  8. ...
  9. 40
eLantern wrote:
ZaedynFel wrote:
ZaedynFel wrote:
Chimera30 wrote:
ZaedynFel, what is the maximum delta (positive or negative) that MMR can shift from a single game? For example, if you lose a match that trueskill2 determined you had 70% odds of winning (or whatever the max odds the matchmaker allows when expanded skill gaps).

What is the maximum delta CSR can shift?
In Slayer, in the last 24 hours, the largest change was an equivalent of 1205 in CSR.
For reference, someone expected to go 9-6 went 34-4
Wouldn't that get flagged as a potential outlier which would then simply get ignored in how it would affect a player's MMR/CSR?
That wasn't crazy enough given all the other circumstances of the match / expected randomness.

Also, our flagging alert has a high tolerance to prevent false positives.
eLantern wrote:
Because I carried your -Yoink- buddy... lol. j/k

It might be in-part because we were playing in a full party and winning most of our matches during the placement which also didn't feature the typical -100/200 CSR after placement like previous seasons. A few unexpected wins may have boosted us up -- I mean that's probably what landed me into Onyx a couple times before receding down to Diamond 5.
I don't think I ranked in with you on this account...

ZaedynFel wrote:
Oh, that was your average across all your matches, so not really a recent measure. I think you're currently low Diamond though. Here's your graph for the season. Also, a good example of how much MMR can move around.
Wow. Lots of fluctuation there. Also, why the gaps? (And I think thats more than just this season, as I only have 180 Slayer games played, but that's okay.)

ZaedynFel wrote:
Does MMR ever overadjust and then have to correct itself after such a large adjustment?
Hmm. It would be more accurate to say that sometimes MMR isn't confident in your skill, so allows itself to make large adjustments back and forth until it is.
Are the large jumps typical, or do I just over/under-perform frequently, so MMR isn't very confident in my skill?

I don't know, I guess I was just expecting something a little smoother...

Edit: Does TS2 still have the two variables (mean and variance/confidence) to identify skill, or has the need for that been eliminated with the new calculations?
ZaedynFel wrote:
Chimera30 wrote:
ZaedynFel wrote:
Chimera30 wrote:
ZaedynFel, what is the maximum delta (positive or negative) that MMR can shift from a single game? For example, if you lose a match that trueskill2 determined you had 70% odds of winning (or whatever the max odds the matchmaker allows when expanded skill gaps).

What is the maximum delta CSR can shift?
In Slayer, in the last 24 hours, the largest change was an equivalent of 1205 in CSR.
And that was a change in MMR, yes? What would that change be like in terms of tiers (Bronze, Silver, Gold, etc.)? Is going up 1205 points in MMR like going from Plat 1 to Diamond 1 in a single game?

Are changes in CSR limited to a maximum of +/- 30 for any given game?
1205 points is like going from the very bottom of bronze (which is 0) to Diamond 1 in one match.
Sweet Christmas.

Just speculating but if MMR can shift so greatly in a single game, is that why CSR is used as the public rank? Because CSR fluctuations are capped so it's a more stable value?
Chimera30 wrote:
ZaedynFel wrote:
Chimera30 wrote:
ZaedynFel wrote:
Chimera30 wrote:
ZaedynFel, what is the maximum delta (positive or negative) that MMR can shift from a single game? For example, if you lose a match that trueskill2 determined you had 70% odds of winning (or whatever the max odds the matchmaker allows when expanded skill gaps).

What is the maximum delta CSR can shift?
In Slayer, in the last 24 hours, the largest change was an equivalent of 1205 in CSR.
And that was a change in MMR, yes? What would that change be like in terms of tiers (Bronze, Silver, Gold, etc.)? Is going up 1205 points in MMR like going from Plat 1 to Diamond 1 in a single game?

Are changes in CSR limited to a maximum of +/- 30 for any given game?
1205 points is like going from the very bottom of bronze (which is 0) to Diamond 1 in one match.
Sweet Christmas.

Just speculating but if MMR can shift so greatly in a single game, is that why CSR is used as the public rank? Because CSR fluctuations are capped so it's a more stable value?
Yeah, that was part of the original reasoning behind the first time I used an MMR in a game.
eLantern wrote:
Because I carried your -Yoink- buddy... lol. j/k

It might be in-part because we were playing in a full party and winning most of our matches during the placement which also didn't feature the typical -100/200 CSR after placement like previous seasons. A few unexpected wins may have boosted us up -- I mean that's probably what landed me into Onyx a couple times before receding down to Diamond 5.
I don't think I ranked in with you on this account...

ZaedynFel wrote:
Oh, that was your average across all your matches, so not really a recent measure. I think you're currently low Diamond though. Here's your graph for the season. Also, a good example of how much MMR can move around.
Wow. Lots of fluctuation there. Also, why the gaps? (And I think thats more than just this season, as I only have 180 Slayer games played, but that's okay.)

ZaedynFel wrote:
Does MMR ever overadjust and then have to correct itself after such a large adjustment?
Hmm. It would be more accurate to say that sometimes MMR isn't confident in your skill, so allows itself to make large adjustments back and forth until it is.
Are the large jumps typical, or do I just over/under-perform frequently, so MMR isn't very confident in my skill?

I don't know, I guess I was just expecting something a little smoother...

Edit: Does TS2 still have the two variables (mean and variance/confidence) to identify skill, or has the need for that been eliminated with the new calculations?
It has both a mean and variance.

The gaps are where you were playing other playlists. The app I'm using doesn't make it easy to just graph on continuous rating, and I'm too lazy to switch.
eLantern wrote:
Because I carried your -Yoink- buddy... lol. j/k

It might be in-part because we were playing in a full party and winning most of our matches during the placement which also didn't feature the typical -100/200 CSR after placement like previous seasons. A few unexpected wins may have boosted us up -- I mean that's probably what landed me into Onyx a couple times before receding down to Diamond 5.
I don't think I ranked in with you on this account...
I thought you did, but you and others had a few games in already before I joined the group.
Couple questions:

1. When will the season end? Really would have been better to do a reset as opposed to automatically forcing people to rank down with no clue how far they will drop since MMR is hidden (unless they win 94% of their games). IMO

2. Would you be able to tell me (text or graph) where my skill is in Slayer (bearhound) and the same for timetogetnastyy? As I had previously mentioned, we have a 60% win rate this season with positive KD but are rank locked, and we really have no plans to jump back into Slayer until next season since we have no idea how far we’ll drop. At least if we know where the drop will end, we could rank down to that and then at least see how bad this new rank lock is.

3. We saw a previous failed attempt (halo: reach) to base rank on Individual performance instead of just win/loss. I’m curious why we are going back to this? I don’t know if you’ll respond to this point, but it seems odd to me and my friends that MMR is your ultimate skill cap, which accounts for individual performance. Won’t this lead to more contesting over power weapons within your team and a generally aggressive play style since you need to outperform instead of just focus on making the plays to win?
ZaedynFel wrote:
Chimera30 wrote:
ZaedynFel wrote:
Chimera30 wrote:
ZaedynFel, what is the maximum delta (positive or negative) that MMR can shift from a single game? For example, if you lose a match that trueskill2 determined you had 70% odds of winning (or whatever the max odds the matchmaker allows when expanded skill gaps).

What is the maximum delta CSR can shift?
In Slayer, in the last 24 hours, the largest change was an equivalent of 1205 in CSR.
And that was a change in MMR, yes? What would that change be like in terms of tiers (Bronze, Silver, Gold, etc.)? Is going up 1205 points in MMR like going from Plat 1 to Diamond 1 in a single game?

Are changes in CSR limited to a maximum of +/- 30 for any given game?
1205 points is like going from the very bottom of bronze (which is 0) to Diamond 1 in one match.
Holy smokes. No wonder smurf accounts are basically non-existent now. Good work
bearhound wrote:
Couple questions:

1. When will the season end? Really would have been better to do a reset as opposed to automatically forcing people to rank down with no clue how far they will drop since MMR is hidden (unless they win 94% of their games). IMO

2. Would you be able to tell me (text or graph) where my skill is in Slayer (bearhound) and the same for timetogetnastyy? As I had previously mentioned, we have a 60% win rate this season with positive KD but are rank locked, and we really have no plans to jump back into Slayer until next season since we have no idea how far we’ll drop. At least if we know where the drop will end, we could rank down to that and then at least see how bad this new rank lock is.

3. We saw a previous failed attempt (halo: reach) to base rank on Individual performance instead of just win/loss. I’m curious why we are going back to this? I don’t know if you’ll respond to this point, but it seems odd to me and my friends that MMR is your ultimate skill cap, which accounts for individual performance. Won’t this lead to more contesting over power weapons within your team and a generally aggressive play style since you need to outperform instead of just focus on making the plays to win?
1. We haven't announced a date yet, but probably in line with the usual length.
2. I show bearhound as Diamond 3 and timetogetnastyy as Diamond 2
3. This is a completely different system from Reach's. Reach used a bunch of adhoc weights without statistically associating them with wins. It was awful at actually predicting who would win, which is why the motivations got all out of wack. This is TrueSkill2, which is just an extension of the same system we've always used, not some separate system. As such, everything in the system is anchored on wins. We don't weight kills towards your MMR, we predict both kills and wins from the same rating, holding wins as the absolute gate (no team can go up in rating if they lose). Because it's super accurate, it doesn't have issues with incentives (see also my GDC talk). You have to win still. Kills, deaths, etc., are only considered inasmuch as they inform the win. The same system has been running in Gears 4 Ranks since its launch without issues (20 months).
ZaedynFel wrote:
bearhound wrote:
Couple questions:

1. When will the season end? Really would have been better to do a reset as opposed to automatically forcing people to rank down with no clue how far they will drop since MMR is hidden (unless they win 94% of their games). IMO

2. Would you be able to tell me (text or graph) where my skill is in Slayer (bearhound) and the same for timetogetnastyy? As I had previously mentioned, we have a 60% win rate this season with positive KD but are rank locked, and we really have no plans to jump back into Slayer until next season since we have no idea how far we’ll drop. At least if we know where the drop will end, we could rank down to that and then at least see how bad this new rank lock is.

3. We saw a previous failed attempt (halo: reach) to base rank on Individual performance instead of just win/loss. I’m curious why we are going back to this? I don’t know if you’ll respond to this point, but it seems odd to me and my friends that MMR is your ultimate skill cap, which accounts for individual performance. Won’t this lead to more contesting over power weapons within your team and a generally aggressive play style since you need to outperform instead of just focus on making the plays to win?
1. We haven't announced a date yet, but probably in line with the usual length.
2. I show bearhound as Diamond 3 and timetogetnastyy as Diamond 2
3. This is a completely different system from Reach's. Reach used a bunch of adhoc weights without statistically associating them with wins. It was awful at actually predicting who would win, which is why the motivations got all out of wack. This is TrueSkill2, which is just an extension of the same system we've always used, not some separate system. As such, everything in the system is anchored on wins. We don't weight kills towards your MMR, we predict both kills and wins from the same rating, holding wins as the absolute gate (no team can go up in rating if they lose). Because it's super accurate, it doesn't have issues with incentives (see also my GDC talk). You have to win still. Kills, deaths, etc., are only considered inasmuch as they inform the win. The same system has been running in Gears 4 Ranks since its launch without issues (20 months).
Thanks so much for the reply. As it pertains to this season, at least with my duo, we’ll likely play less (at least in Slayer) because in our minds there’s no point in grinding for wins when we have so far to drop due to inflated CSR. I’ll reserve judgment on the new way CSR is distributed until next season (or at least in a different playlist — we ranked in HCS after this New implementatation so I would assume we don’t have inflated CSR there).
bearhound wrote:
ZaedynFel wrote:
bearhound wrote:
Couple questions:

1. When will the season end? Really would have been better to do a reset as opposed to automatically forcing people to rank down with no clue how far they will drop since MMR is hidden (unless they win 94% of their games). IMO

2. Would you be able to tell me (text or graph) where my skill is in Slayer (bearhound) and the same for timetogetnastyy? As I had previously mentioned, we have a 60% win rate this season with positive KD but are rank locked, and we really have no plans to jump back into Slayer until next season since we have no idea how far we’ll drop. At least if we know where the drop will end, we could rank down to that and then at least see how bad this new rank lock is.

3. We saw a previous failed attempt (halo: reach) to base rank on Individual performance instead of just win/loss. I’m curious why we are going back to this? I don’t know if you’ll respond to this point, but it seems odd to me and my friends that MMR is your ultimate skill cap, which accounts for individual performance. Won’t this lead to more contesting over power weapons within your team and a generally aggressive play style since you need to outperform instead of just focus on making the plays to win?
1. We haven't announced a date yet, but probably in line with the usual length.
2. I show bearhound as Diamond 3 and timetogetnastyy as Diamond 2
3. This is a completely different system from Reach's. Reach used a bunch of adhoc weights without statistically associating them with wins. It was awful at actually predicting who would win, which is why the motivations got all out of wack. This is TrueSkill2, which is just an extension of the same system we've always used, not some separate system. As such, everything in the system is anchored on wins. We don't weight kills towards your MMR, we predict both kills and wins from the same rating, holding wins as the absolute gate (no team can go up in rating if they lose). Because it's super accurate, it doesn't have issues with incentives (see also my GDC talk). You have to win still. Kills, deaths, etc., are only considered inasmuch as they inform the win. The same system has been running in Gears 4 Ranks since its launch without issues (20 months).
Thanks so much for the reply. As it pertains to this season, at least with my duo, we’ll likely play less (at least in Slayer) because in our minds there’s no point in grinding for wins when we have so far to drop due to inflated CSR. I’ll reserve judgment on the new way CSR is distributed until next season (or at least in a different playlist — we ranked in HCS after this New implementatation so I would assume we don’t have inflated CSR there).
That's right, if you recently ranked in HCS, it'll be in sync. Though as a warning, HCS has a much pickier distribution. It's really hard to get above Diamond.
ZaedynFel wrote:
bearhound wrote:
ZaedynFel wrote:
bearhound wrote:
Couple questions:

1. When will the season end? Really would have been better to do a reset as opposed to automatically forcing people to rank down with no clue how far they will drop since MMR is hidden (unless they win 94% of their games). IMO

2. Would you be able to tell me (text or graph) where my skill is in Slayer (bearhound) and the same for timetogetnastyy? As I had previously mentioned, we have a 60% win rate this season with positive KD but are rank locked, and we really have no plans to jump back into Slayer until next season since we have no idea how far we’ll drop. At least if we know where the drop will end, we could rank down to that and then at least see how bad this new rank lock is.

3. We saw a previous failed attempt (halo: reach) to base rank on Individual performance instead of just win/loss. I’m curious why we are going back to this? I don’t know if you’ll respond to this point, but it seems odd to me and my friends that MMR is your ultimate skill cap, which accounts for individual performance. Won’t this lead to more contesting over power weapons within your team and a generally aggressive play style since you need to outperform instead of just focus on making the plays to win?
1. We haven't announced a date yet, but probably in line with the usual length.
2. I show bearhound as Diamond 3 and timetogetnastyy as Diamond 2
3. This is a completely different system from Reach's. Reach used a bunch of adhoc weights without statistically associating them with wins. It was awful at actually predicting who would win, which is why the motivations got all out of wack. This is TrueSkill2, which is just an extension of the same system we've always used, not some separate system. As such, everything in the system is anchored on wins. We don't weight kills towards your MMR, we predict both kills and wins from the same rating, holding wins as the absolute gate (no team can go up in rating if they lose). Because it's super accurate, it doesn't have issues with incentives (see also my GDC talk). You have to win still. Kills, deaths, etc., are only considered inasmuch as they inform the win. The same system has been running in Gears 4 Ranks since its launch without issues (20 months).
Thanks so much for the reply. As it pertains to this season, at least with my duo, we’ll likely play less (at least in Slayer) because in our minds there’s no point in grinding for wins when we have so far to drop due to inflated CSR. I’ll reserve judgment on the new way CSR is distributed until next season (or at least in a different playlist — we ranked in HCS after this New implementatation so I would assume we don’t have inflated CSR there).
That's right, if you recently ranked in HCS, it'll be in sync. Though as a warning, HCS has a much pickier distribution. It's really hard to get above Diamond.
Yeah we’ve seen that in the past from our brief time in that playlist. But at least we’re not inflated so we can see how it goes. Thanks for the info.

One other question, I know you’re busy so if you can’t get to it that’s okay: it was my understanding that at the beginning of this season everyone still got ranked below their MMR, and then the inflation occurred if they continued placing and earning too much CSR. If that’s correct, I initially placed diamond 4 and my friend low onyx, but now, after winning 60% of our matches and going positive, our CSR is inflated and our true ranks are D3 and D2. If our initial placements were accurate, why did our MMR drop so much? (Or were we initially inflated from the get go?)
bearhound wrote:
ZaedynFel wrote:
bearhound wrote:
ZaedynFel wrote:
bearhound wrote:
Couple questions:

1. When will the season end? Really would have been better to do a reset as opposed to automatically forcing people to rank down with no clue how far they will drop since MMR is hidden (unless they win 94% of their games). IMO

2. Would you be able to tell me (text or graph) where my skill is in Slayer (bearhound) and the same for timetogetnastyy? As I had previously mentioned, we have a 60% win rate this season with positive KD but are rank locked, and we really have no plans to jump back into Slayer until next season since we have no idea how far we’ll drop. At least if we know where the drop will end, we could rank down to that and then at least see how bad this new rank lock is.

3. We saw a previous failed attempt (halo: reach) to base rank on Individual performance instead of just win/loss. I’m curious why we are going back to this? I don’t know if you’ll respond to this point, but it seems odd to me and my friends that MMR is your ultimate skill cap, which accounts for individual performance. Won’t this lead to more contesting over power weapons within your team and a generally aggressive play style since you need to outperform instead of just focus on making the plays to win?
1. We haven't announced a date yet, but probably in line with the usual length.
2. I show bearhound as Diamond 3 and timetogetnastyy as Diamond 2
3. This is a completely different system from Reach's. Reach used a bunch of adhoc weights without statistically associating them with wins. It was awful at actually predicting who would win, which is why the motivations got all out of wack. This is TrueSkill2, which is just an extension of the same system we've always used, not some separate system. As such, everything in the system is anchored on wins. We don't weight kills towards your MMR, we predict both kills and wins from the same rating, holding wins as the absolute gate (no team can go up in rating if they lose). Because it's super accurate, it doesn't have issues with incentives (see also my GDC talk). You have to win still. Kills, deaths, etc., are only considered inasmuch as they inform the win. The same system has been running in Gears 4 Ranks since its launch without issues (20 months).
Thanks so much for the reply. As it pertains to this season, at least with my duo, we’ll likely play less (at least in Slayer) because in our minds there’s no point in grinding for wins when we have so far to drop due to inflated CSR. I’ll reserve judgment on the new way CSR is distributed until next season (or at least in a different playlist — we ranked in HCS after this New implementatation so I would assume we don’t have inflated CSR there).
That's right, if you recently ranked in HCS, it'll be in sync. Though as a warning, HCS has a much pickier distribution. It's really hard to get above Diamond.
Yeah we’ve seen that in the past from our brief time in that playlist. But at least we’re not inflated so we can see how it goes. Thanks for the info.

One other question, I know you’re busy so if you can’t get to it that’s okay: it was my understanding that at the beginning of this season everyone still got ranked below their MMR, and then the inflation occurred if they continued placing and earning too much CSR. If that’s correct, I initially placed diamond 4 and my friend low onyx, but now, after winning 60% of our matches and going positive, our CSR is inflated and our true ranks are D3 and D2. If our initial placements were accurate, why did our MMR drop so much? (Or were we initially inflated from the get go?)
I still have time's graph up. He has kissed onyx a couple of times. He fluctuates top to bottom of Diamond. So not so much inflation / deflation, just some players tilt and some don't.

Slayer also tends to have more tilt to it than, e.g. the HCS ones. Maybe because of the range of maps and weapons is smaller.
ZaedynFel wrote:
bearhound wrote:
ZaedynFel wrote:
bearhound wrote:
ZaedynFel wrote:
bearhound wrote:
Couple questions:

1. When will the season end? Really would have been better to do a reset as opposed to automatically forcing people to rank down with no clue how far they will drop since MMR is hidden (unless they win 94% of their games). IMO

2. Would you be able to tell me (text or graph) where my skill is in Slayer (bearhound) and the same for timetogetnastyy? As I had previously mentioned, we have a 60% win rate this season with positive KD but are rank locked, and we really have no plans to jump back into Slayer until next season since we have no idea how far we’ll drop. At least if we know where the drop will end, we could rank down to that and then at least see how bad this new rank lock is.

3. We saw a previous failed attempt (halo: reach) to base rank on Individual performance instead of just win/loss. I’m curious why we are going back to this? I don’t know if you’ll respond to this point, but it seems odd to me and my friends that MMR is your ultimate skill cap, which accounts for individual performance. Won’t this lead to more contesting over power weapons within your team and a generally aggressive play style since you need to outperform instead of just focus on making the plays to win?
1. We haven't announced a date yet, but probably in line with the usual length.
2. I show bearhound as Diamond 3 and timetogetnastyy as Diamond 2
3. This is a completely different system from Reach's. Reach used a bunch of adhoc weights without statistically associating them with wins. It was awful at actually predicting who would win, which is why the motivations got all out of wack. This is TrueSkill2, which is just an extension of the same system we've always used, not some separate system. As such, everything in the system is anchored on wins. We don't weight kills towards your MMR, we predict both kills and wins from the same rating, holding wins as the absolute gate (no team can go up in rating if they lose). Because it's super accurate, it doesn't have issues with incentives (see also my GDC talk). You have to win still. Kills, deaths, etc., are only considered inasmuch as they inform the win. The same system has been running in Gears 4 Ranks since its launch without issues (20 months).
Thanks so much for the reply. As it pertains to this season, at least with my duo, we’ll likely play less (at least in Slayer) because in our minds there’s no point in grinding for wins when we have so far to drop due to inflated CSR. I’ll reserve judgment on the new way CSR is distributed until next season (or at least in a different playlist — we ranked in HCS after this New implementatation so I would assume we don’t have inflated CSR there).
That's right, if you recently ranked in HCS, it'll be in sync. Though as a warning, HCS has a much pickier distribution. It's really hard to get above Diamond.
Yeah we’ve seen that in the past from our brief time in that playlist. But at least we’re not inflated so we can see how it goes. Thanks for the info.

One other question, I know you’re busy so if you can’t get to it that’s okay: it was my understanding that at the beginning of this season everyone still got ranked below their MMR, and then the inflation occurred if they continued placing and earning too much CSR. If that’s correct, I initially placed diamond 4 and my friend low onyx, but now, after winning 60% of our matches and going positive, our CSR is inflated and our true ranks are D3 and D2. If our initial placements were accurate, why did our MMR drop so much? (Or were we initially inflated from the get go?)
I still have time's graph up. He has kissed onyx a couple of times. He fluctuates top to bottom of Diamond. So not so much inflation / deflation, just some players tilt and some don't.

Slayer also tends to have more tilt to it than, e.g. the HCS ones. Maybe because of the range of maps and weapons is smaller.
Huh, interesting. Thanks for the response.
ZaedynFel wrote:
This is TrueSkill2, which is just an extension of the same system we've always used, not some separate system. As such, everything in the system is anchored on wins. We don't weight kills towards your MMR, we predict both kills and wins from the same rating, holding wins as the absolute gate (no team can go up in rating if they lose). Because it's super accurate, it doesn't have issues with incentives (see also my GDC talk). You have to win still. Kills, deaths, etc., are only considered inasmuch as they inform the win.
I found this comment particularly interesting.

I was under the impression that the distribution of MMR gain or loss per team could vary. Thus an individual's TS2 Match-Make Rating (MMR) could potentially go up or hold pat based on their performance even if their team loss, but it may require some unique circumstances to occur such as the system predicting the team that player was on as underdogs (aka expected to lose) while the player's individual performance (Kills, Deaths, & K/D ratio) managed to greatly exceed the system's expectations for them.

Now, your comment made mention that, "no TEAM can go up in rating if they lose" which will still remain true because other players on the team can simply take a greater negative impact to their MMRs in order to offset any positive or non-impact granted to the MMR of the player who performed exceedingly well. Therefore, while you may not over-weigh kills, deaths, etc towards a player's MMR because it still holds that vital connection to TEAM win or loss it still plays a fairly influential role toward individual adjustments; meaning, that player performance (out-performing or under-performing) in relation to a win or loss can modify the MMR impact at the individual level. In other words, if a team loses the team won't see a net MMR positive gain, but it is entirely possible that an individual or individuals could potentially see a positive gain or a hold on their MMR based on their performance, right? Also, on a winning team each individual isn't necessarily awarded the same amount of positive gain toward their MMR as the amount can be influenced by how each individual performed in accordance with their expectations, right?
eLantern wrote:
ZaedynFel wrote:
This is TrueSkill2, which is just an extension of the same system we've always used, not some separate system. As such, everything in the system is anchored on wins. We don't weight kills towards your MMR, we predict both kills and wins from the same rating, holding wins as the absolute gate (no team can go up in rating if they lose). Because it's super accurate, it doesn't have issues with incentives (see also my GDC talk). You have to win still. Kills, deaths, etc., are only considered inasmuch as they inform the win.
I found this comment particularly interesting.

I was under the impression that the distribution of MMR gain or loss per team could vary. Thus an individual's TS2 Match-Make Rating (MMR) could potentially go up based on their performance even if their team loss, but it may require some unique circumstances to occur such as the system predicting the team that player was on as underdogs (aka expected to lose) while the player's individual performance (Kills, Deaths, & K/D ratio) managed to greatly exceed the system's expectations for them. Now your comment made mention that, "no TEAM can go up in rating if they lose" which can still remain true because other players on the team could simply take a greater negative impact to their MMRs in order to offset any positive impact granted to the MMR of the player who performed exceedingly well. Therefore, while you may not over-weigh kills, deaths, etc towards a player's MMR because it still holds that vital connection to team win or loss it still plays a fairly influential role toward individual adjustments; meaning, that player performance (out-performing or under-performing) in relation to a win or loss can modify the MMR impact at the individual level. In other words, if a team loses the team won't see a net MMR positive gain, but it is entirely possible that an individual or individuals could potentially see a positive gain to their MMR based on their performance, right? Also, on a winning team each individual isn't necessarily awarded the same amount of positive gain toward their MMR as the amount can be influenced by how each individual performed in accordance with their expectations, right?
Yes. It's like there's a pot of points to go around. If two teams also perform almost exactly the same (like 50-49 slayer, etc.), it's possible for the overall change between teams to be very small.
ZaedynFel wrote:
eLantern wrote:
ZaedynFel wrote:
This is TrueSkill2, which is just an extension of the same system we've always used, not some separate system. As such, everything in the system is anchored on wins. We don't weight kills towards your MMR, we predict both kills and wins from the same rating, holding wins as the absolute gate (no team can go up in rating if they lose). Because it's super accurate, it doesn't have issues with incentives (see also my GDC talk). You have to win still. Kills, deaths, etc., are only considered inasmuch as they inform the win.
I found this comment particularly interesting.

I was under the impression that the distribution of MMR gain or loss per team could vary. Thus an individual's TS2 Match-Make Rating (MMR) could potentially go up based on their performance even if their team loss, but it may require some unique circumstances to occur such as the system predicting the team that player was on as underdogs (aka expected to lose) while the player's individual performance (Kills, Deaths, & K/D ratio) managed to greatly exceed the system's expectations for them. Now your comment made mention that, "no TEAM can go up in rating if they lose" which can still remain true because other players on the team could simply take a greater negative impact to their MMRs in order to offset any positive impact granted to the MMR of the player who performed exceedingly well. Therefore, while you may not over-weigh kills, deaths, etc towards a player's MMR because it still holds that vital connection to team win or loss it still plays a fairly influential role toward individual adjustments; meaning, that player performance (out-performing or under-performing) in relation to a win or loss can modify the MMR impact at the individual level. In other words, if a team loses the team won't see a net MMR positive gain, but it is entirely possible that an individual or individuals could potentially see a positive gain to their MMR based on their performance, right? Also, on a winning team each individual isn't necessarily awarded the same amount of positive gain toward their MMR as the amount can be influenced by how each individual performed in accordance with their expectations, right?
Yes. It's like there's a pot of points to go around. If two teams also perform almost exactly the same (like 50-49 slayer, etc.), it's possible for the overall change between teams to be very small.
Doesn’t that make it similar to Reach, in that you could rank up more if you individually perform better? You said you don’t weight kills towards your MMR, but then from the next post it sounds like performance does factor. So i have a better chance of individually ranking up my MMR by taking rockets and getting more kills, than I do by sitting back and holding a spawn point while my teammate takes rockets (assuming we win either way), right? Am I misunderstanding? The example above even goes as far to say if you play very well, individually you could go up despite a loss (because your team as a whole goes down). To me that sounds like it opens the can of worms that the reach system did (I understand this is much more accurate than the reach system, but once players realize that individual performance matters after placement matches, it could lead to more solo performances instead of team play.. or maybe h5 has always been this way and I never realized..). Just my two cents.
ZaedynFel wrote:
KR3MIS wrote:
So is their some sort of weird crappy to be in middle ground with the new ranking system. I’m sure population has to come into play but when I’m matching solo I match with players around my skill level which isn’t a problem regardless of our difference in rank, the problem I face is the up hill climb the teammates I’m matched with face. Their obviously out matched in just about every game and game type most of the time dropping kills in the single digits and making what could be a good evenly matched experience into a constant struggle and loss of rank. After a while I’ll get matched with people so far below me that my teammates no longer can help the other team win. So that’s the spot I’m in.
Do you have a playlist or links?

Lower pop lists will often result in wide gaps between the best and worse player in the match. But, when this happens, it means the other team also has an equivalent amount of skill.
I’ve had it happen in just about every playlist at least once but due to this being a problem I have I feel like if I’m going to have to deal with unhelpful teammates then playing swat gives me the best chance to carry so that’s where I’ve had the most of this issue happen when I’m soloing

https://www.halowaypoint.com/en-us/games/halo-5-guardians/xbox-one/mode/arena/matches/f2d8099c-6ae2-4ae8-8b2d-ec1574242e99/players/kr3mis?gameHistoryMatchIndex=41&gameHistoryGameModeFilter=Arena

https://www.halowaypoint.com/en-us/games/halo-5-guardians/xbox-one/mode/arena/matches/b448f13b-1ef4-4247-8295-b189c2e4c3f8/playgameHistoryMatchIndex=49&gameHiers/kr3mis?storyGameModeFilter=Arena

https://www.halowaypoint.com/en-us/games/halo-5-guardians/xbox-one/mode/arena/matches/f611f2fd-fdc8-4149-99c1-d9b1a4ec2b28/players/kr3mis?gameHistoryMatchIndex=82&gameHistoryGameModeFilter=Arena
what is my MMR in SWAT?
bearhound wrote:
Doesn’t that make it similar to Reach, in that you could rank up more if you individually perform better? You said you don’t weight kills towards your MMR, but then from the next post it sounds like performance does factor. So i have a better chance of individually ranking up my MMR by taking rockets and getting more kills, than I do by sitting back and holding a spawn point while my teammate takes rockets (assuming we win either way), right? Am I misunderstanding? The example above even goes as far to say if you play very well, individually you could go up despite a loss (because your team as a whole goes down). To me that sounds like it opens the can of worms that the reach system did (I understand this is much more accurate than the reach system, but once players realize that individual performance matters after placement matches, it could lead to more solo performances instead of team play.. or maybe h5 has always been this way and I never realized..). Just my two cents.
The accuracy makes a huge difference though, and is why in Gears 4, this hasn't caused any issues.

The thing is, if you are playing with near equally-skilled players, grabbing rockets isn't going to get you enough more kills to make a difference.

If you are a better player, then maybe you could, but you're better anyways, and that's not changing the update.

If you're a worse player, you're unlikely to make an impact doing that.

So as long as the balance between wins and the rest is right, we're good.

Not to mention the fact that too much selfish play leads to losses, and leads to the system adapting to downweight that behavior anyways. So either way, still gotta win.
  1. 1
  2. ...
  3. 8
  4. 9
  5. 10
  6. 11
  7. 12
  8. ...
  9. 40