Skip to main content

Forums / Community / Matchmaking Feedback & Discussion

Matchmaking Update 3/17/2020

OP ZaedynFel

  1. 1
  2. ...
  3. 22
  4. 23
  5. 24
  6. 25
  7. 26
  8. ...
  9. 37
Some explanation of why only 1 csr goes up
https://www.halowaypoint.com/en-us/games/halo-5-guardians/xbox-one/mode/arena/matches/56d4d623-2ab0-400b-a80f-b06f610bc8cd/players/noahchief05?gameHistoryMatchIndex=16&gameHistoryGameModeFilter=Arena
CSR update is delayed by 1 game and is based on MMR, Check out the FAQ on this forum: FAQ

But that being said, looks like your team is significantly higher (visible) rank than the other team.
Cody15x wrote:
Hey I’m just curious about this game. Is the enemy team a to4 or was the matchmaker just doing its best with low population?
Even though this match appears “fair,” i.e. avg of plat v plat, having a good player and a bad player just seems to be better than 2 avg players.
Given it was 50-46 despite you losing a teammate, I would say it was fine.
This post has been edited by a moderator. Please do not post spam.
*Original post. Click at your own discretion.
Spoiler:
Show
Hi Josh,
Just wondering why there are so many smurf accounts in H5? Also most of them are streamers/you tubers and some are quite famous .They even laugh about it on their streams and play in tournaments. How is this allowed?
As a solid silver player I wish I had onyx problems! LOL
I just want to know why I get placed with champion players in swat when I haven't even completed my placing yet! Does it use previous ranking to determine your player pool?
Placing is CSR
Matching uses MMR, which is a history of everything you've done in that kind of playlist before.
The answer is probably no, because Halo 5 is a very old game and Infinite clearly needs all the resources it can muster, but is there ever thought of looking back at TS2 to see if the model performs similarly on more recent data? The paper was published in early 2018, presumably based on older historical data, and I'm curious if those assumptions hold for a much smaller, much different-looking player base. That guy earlier in the thread talking about assists got me thinking about that. Despite making sense, assists didn't do a good job of predicting wins at a population level when we had a giant population. When our population's much smaller, and probably more skilled since there are much more vets than new players (I assume), I'd be curious if things like assists which can be a low-skill indicator when you're new (e.g. you can't finish a kill) but high-skill indicator as a veteran (you play well with a team, don't rush to trade when you have help, etc).
The answer is probably no, because Halo 5 is a very old game and Infinite clearly needs all the resources it can muster, but is there ever thought of looking back at TS2 to see if the model performs similarly on more recent data? The paper was published in early 2018, presumably based on older historical data, and I'm curious if those assumptions hold for a much smaller, much different-looking player base. That guy earlier in the thread talking about assists got me thinking about that. Despite making sense, assists didn't do a good job of predicting wins at a population level when we had a giant population. When our population's much smaller, and probably more skilled since there are much more vets than new players (I assume), I'd be curious if things like assists which can be a low-skill indicator when you're new (e.g. you can't finish a kill) but high-skill indicator as a veteran (you play well with a team, don't rush to trade when you have help, etc).
If I had to guess, 343 is constantly checking the data and making minor tweaks here and there. I think the paper was just to explain the system when the were rolling it out and since there have not been any major changes no need to update the paper. Why continue to prove what the data and model have sufficiently shown?

On a separate note when does the current season end? Trying to get to diamond in all the playlists I have ranked in so far and want to make sure I have time.
https://www.halowaypoint.com/en-us/games/halo-5-guardians/xbox-one/mode/arena/matches/d8ef67ad-a4bd-414c-a897-67d15d3acdd9/players/noahchief05?gameHistoryMatchIndex=5&gameHistoryGameModeFilter=Arena
What is that?
https://www.halowaypoint.com/en-us/games/halo-5-guardians/xbox-one/mode/arena/matches/d8ef67ad-a4bd-414c-a897-67d15d3acdd9/players/noahchief05?gameHistoryMatchIndex=5&gameHistoryGameModeFilter=Arena
What is that?
Looks like a close match. Your opponents had two high ranked players and two low ranked whereas you had 4 mid ranked players. Kills were almost even and the game could have gone either way, which is exactly what the system is aiming for.
Idk that day I lost against lower ranks and I don't go down much like that game of two champs
r u ever going to implement the d6 to champ system back. personally i like the champ threshold being at 1600 hitting multiple champs this season was fun.
@ZaedynFel could you tell me what my current arena matchmaking MMR is please? I've never seen it before and I'm real interested to have a look.
The answer is probably no, because Halo 5 is a very old game and Infinite clearly needs all the resources it can muster, but is there ever thought of looking back at TS2 to see if the model performs similarly on more recent data? The paper was published in early 2018, presumably based on older historical data, and I'm curious if those assumptions hold for a much smaller, much different-looking player base. That guy earlier in the thread talking about assists got me thinking about that. Despite making sense, assists didn't do a good job of predicting wins at a population level when we had a giant population. When our population's much smaller, and probably more skilled since there are much more vets than new players (I assume), I'd be curious if things like assists which can be a low-skill indicator when you're new (e.g. you can't finish a kill) but high-skill indicator as a veteran (you play well with a team, don't rush to trade when you have help, etc).
If I had to guess, 343 is constantly checking the data and making minor tweaks here and there. I think the paper was just to explain the system when the were rolling it out and since there have not been any major changes no need to update the paper. Why continue to prove what the data and model have sufficiently shown?

On a separate note when does the current season end? Trying to get to diamond in all the playlists I have ranked in so far and want to make sure I have time.
Seasons are basically Calendar seasons. So June/July/August is Summer. Sept/Oct/Nov Fall, Dec/Jan/Feb Winter, etc.

And, yes, we are constantly testing the importance of all of the variables (including Assists) because information like that is helpful not only to H5, but informs how we prioritize stuff in Infinite.

So far, nothing has changed. Pop also isn't *that* much lower since 2018.
I just want to know why I get placed with champion players in swat when I haven't even completed my placing yet! Does it use previous ranking to determine your player pool?
Yep. Like AshamanND said, we never reset your MMR, so we know exactly how good you are day 1 of the next season. It's up to you to prove if you're still as good as you were.

Though it only takes 1-2 games to know to match you with Champs even on a reset.

This post has been hidden.

3
"
  • Finding a group of players that the team balancer (a separate system) will be able to create equally skilled teams with (50/50 win chance).
This sounds good on paper but in actual practice you're pairing skilled players with teammates who factually aren't hitting their shots. This forces the skilled players into a constant cycle of fighting with no team shot and being at a (frustrating) handicap. Your MMR doesn't take into account player accuracy and map knowledge which are the MOST important factors for determining skill.

This is EVERY game of solo searching 4s as a skilled player. We are forced against our will to HAVE to strap on a backpack even in SOCIAL gametypes.

Add accuracy into your equation or remove it altogether as it factually does not balance games. There's no need to redesign something that wasn't broken.
"
  • Finding a group of players that the team balancer (a separate system) will be able to create equally skilled teams with (50/50 win chance).
This sounds good on paper but in actual practice you're pairing skilled players with teammates who factually aren't hitting their shots. This forces the skilled players into a constant cycle of fighting with no team shot and being at a (frustrating) handicap. Your MMR doesn't take into account player accuracy and map knowledge which are the MOST important factors for determining skill.

This is EVERY game of solo searching 4s as a skilled player. We are forced against our will to HAVE to strap on a backpack even in SOCIAL gametypes.

Add accuracy into your equation or remove it altogether as it factually does not balance games. There's no need to redesign something that wasn't broken.
We tested accuracy and it did not improve results. The current system already predicts accuracy near exactly without even including it in the model. And not just "on paper", when tested on millions and millions of actual matches and real players. It's extremely capable of balancing matches. When the system predicts 50%, the score is extremely close every time.

As for "redesigning something that wasn't broken" we haven't touched the team balancer code since the game launched. That has never changed, the game has always balanced the teams the same exact way. The only difference is the number the team balancer is now being handed is incomparably better at predicting a player's performance than before.

The problem you're complaining about has nothing to do with how well the system predicts player performance, and also has nothing to do with skill-based matchmaking, it has everything to do with population size.

When you play an unpopular playlist, the matchmaker has no choice but to dump whichever 8 players are actually available into the same match. If you are the best player out of the ONLY 8 players available in a low pop list, you will always have terrible teammates.

Same thing goes if there is no skill-based matchmaking. If it is totally random matchmaking, then you get a mix of great and terrible players in the same match, and the classic team-balancer (from every Halo) will put the best with the worst like it's always done since 2004.

Your issue isn't the skill system, and it's not even the matchmaking, it's straight up just low pop. There's no way around that though if you want a match.

And there's no way we are going to put the other good players in a match all on the same team to just destroy bad players in the match. That drives way more players out of the game than you having to backpack.
ZaedynFel wrote:
"
  • Finding a group of players that the team balancer (a separate system) will be able to create equally skilled teams with (50/50 win chance).
This sounds good on paper but in actual practice you're pairing skilled players with teammates who factually aren't hitting their shots. This forces the skilled players into a constant cycle of fighting with no team shot and being at a (frustrating) handicap. Your MMR doesn't take into account player accuracy and map knowledge which are the MOST important factors for determining skill.

This is EVERY game of solo searching 4s as a skilled player. We are forced against our will to HAVE to strap on a backpack even in SOCIAL gametypes.

Add accuracy into your equation or remove it altogether as it factually does not balance games. There's no need to redesign something that wasn't broken.
We tested accuracy and it did not improve results. The current system already predicts accuracy near exactly without even including it in the model. And not just "on paper", when tested on millions and millions of actual matches and real players. It's extremely capable of balancing matches. When the system predicts 50%, the score is extremely close every time.

As for "redesigning something that wasn't broken" we haven't touched the team balancer code since the game launched. That has never changed, the game has always balanced the teams the same exact way. The only difference is the number the team balancer is now being handed is incomparably better at predicting a player's performance than before.

The problem you're complaining about has nothing to do with how well the system predicts player performance, and also has nothing to do with skill-based matchmaking, it has everything to do with population size.

When you play an unpopular playlist, the matchmaker has no choice but to dump whichever 8 players are actually available into the same match. If you are the best player out of the ONLY 8 players available in a low pop list, you will always have terrible teammates.

Same thing goes if there is no skill-based matchmaking. If it is totally random matchmaking, then you get a mix of great and terrible players in the same match, and the classic team-balancer (from every Halo) will put the best with the worst like it's always done since 2004.

Your issue isn't the skill system, and it's not even the matchmaking, it's straight up just low pop. There's no way around that though if you want a match.

And there's no way we are going to put the other good players in a match all on the same team to just destroy bad players in the match. That drives way more players out of the game than you having to backpack.
1: I said accuracy AND map awareness. The ONLY factor that should be taken into account for matching people is their location/factual playlist rank. Hypothetically: 8 players all around level 20 should be mixed around every game instead of FORCING the 3 bad players with the better overall level 20. It's an unfair bias towards higher skilled players, especially when searching solo que. There's two outcomes to this: smurfs and skilled players leaving the game (hence your population problem) Novody wants to play a game where they are intentionally punished for working hard at bettering themselves.

2: Sure the score is close, but nobody is enjoying the game. Good players are FORCED to backpack and bad players continue to go negative due to them being carried and constantly fighting better players. This causes people to create smurfs, which you guys seem to be okay with because it artificially inflates your player count for that sweet investment money.

3: Redesigning something that wasn't broken was in regards to Bungie's MM. I clocked thousands of hours in H2 and H3 online (I was also a H3 Gamebattles Admin where I saw oversaw and managed the most populated Halo RANKED matchmaking system that bypassed *yours/bungie*). Our system of matchmaking was never under dispute for balancing games because there was no secondary bias calculation of "skill" you either won or lost and that's what determined your rank. The social slayer/ranked playlists never had imbalances this large. The ONLY time this "new system that isn't new because it's been the same since 04"(lol wat) should be implemented is in ranked.

4: "it has to do with population size" More people would play if they weren't forced into a handicap, I think that's the biggest struggle your team fails to understand. Nobody wants to backpack 3 little Billy's and Timmy's in a social playlist EVERY game. Regular people bust their -Yoink- at a full time job, and shouldn't have to come home and be sweating every fight in a social gametype with a constant 3v1 scenario because Billy and Timmy JUST got game pass and don't know the map layouts. Your numbers are so low youre willing to do anything to make sure they don't leave, I get it. (Also, intentionally ignoring hit registration/spawning issues for 5 years to artificially lower the skill ceiling and raise the ttk against lower skilled players.) It's either negligence or intent. I HOPE it's just incompetence.

5: Bungie designed the MM system with the intention to emulate real life LAN events. Social games need to be randomly assorted to recreate that LAN feeling of mixing teams up. It doesnt matter if Billy and Timmy need help winning, that's up to THEM to get better to improve their gameplay. I don't owe you guys mentally exhausting labor to improve your playerbase's statistical performance. If it doesn't let up, I'll start searching with myself and 3 other smurf accounts and show you how fast we can ACTUALLY depopulate your playlists, because at this point this is so frustrating that I'm not even enjoying the game.

Your whole teams intention is to make the match close by forcing good players against their will to statistically never have a teamshot in social slayer and it's siphoning the actual enjoyment out of the casual play. That's the factual byproduct of these practices.

Here's what you need to do to be successful, next patch IMMEDIATLY make custom game browsing and cross-platform accessible. Now is NOT the time to be allocating so many resources into focusing on skin development.

Every single stream that go into and every lobby has the SAME concerns as me. Ive seen your staff's Twitter posts about how much you hate good players succeeding when they search 4s. Curious what that's all about.
  1. 1
  2. ...
  3. 22
  4. 23
  5. 24
  6. 25
  7. 26
  8. ...
  9. 37