Why do I see a large skill gap in my match?
First, matchmaking is based on MMR not CSR. If you are seeing large gaps, it means someone is changing in MMR faster than their CSR can catch them. Going by MMR, matches are always tight, so they should be reasonable in how they play out, regardless of what CSRs you see. The exception to this is Champion-level players that are so good we have to match them vs. high-Onyx players who they usually destroy. Fortunately this is rare. The other exception is when higher-skill players create a fireteam with much lower-skilled ones, but I’ve addressed this elsewhere.
Number of Placement Matches
Some of you have asked about reducing the number of placement matches. If I can confirm the skill system is confident enough in skills early on, we could change this. It's something I'm investigating.
If you mess with Elo’s math to force Onyx to be smaller it’ll mess everything up
I don’t need to touch the math, just the boundaries on the raw numbers that are used to decide your ranks. For example, right now, 1500 is the boundary for Onyx. Maybe that needs to be 1700. But your raw CSR and the math behind it would stay the same.
As a side note, I reserve the right to mess with Elo’s math as much as I want since I've published academic papers in that type of math(s).
Demotions
I’ve read some concerns about getting a promotion and then an immediate demotion in the next match. The reverse also happens. Right now, we both promote and demote immediately when a player’s CSR crosses the boundary for the higher / lower rank, and we don’t do anything to slow that down. We could consider adding a buffer if this is too jarring, though it’s not a high priority. One way to do this would be to add or subtract a few extra games worth of CSR when you cross a boundary. For example, when you get promoted to Gold, we add an extra 15-30 CSR so you would have to lose at least 2 or 3 in a row to get demoted. If we did that, we would also need to subtract an extra 15-30 CSR when you get demoted to remain balanced. It does have the drawback of slightly over- / under-estimating the CSR, though not enough to matter much in my opinion.
Placement
Some of you have expressed the concern that you always start behind where you ended last season. We intentionally place you a bit back from where your MMR would place you. If you ended last season after just getting into the next higher rank, you will always fall back to the previous one and have to work your way up again. Two reasons we do that:
We may be improving the placement accuracy in the coming months, but not in the short term.
Keep in mind when you have an unranked player in a match, they may have been ranked previously, so don’t assume they’re terrible. The system tries to place unranked players into the right place faster than we give them a CSR.
Shouldn’t FFA give a win for 1-3rd place?
Yes, I agree with this. I’m not sure I can do it soon, but I’ll check. Note: CSR gains are usually positive for 1-3, negative for 4-6.
Not Using the Average Skill in Party
I’ve heard people suggest that we should use the max skill of a party instead of its average in order to discourage partying up with smurfs. I’m not opposed to that and may try it in the future, but:
Team Rankings vs. Personal
Yes, I like the idea of having a Personal Ranking that moves based on individual in-game performance, and one that is based on winning. I would want these both separate and want to make it clear that the winning version is more important.
Keep in mind we don’t use Arpad Elo’s method for Ranks. We use something more advanced. We can integrate Kills, Deaths, Assists, Headshots, Objectives, etc., into your skills. We can also create a win-based one CSR separate from your individual performance-based one. I’m constantly evaluating which would be best for each part of the game.
Champion Ranks Rewards Follow-up
A few of you have suggested we have different rewards for each rank each season. While I think that’s super cool, let me be more clear why I gave the design I proposed.
At this stage in Halo 5’s life cycle, I only have so many rewards left I can give. If I had unlimited reward content, I could have something for every rank for every season, but I don’t.
Instead, I have a limited set of cool rewards and a goal to make Champion feel exclusive. This means some of you will have a harder time getting these rewards at first, and I know that’s sad. But my focus right now is how to first make it worth the fight for the Champions for a few seasons, and secondarily to have something else for the lower Ranks.
One compromise is to let Champions get the current season’s reward, Onyx the previous season’s reward, Diamond the reward from 2 seasons ago, and so on down the chain. So everyone can eventually get them all, but better players get them sooner.
Aren’t people quitting because they don’t like the Map and / or Mode?
For most playlists, there’s no pattern on which maps and / or modes people quit the most. It’s equal. This is the case in team arena. This means removing a map-mode combination wouldn’t make a difference in quit rates. Neither would vetoes because players are just as likely to quit the alternative as they are the vetoed map.
Do you only talk about arena? What about Warzone / Warzone Assault
No, not just arena. I will address Warzone questions. I haven’t seen many yet though, and sorry if I’ve missed them. I did address Warlords recently.
( 1 / 2 )
First, matchmaking is based on MMR not CSR. If you are seeing large gaps, it means someone is changing in MMR faster than their CSR can catch them. Going by MMR, matches are always tight, so they should be reasonable in how they play out, regardless of what CSRs you see. The exception to this is Champion-level players that are so good we have to match them vs. high-Onyx players who they usually destroy. Fortunately this is rare. The other exception is when higher-skill players create a fireteam with much lower-skilled ones, but I’ve addressed this elsewhere.
Number of Placement Matches
Some of you have asked about reducing the number of placement matches. If I can confirm the skill system is confident enough in skills early on, we could change this. It's something I'm investigating.
If you mess with Elo’s math to force Onyx to be smaller it’ll mess everything up
I don’t need to touch the math, just the boundaries on the raw numbers that are used to decide your ranks. For example, right now, 1500 is the boundary for Onyx. Maybe that needs to be 1700. But your raw CSR and the math behind it would stay the same.
As a side note, I reserve the right to mess with Elo’s math as much as I want since I've published academic papers in that type of math(s).
Demotions
I’ve read some concerns about getting a promotion and then an immediate demotion in the next match. The reverse also happens. Right now, we both promote and demote immediately when a player’s CSR crosses the boundary for the higher / lower rank, and we don’t do anything to slow that down. We could consider adding a buffer if this is too jarring, though it’s not a high priority. One way to do this would be to add or subtract a few extra games worth of CSR when you cross a boundary. For example, when you get promoted to Gold, we add an extra 15-30 CSR so you would have to lose at least 2 or 3 in a row to get demoted. If we did that, we would also need to subtract an extra 15-30 CSR when you get demoted to remain balanced. It does have the drawback of slightly over- / under-estimating the CSR, though not enough to matter much in my opinion.
Placement
Some of you have expressed the concern that you always start behind where you ended last season. We intentionally place you a bit back from where your MMR would place you. If you ended last season after just getting into the next higher rank, you will always fall back to the previous one and have to work your way up again. Two reasons we do that:
- To require some time investment this season to earn your rank back
- To prove to the system you deserve that rank
We may be improving the placement accuracy in the coming months, but not in the short term.
Keep in mind when you have an unranked player in a match, they may have been ranked previously, so don’t assume they’re terrible. The system tries to place unranked players into the right place faster than we give them a CSR.
Shouldn’t FFA give a win for 1-3rd place?
Yes, I agree with this. I’m not sure I can do it soon, but I’ll check. Note: CSR gains are usually positive for 1-3, negative for 4-6.
Not Using the Average Skill in Party
I’ve heard people suggest that we should use the max skill of a party instead of its average in order to discourage partying up with smurfs. I’m not opposed to that and may try it in the future, but:
- It discourages playing with legit worse friends. It will make it painful for them, frustrating for the good player, and drags the good player CSR down in cases where your friends are legit bad. We’re careful about putting up barriers to playing with friends since it’s so fun. But also understand that maybe ranked isn’t the place for that if your friends are bad.
- I want to first to see how well the skill improvements alone can mitigate this problem.
- If still needed I may do it for only Team Arena
- A softer option is to force your CSR to move towards your fireteam’s average after every match rather than your personal MMR. For example, in the case of an Onyx player playing with Bronze, both will see their CSR drift towards Gold over time. This will have the result of discouraging smurfing since it will drag the better player’s CSR down. In cases where your friends really are bad, it also will still give the more balanced matches for everyone that we have today. It does have the downside that when you play with bad friends, your CSR goes down. But so does matchmaking on the max. I think this option is better overall.
Team Rankings vs. Personal
Yes, I like the idea of having a Personal Ranking that moves based on individual in-game performance, and one that is based on winning. I would want these both separate and want to make it clear that the winning version is more important.
Keep in mind we don’t use Arpad Elo’s method for Ranks. We use something more advanced. We can integrate Kills, Deaths, Assists, Headshots, Objectives, etc., into your skills. We can also create a win-based one CSR separate from your individual performance-based one. I’m constantly evaluating which would be best for each part of the game.
Champion Ranks Rewards Follow-up
A few of you have suggested we have different rewards for each rank each season. While I think that’s super cool, let me be more clear why I gave the design I proposed.
At this stage in Halo 5’s life cycle, I only have so many rewards left I can give. If I had unlimited reward content, I could have something for every rank for every season, but I don’t.
Instead, I have a limited set of cool rewards and a goal to make Champion feel exclusive. This means some of you will have a harder time getting these rewards at first, and I know that’s sad. But my focus right now is how to first make it worth the fight for the Champions for a few seasons, and secondarily to have something else for the lower Ranks.
One compromise is to let Champions get the current season’s reward, Onyx the previous season’s reward, Diamond the reward from 2 seasons ago, and so on down the chain. So everyone can eventually get them all, but better players get them sooner.
Aren’t people quitting because they don’t like the Map and / or Mode?
For most playlists, there’s no pattern on which maps and / or modes people quit the most. It’s equal. This is the case in team arena. This means removing a map-mode combination wouldn’t make a difference in quit rates. Neither would vetoes because players are just as likely to quit the alternative as they are the vetoed map.
Do you only talk about arena? What about Warzone / Warzone Assault
No, not just arena. I will address Warzone questions. I haven’t seen many yet though, and sorry if I’ve missed them. I did address Warlords recently.
( 1 / 2 )
