[Locked] Matchmaking Feedback Update – April 16

DaxSeven09 wrote:

Co2 OUTBR3AK wrote:

D M4N8 wrote:

Wait- so this system no longer matches you to where there will be very close games?

It will be more of a 50/50 based system instead because that's what it sounds like...

Or was that just an example?

50/50 means matches will be tighter. You have that chance of winning, so less chance for consistent blowouts.

Unless I'm matching up against you! Ugh, it seems like every night I meet up against you and never with you. :)

Anyways, it seemed like the few games I played last night that I was matched with more players closer to my caliber.

Lol I understand completely. I am a fairly high MMR player, but I run solo mostly in BTB and never full team squad. So I may play ok teams that I could easily slay through, but to even it out I in turn get random on my team that barely know how to walk in 3d or even pick up weapons.

It'san average, basically. The system creates better teams so each team has an even chance (50/50) to win based on MMR and win/loss, kd, etc.
I could have a skill of 90/100, while my teammates help average our team skill down to say 60. But the opposing team could have multiple people with a skill of 75 and randoms that get their team skill to 60 as well. It's about numbers among factors. ;)

And you may see me tonight again too ;)

radar3301enigma wrote:

ZaedynFel wrote:

radar3301enigma wrote:

@ZaedynFel: I know you're planning on rolling this out to Ranked at some point. Do you have a rough timeline when that will be? In the next week or two, or when the season resets in a couple months? Or sooner than that?

We're aiming for season roll. So if everything goes well it shouldn't be too far off.

Excellent, thanks.

What kind of differences in ranks will we see? For example, if my skill is really a 0.8, but my displayed skill is a 1.3, will that be reflected?

Well, we aren't making major changes to the Ranking system itself, just the skill that drives part of it.

So it's still:

1. Play 10 matches
2. Convert MMR into CSR
3. Subtract off 100 CSR to make sure you earn the CSR
4. Future matches compare your CSR to your opponents' MMR

The difference it that the MMR above comes from TrueSkill2 instead.

LUKEPOWA wrote:

Here are the three games I played last night. Just curious if these games were predicted correctly.

Link. Lost 3-0. Skill seemed even.
Link. Lost. Party of 4 on other team.
Link. Tie game.

There's still a lot of times that games won't be even. Plus even with matched skills, 1 or 2 people that are good can still overrun games and create extreme blowouts. Take these 2 matches I played last night. I steamrolled:

Won 2-1; K/D 70-8Won 2-1; K/D 59-9Won 2-0: K/D 96-3 (ouch)I get hate messages over it somewhat often and reported for cheating. Buuuuut it's btb and I never play with full teams. Maybe 3 or 4 people, so randoms still don't or can't grab flags or arm bombs even when I'm literally slaying over and over. While my win/loss has inevitably suffered from this, the end result isn't so much doing well in matches compared to the logical outcome of "winning is more important."

I still feel that better players and teams will still steamroll and win more often, but the number of matches where I absolutely destroy people hasn't really changed much at all unless I run into teams of 8 while playing solo or 3s and 4s. (It happens a bit too.)

LUKEPOWA wrote:

Here are the three games I played last night. Just curious if these games were predicted correctly.

Link. Lost 3-0. Skill seemed even.
Link. Lost. Party of 4 on other team.
Link. Tie game.

First Match: Winning team had a 60% chance to win
Second Match: Winning team had a 46% chance to win (slight upset)
Third Match: Better team had a 52% chance of winning, 5% chance to tie.

eLantern wrote:

Pablo Says Hi wrote:

hi,

will trueskill2 be applied to ranked matches?

I wrote a couple weeks ago about being constantly matched in unfair ranked games and not much has changed. I took your advice in changing my search preferences to focused but it feels the same as balanced.

my last match:

https://www.halowaypoint.com/en-ca/games/halo-5-guardians/xbox-one/mode/arena/matches/e5f86be8-727e-4c76-9808-53fdf1721bd0/players/pablo%20says%20hi?gameHistoryMatchIndex=0&gameHistoryGameModeFilter=All

red team had two champions and tow onyx players while we (blue team) had a platinum, two low level diamonds and an unranked.

could you please explain to me why the current system allows games like these to be played? Why doesnt the current system recognize this as unfair?

thanks,

Pablo

May I ask what region you play in?

east coast. does that matter?

ZaedynFel wrote:

Co2 OUTBR3AK wrote:

LUKEPOWA wrote:

Here are the three games I played last night. Just curious if these games were predicted correctly.

Link. Lost 3-0. Skill seemed even.
Link. Lost. Party of 4 on other team.
Link. Tie game.

There's still a lot of times that games won't be even. Plus even with matched skills, 1 or 2 people that are good can still overrun games and create extreme blowouts. Take these 2 matches I played last night. I steamrolled:

Won 2-1; K/D 70-8Won 2-1; K/D 59-9Won 2-0: K/D 96-3 (ouch)I get hate messages over it somewhat often and reported for cheating. Buuuuut it's btb and I never play with full teams. Maybe 3 or 4 people, so randoms still don't or can't grab flags or arm bombs even when I'm literally slaying over and over. While my win/loss has inevitably suffered from this, the end result isn't so much doing well in matches compared to the logical outcome of "winning is more important."

I still feel that better players and teams will still steamroll and win more often, but the number of matches where I absolutely destroy people hasn't really changed much at all unless I run into teams of 8 while playing solo or 3s and 4s. (It happens a bit too.)

Your first match: 98% chance of winning (because of quits). TrueSkill2 predicted 48-8 and you went 70-8 instead. It was spot on about your deaths.
Second Match: Bad link
Third match: 90% chance of winning. Predicted 50-7 instead of 96-3.

So you played even better than expected (you're an outlier on the high end), it did expect really good performances out of you already, and it knew who would win.

This means the reason for the imbalance is not skill system, but instead either the matchmaker, or quitters.

I agree. It's a hugely prominent problem in BTB. I'm honestly hugely surprised there's no mercy rules in place for these types of blowouts because I've been on that receiving end and it sucks. Hell, it took me 10 minutes in Forge to script a mercy rule for slayer that ends the match if one team is beating another by margin of greater than or equal to 40 points. Quitters will always be a problem .

However, on the topic of TrueSkill2, hopefully matches become better balanced. Quitting is usually due to I'm balance and quitting only aggravates that fact even more. So more quit. We should hopefully see a reduction in quits though,I'm assuming.

ZaedynFel wrote:

UPDATE: Trueskill2 is now live in all Social and Warzone hoppers.

12 mans have to search for 15+ minutes in both Assault and regular Warzone before finding a match. Before we could find matches quickly very consistently even midday weekdays, so this is extremely disappointing for us. I understand that making matches as even as possible is very important to you, but I would rather join a uneven match then sit in lobbies for 20 minutes between games.

eLantern wrote:

ZaedynFel wrote:

Your first match: 98% chance of winning (because of quits). TrueSkill2 predicted 48-8 and you went 70-8 instead. It was spot on about your deaths.
Second Match: Bad link
Third match: 90% chance of winning. Predicted 50-7 instead of 96-3.

So you played even better than expected (you're an outlier on the high end), it did expect really good performances out of you already, and it knew who would win.

This means the reason for the imbalance is not skill system, but instead either the matchmaker, or quitters.

I understand that quitters will significantly alter the win probability, but what do you mean when you say that it could be the "matchmaker"?

Is this in reference to what imbalance the matchmaker will allow? In other words, the matchmaker could not have constructed a better match-up with a more competitive win probability because it was also giving some priority to player connections (data centers) and where some players were in their search times because it tries to ensure that all players (if solo) will eventually find a match in social?

Yeah, it means the matchmaker may have known this match was bad, but let it happen anyways. Usually this happens because of population.

Basically, we try and tweak the matchmaking rules so they assume we're at "Peak" population at first, and then they expand eventually down to "trough" so everyone should be able to match.

It's not perfect because even the peaks and troughs change over time, and you can end up expanding more than you would have wanted.

I have been really struggling to find matches in super fiesta and the few games I have gotten have had a significant increase in player lag rendering the games almost unplayable. . Does trueskill 2 prioritize skill ranking over connection quality while searching for a match?

GeoNeo2582 wrote:

I have been really struggling to find matches in super fiesta and the few games I have gotten have had a significant increase in player lag rendering the games almost unplayable. . Does trueskill 2 prioritize skill ranking over connection quality while searching for a match?

Search Focused to prioritize connection more. Just realize it'll come at a cost of likely lengthier or more times out searches.