View Full Version : Matchmaking findings...

03-13-2013, 01:44 PM
I'm simply copy/pasting this from an email John sent out yesterday. He looked into this based on feedback from the community.

I've just mined the data since we put in the change (and the rank 2s and 3s). The results are that the mis-matches are VERY infrequent:

power delta battle count percent
0 86259 81.28
1 14484 13.65
2 3919 3.69
3 1040 0.98
4 325 0.31
5 139 0.13
6 41 0.04

In other words, power delta 0 and 1 matches make up for 95% of all matches. Only 41 delta-6 matches have occurred in the last 5 days (out of 106,126 total matches.

Internally, the matchmaker computes a 'disparity' rating that is a combination of Power, Elo, and Timer. Here is the data representing how well the matchmaker thinks the matches are being made. The 'disparity' number here runs from 0 to 10, with 10 being the worst match possible and 0 being 'perfect'.

battle count percent total disparity
80952 76.28 0
4498 4.24 1
2886 2.72 2
6050 5.7 3
9746 9.18 4
1479 1.39 5
392 0.37 6
110 0.1 7
54 0.05 8
20 0.02 9
8 0.01 10

I guess the point is that for the large majority of players the game should be matching very well indeed. We apologize for the times there simply isn't another closely matched player online.

03-13-2013, 01:53 PM
Take into account that the majority of players has rank <= 6 and elo around 1000 (new players). So people who try out the game currently get mostly good matches - could be worse. :)

03-13-2013, 02:00 PM
power delta battle count

6 41 0.04

those are all my games :D

03-13-2013, 02:52 PM
So does that mean less than 1.5% of games have had a 3 or more rank difference? How have I been unlucky enough to be involved in more than one!? What is the popular time to play on your servers? I have had a lot of mismatched games lately...

[edit: In fact now that I think about it, if the occurrences are so low, why not just remove mismatched games altogether? Or make it a maximum of 1 or 2 power difference...

It seems like you've stated these figures as a way of showing how much of a non-issue it is, but to me all it shows is that the feature is unnecessary :/ ]

03-13-2013, 04:52 PM
It's probably the case that there are very few people on the list of those eating lopsided matchups, and they're all getting all of them while the newer players get their same-elo same-power matches while they grind up their teams.

Slimsy Platypus
03-13-2013, 05:23 PM
Thanks for posting this. Personally, this is quite a relief for me to see that the majority of players have been getting excellent matches!

Although, I must say that from my experience I think that more than 4% of my matches have had power differences greater than 2 (and most of them being my opponnent with upgraded units while I was using a standard TP:6 group). Of those 41 power 6 disparities, two of those were me while using a TP:6 group (which is 5% of those occurances). Perhaps I payed very close attention to my mismatches and have disregarded the good matches, just because I got it in my head that it wasn't working perfectly. Perhaps I was just a victim of good ol' natural stastical variance.

If I had to guess I would think that of my matches 1/6 to 1/7 were matches with disparity greater than 2 (about 15%). I'm wondering now if waiting for a longer period of time while the hard Elo window opens up somehow increases the probability that you will get "mismtached". I certainly don't know all the inner workings of the matchmaking, so I wont try to speculate too much.

Regardless, if its getting fixed it doesn't matter! Thanks Stoic!

03-13-2013, 06:15 PM
I think power mismatch may be occuring more for high Elo players than for low or medium Elo players

03-13-2013, 06:37 PM
Well, I for one haven't played a match since the day of the last patch. Personally I don't *ever* want to play a rated match that is unbalance by *even 1* power.

Playing a rated game with even the possibility of unbalance is enough to make me not want to play the game at all. And I logged alot of time the first week and spent money on the perma boost because I liked the game so much.

Some people may enjoy unbalanced games and I think the *option* for that would be great. That said, I want the option of waiting for a fair fight.

Slimsy Platypus
03-13-2013, 07:36 PM
Just as an example. Tonight I have played two matches. Both had a TP differential of 2. But only 4% of the matches occuring when those numbers were run had this much disparity. Maybe I am just unlucky.

I saw Arnie say in the chat box earlier "We're subdividing the versus banner into "quick match" which finds you the closest match and you play for renown and fun, or "Ranked Match" which will find an exact Power match." which will certainly solve the issue. But if the current MM will be used for "quick match" you may want to investigate if there is a direct relationship between Elo and your disparity rating (as I know Tirean has voiced a similar concern with some pretty wonky matchmaking).

Also, for the record let me just state that I'm not unhappy with the game at all. Actually, I'm still absolutely having a blast! I'm just trying to make this issue transparant. I sort of feel like an old crazy man on a soap box, because my experience does not seem like it is one your data supports. Perhaps it is all in my head.

03-13-2013, 07:37 PM
Having been in 5 or so matches of differential 3, Im guessing high elo is largely responsible. And Im also taking into account that that means its also largely irrelevant, in regards to the overall populace. Thanks for the heads up.

03-13-2013, 08:26 PM
You've got an interesting result with that

9746 9.18 4
disparity rating. Wonder why it jumps there?

03-13-2013, 10:20 PM
Just like Slimsy, relieved that this is working fine for most players.

My experience is consistent with high elo players getting the short end of the stick. The issue being probably accentuated by underpopulated servers. I had two 6-power delta matches *in a row* a couple of days ago, and most of my other matches were also (at least slightly) mismatched. The most disturbing thing is that I found myself multiple times in the position of having both the higher elo and the higher powered team, which is no fun for anyone involved. Needless to say I'm playing significantly less in these conditions.
Me being in an "exotic" time zone (GMT+9, Japan) is probably a factor, as the servers are likely to be hitting their daily low in player count when I am online.

03-14-2013, 12:59 AM
These are interesting findings. Like Yuli I have stopped playing matches since the day the patch came out. I am not even interested in the possibility that I will be matched against someone much stronger or weaker than me. I really hope that a play same rank button or similar is implemented soon so I can come back to playing. I really enjoy the game though so continue the good work.

03-15-2013, 10:16 AM
Okay. Well, that's fine for matchmaking, but I don't want my uneven matches ranked. Go ahead and punish those who use (currently) unusual team powers > 6 by making them wait two hours to play each other.