Matchmaker 0_0
-
Maybe the matchmaker could use a little improvement on the balance side of things?
not to mention the -322 guy was afk this was a sad game. -
Can only do so much with a lack of people in queue.
-
3039 vs 978 is not a match in any humans mind, I would rather not have a game at all than have games like this.
Does the matchmaker not pair up equal opponents? at least on rating.
-
It should as far as I know, but perhaps the other team were queued together so MM could not split them.
-
there should be a minimum quality needed to match a game
-
There is. I think what happened here is that one team ended up with a lot of players that the matchmaking system is highly uncertain of. Players start with 1500 rating and 1500 uncertainty, that bottles down to 0 on the in-game scoreboard. Meaning the teams weren't that off for the system, it just wasn't sure of some players yet.
-
The rating you see is the "minimum" rating, the system believes you should have, which is your rating minus some uncertainty.
All players on your team had less than 10 games, so the uncertainty was very high. You should basically not try to judge balance by the shown numbers for players with very few games.
The balance (as you can see in the replay) is 66%, so it's probably still on the lower end. -
There were five people with less than 10 games in that game. The matchmaker tries to find games faster for these people. So it is more lenient with match quality the more new people are in the game. On top of that there must have been some premade teams here, otherwise the matchmaker would have distributed the players into the teams differently.
I checked the numbers and this had to be on average the fourth matching attempt to accumulate enough time bonus to offset the bad match quality.
A game needs at least 0.4 quality to be considered. This is an internal quality metric, not the 66% you see in the replay. This match had a quality of about -2.5. Normally the time bonus is capped at 0.4, but is increased if new people are in the match. With five new people the average wait time had to be at least four attempts to accumulate more than 2.9 time bonus.
If there was a better alternative in the three attempts before it would have chosen one of those.I can understand that this was a frustrating experience but the chances of it happening were very low.
-
I think this is a good example of why we might want to display leagues instead of ratings in TMM. I think if it had shown "In placement" for the new players and the leagues for the rest, it would be less confusing for players to understand why the matchmaker put this game together and more accurately communicate what to expect.
-
Perhaps if the players are new and within their very first few games the system should just assume they perform more towards the lower ability and match them appropraitly with other low ranks.
and if a new player happens to be quite good they should rank up quickly.
As these players only had 1 or 2 games im shure this was not a very pleasant introduction to FAF and may have been put of by this experiance.
Also might it be reasonable to start players of who are new to matchmaking with their global score to begin with? if they have a global score.
-
@caliber it does.
Even though their rating is 1500 they are matched as if their rating was 500 for their first 10 games. (bit more complicated but roughly this) -
so the older tmm gets the more players play the better get the balance?
basic stochastics?=P -
@caliber said in Matchmaker 0_0:
Also might it be reasonable to start players of who are new to matchmaking with their global score to begin with? if they have a global score.
SOmething like this is acutally happening. The 0 rating for the first game is just a visual bug from the client. If you look up the replay you will see that the 0 rated guy actually had a higher rating.
-
B BlackYps referenced this topic on
