OR  Zero-K Name:    Password:   
Title: Cyka nuggets alive again
Host: THrankPootis
Game version: Zero-K
Engine version: 104.0.1-287-gf7b0fcc
Started: 3 months ago
Duration: 8 minutes
Players: 6
Bots: False
Mission: False
Watch Replay Now
Manual download

Team 1
Chance of victory: 34.1%

Team 2
Chance of victory: 65.9%


Show winners

Filter:    Player:  

3 months ago
So how did the autobalancer come up with this one? Everyone on team 2 is better (in rating + my opinion) than their team 1 counterpart:

Anir > ehtom
Darkstaron > Pootis
BOOCH >>>>>>>>>>> ellysonjunior01.

Obviously a game like that is not really 40% chance to win but the game thinks it is somehow.
+0 / -0

3 months ago
While I don't know the mathematical ins and outs of how the matchmaker picks sides, it looks like in this case it simply went by casual rankings. Anir and Dark were on the same team because Anir is higher rated casually than you, but Dark is rated lower than Pootis.

Team one has the 2nd and 3rd highest rated players. Team two has the 1st and 4th. I'm disregarding the two newbies since they're both casually unranked. I'd say this is matchmaker working as intended with the resources available. This was a casual game, you had a rough idea of what you were likely to face before you begun.

Were this a matchmaker game, I would have expected you and Dark to be drawn against Anir and Pootis. Pootis is only just rated inside subgiant bracket for matchmaking while Dark is a very proficient matchmaker giant (indeed, he's only just behind you).

Due to the disparity between the amount of time playing matchmaker and casual games (and then only doing lobster pots where chance is a much bigger player than in smaller games), I'm casually unrated, but just inside the top 20% for matchmaking at the moment. Were I in this game rather than Dark, I imagine it'd have put me in Anir's team as well.

Since there are more casually rated players than competitive players, a top 20% casual player is likely to be worse at the game than a top 20% competitive player. I read in the old matchmaker debates that the two ratings were consciously decoupled. If the format of a casual game is close to that of a matchmaker game (as this one seems to be), Dark likely played as his competitive rating suggested he would- extremely well (and much better than Pootis who as I noted earlier, is only subgiant rated in matchmaker). Had it been a lobster pot, he might not have made much of an impact.

Due to the wider range of games possible in custom rooms, I'm inclined to think that rating is much less reliable in predicting relative skill. This makes the custom matchmaker less reliable than the competitive matchmaker since there are fewer formats for the latter to take.
+0 / -0

3 months ago
(edited 3 months ago)

Since there are more casually rated players than competitive players, a top 20% casual player is likely to be worse at the game than a top 20% competitive player.

I suspect this is not the case any more. For example, DEranksnoke is in a similar position on both ladders, and is in the top 5% of casual and the top 8.6% of matchmaker. For various reasons I think it's quite a bit harder for a new player to break into high casual ranking as opposed to MM, so highly ranked casual players will tend to at least have experience if not skill as such.

Given that the algorithm had basically no data to work with about two of the players I would not expect a good result.
+0 / -0

3 months ago
This is a casual game. By popular request, casual games are only balanced by the data given from other casual games. So 1v1s or other MM games don't count here. There's no way to reliably determine the skill of somebody as new as AUrankBOOCH or BRrankellysonjunior01 only using team games.

As you can see, the balancer correctly put the worst player AUrankBOOCH with the best player LUrankAnir. Apparently AUrankBOOCH performed well above his assigned skill in this match. Thus he will receive a higher rating in the future.
+1 / -0

3 months ago
I did actually watch the replay. Booch did almost nothing of use (and took several minutes to plop his factory). Ellyson tried, but being a newbie likely didn't realise claymores are a bad call on a land map (and also got Anir in the face which was inevitably going to be messy).

It's part of an experienced player's role to educate a novice team mate about the game. Anir and Dark both gave Booch instructions which he belatedly followed. I gave Ellyson instructions when we were on the same side in Multiplayer B587282 4 on Calamity 1.1 some of which were followed and we won (albeit not before his base was wrecked by gunships that my fighters didn't put down quickly enough).
+1 / -0

3 months ago
I think even by hand this stack of players would be very hard to balance because the skillrange is so big.

At the same time i'm a bit surprised that i didnt got THrankPootis in my game instead of PLrankDarkstaron.
But thats allright, because with other players the system would probably have done a bether job.

But as the_green_squig sad, helping your noobs can often help alot ;)

+0 / -0

3 months ago
I don´t think you can have a decent balance if the skillrange between players is more than maybe 200 whr.
+2 / -0