Loading...
  OR  Zero-K Name:    Password:   
Map: TheHunters Redux 1.3
 TheHunters Redux 1.3
Title: [A] Free For All
Host: Nobody
Game version: Zero-K v1.9.12.0
Engine version: 104.0.1-1544-ge1f249f
Battle ID: 1261237
Started: 2 years ago
Duration: 31 minutes
Players: 8
Bots: False
Mission: False
Rating: Casual
Watch Replay Now
Manual download

Team 1
Chance of victory: 1.9%

GBrankSab
Team 2
Chance of victory: 1.2%

ATrankPLT_Atosty
Team 3
Chance of victory: 0%

GBrankJonathan1321
Team 4
Chance of victory: 96.9%

SErankPLT_Godde
Team 5
Chance of victory: 0%

unknownrankaioeieoiao
Team 6
Chance of victory: 0%

AMrankDimmu_Borgir
Team 7
Chance of victory: 0%

USrankKids2504
Team 8
Chance of victory: 0%

USrankkids
Spectators
DErankHeldenstein
DErankDerPizzaHouse
CNrankNT

Show winners



Preview
Filter:    Player:  
sort

RUrankizirayd
2 years ago
(edited 2 years ago)
Literally 2 lost ffa matches cost Godde about 180 metal. I think this is absurd. FFA Elo loss for players with 2750+ Elo is very expensive and not balanced. I was 2800+ for a long time and recently 2800, having recently played a couple of ffa, I dropped to 2700, and while playing for a long time I dropped to 2535 Elo, after playing regular casual matches, I got 2870 Elo back. That is, my ffa rating level is about 2500, and my level of ordinary team games is 2700-2800.

I believe that the formula for ffa is described with an error and does not correctly calculate the chance of winning when one of the players has 200 points higher elo. Moreover, the events in ffa are very difficult, and a group of players with 2600 Elo can get a strong economy and super weapons, and the difference between who wins, a player with 2800 Elo or 2600 is not there. Already 3 players for 2600+, the one who gets the starlight first, literally with an 80% chance to win, because the starlight breaks the singu instantly and after the broken singu, the combo does not exist. Therefore, trying to predict the winner in ffa is very strange.

I am sure that playing ffa and staying at 2900 Elo is super difficult.
+0 / -0

RUrankizirayd
2 years ago
I never understood why there is no separate ffa rating, even in the BAR it is
+8 / -0


AUrankAdminGoogleFrog
2 years ago
As far as I recall, GBrankAdminDeinFreund ran tests with and without FFA rating and found better teamgame winrate prediction by including FFA. This is why FFA counts for rating. I think you were already told this.

GBrankAdminDeinFreund implemented it so he got to make decisions like this. I can accept his stats and still disagree that FFA should count for rating. I think a PR that made FFA not affect rating would be pulled.
+1 / -0

RUrankizirayd
2 years ago
I don’t understand how this relates to my experience of playing ffa, I had 5 consecutive defeats in ffa with a 70% predicted chance of winning. This 70% chance is a modifier for rating points reduction and greatly lowers my rating, leveling the rating points so that they match my winrate. The casual rating works pretty well in this regard, it balances well and reflects the overall level of the game - ffa rating does not do it perfectly.

Okay, for example Godde, he lost two ffa games in a row with a 96% prediction. I don’t deny that two consecutive losses at 96% is quite a possible outcome, but it’s almost unbelievable. A player who sets a 1v1 elo record in 3561 elo and for a casual rating of 3360, he suddenly loses the game to a player with a 0.2% chance of winning. What is the value of 10 casual rating points at all, at 3350+? Two games and minus 160 rating points, provided that a 96% win is predicted. Do you see contradictions or only I see them?


--
I think that Daine's work on removing the ffa rating from the general casual rating contains errors either in the calculations or in the conclusions that he made and an error occurs in the general logic of understanding the current situation and requires re-evaluation.
+1 / -0

RUrankizirayd
2 years ago
Do you realize what a 0.2% chance of prediction is? This is 1/500, that is, in the current configuration, all 499 other games GBrankJonathan1321 must lose, of course, maybe he will win, and even a couple of times, the chance of this is possible, but in reality, of course, we are talking about dozens of victories, if not more, I am almost sure of that.

1/500 - a miracle happened in this game
+0 / -0


AUrankAdminGoogleFrog
2 years ago
There are comparatively few FFAs, so I could believe that whatever scoring DF used did not work in their favour. I don't remember what he said he did.

The short answer is that if anyone actually went to remove FFAs from the rating system, I don't think anyone would stop them. Someone has to actually do it though.
+0 / -0


SErankPLT_Godde
2 years ago
I like that it is so easy to lose lobsterrating in FFAs.
+0 / -0


SErankPLT_Godde
2 years ago
Really I should not have tried to spam out like 6 singus at the same time :P
+1 / -0
FRrankmalric
2 years ago
quote:
Do you see contradictions or only I see them?
I think it always depends on what your objective - it is not hard to imagine how including FFA results improves overall balance/prediction, while at the same time makes rating more noisy for extremes (as in players very far from the average player).

I always had the opinion that we should have more (separate) public ratings, even if that implies slightly less balanced games for team games - so I do agree with what you suggest, but don't think that the current way is "wrong", just think it optimizes with other trade-offs.

+0 / -0


PLrankRafalpluk
2 years ago
(edited 2 years ago)
quote:
The short answer is that if anyone actually went to remove FFAs from the rating system, I don't think anyone would stop them. Someone has to actually do it though.

Imho FFA's should not affect Elo because of their inherently random nature. If you play a lot of FFA's you notice a pattern that in a longer timespan it causes the player's Elo to converge to average:
- the purple stars keep losing Elo while playing FFA's because they lose more often than it would be expected from the Elo, because of random factors
- at the same time players with low Elo will keep gaining Elo because they can randomly win some games which they otherwise would never win looking at their Elo, again because of random factors. A good example is this game.

TLDR: FFA is Elo-averaging game mode. Which means it screws up well-balanced teams-Elo when played. Which means it should not affect said Elo.

It can still use the Elo for balancing, for example in team FFA's, but it should be read-only mode, because these games are still random, so it is quite pointless to determine ratings based on them.
+0 / -0


AUrankAdminGoogleFrog
2 years ago
PLrankRafalpluk I'm not sure why you quoted me there.

quote:
It can still use the Elo for balancing, for example in team FFA's, but it should be read-only mode, because these games are still random, so it is quite pointless to determine ratings based on them.

I agree.

I don't think your preceding argument is necessary though, and I would much rather see it carried out with an analysis of the data. I can imagine you being wrong here, but I won't care whether this argument is wrong because there are other reasons to remove rating changes from FFA.
+0 / -0