Loading...
  OR  Zero-K Name:    Password:   
Title: teams
Host: unknownrankShaman
Game version: Zero-K v1.6.5.4
Engine version: 104.0.1-287-gf7b0fcc
Started: 5 years ago
Duration: 11 minutes
Players: 3
Bots: False
Mission: False
Watch Replay Now
Manual download

Team 1
Chance of victory: 87.9%

unknownrankShaman
unknownrankHydris
Team 2
Chance of victory: 12.1%

USranklytzhadow
Spectators
FRrankmalric

Show winners



Preview
Filter:    Player:  
sort
5 years ago
team balance algorithm stupid
+1 / -0
How so? You need to put the worst player with the best player, in order to handicap the good player. Giving him a bad teammate that uses 50% of metal is the most reliable way of reducing his team's performance.

If you had been put with USranklytzhadow your win chance would've been in the <1% range. That's what I'd call bad balance..
+6 / -0

5 years ago
Balance system really is bad. Games should be allowed started only with even numbers of players.
quote:
Giving him a bad teammate that uses 50% of metal is the most reliable way of reducing his team's performance.

Not in this situation because unknownrankHydris and USranklytzhadow are equally bad. So one use 50% metal inefficiently but other 100% metal. This chance of victory is just worthless mathematical calculation in such situations.
+0 / -0
5 years ago
Saying "not in this situation" does not make sense if you think about probabilities. Like flipping a coin - 50% chance to guess correctly - and then if you "loose" say "but it did not work in this situation".

Also I find "equally bad" an exaggeration (rank 281 vs rank 418 is still some difference)
+0 / -0

5 years ago
quote:
Not in this situation because unknownrankHydris and USranklytzhadow are equally bad. So one use 50% metal inefficiently but other 100% metal. This chance of victory is just worthless mathematical calculation in such situations.


Yes in this situation! Or do you think a 1% win chance is the same as a 30% win chance?
+0 / -0

5 years ago
This is one of the cases where it would've been great if double comm didn't exist.
+2 / -0

5 years ago
quote:
Yes in this situation! Or do you think a 1% win chance is the same as a 30% win chance?

In this situation these calculated percentages is worthless. I already write why. Percentage could be more real if in team 1 would be SErankGodde with unknownrankHydris and in Team 2 @_Shaman.
+1 / -0

5 years ago
Do you think the percentages are wrong? How would you say they should be?
+0 / -0


5 years ago
30.9% is close to 1 in 3 chance. It sounds about right. Remember that there are multitudes of factors for performance:

1. Cooperation / Coordination
2. Mood
3. Motivation
4. Initial factories
5. Hunger (yes, this is an actual factor. Performance degrades rapidly on an empty stomach.)
6. Player energy level (Tiredness/fatigue also degrade performance rapidly due to lengthened reaction time)
7. Awareness
8. Explicit planning

I'd say the <1% chance would be an overestimation of my abilities. I'd say it's somewhere in the 20-25% chance range.
+0 / -0


5 years ago
Is there any data to suggest that the 30% chance is correct? 2v1 games are rare so the matchmaker is not trained on them.
+0 / -0


5 years ago
1 in 3 seems to be around the right percentage of wins for a 2 com player if they're aware of certain strategies they can do in 2v1. As the number of players on the bigger side grows though, this chance falls due to the number of potential players who will go air, scout your cheese then promptly shut it down with swifts or thunderbirds as well as the associated opportunity costs being lowered for every player that joins in. If one takes out a player as the 2com player, your odds of winning go up with the highest gains being in 2v1. The best places for uneven teams tend to be in 2v1 and 3v2 since taking out a player in these games is far more likely (and air is far more costly in terms of opportunity costs) than in 6v5 or 5v4. In larger uneven games, the 2 com player has far less influence since taking out a player using drops or some rush is far harder due to the opportunity costs of defending against said rush are far lower than if they were done in 2v1.
+0 / -0

5 years ago
quote:
Is there any data to suggest that the 30% chance is correct? 2v1 games are rare so the matchmaker is not trained on them.

WHR works on people, not teams.
  • if a double comm is perfectly equivalent to two players of that rating, then the entire match history is data to suggest the chance is correct.
  • if it is perfectly skewed (ie. 2-man always wins or always loses) then WHR will never give correct answers.

My impression of double comm is that it it roughly equivalent but depends on circumstances so the prediction can be fairly wrong but should generally be the same ballpark.
+0 / -0
@_Shaman you've given me a sanity check, not data.

PLrankAdminSprung WHR works on people in teams. The common team configurations are likely to have better balance because the volume of games is a weighting for accurate WHR.

My question was whether someone has checked whether the predictions for 2v1 games are accurate. The only way to check this is to look at the data, that is, the outcomes of the games themselves. I am unconvinced by the squishy reasons given for the prediction being accurate.
+0 / -0