Hi all,
So, coh2 charts is a big source of data for balance discussions, specifically win rates.
I don't understand one thing though: why is the average win rate 66%? I mean, correct me if I am wrong here, but for any number of winners, there has to be an equal number of losers correct? So why are the win rates always 0.50+ at least? And average is 0.66?
What exactly is the algorithm they use?
(Mods: I put this in the balance section because it seems pertinent, but move it to somewhere else as you deem fit)
A question about win rates and the coh2 charts algorithm
5 Nov 2015, 12:14 PM
#1
Posts: 509
5 Nov 2015, 12:18 PM
#2
Posts: 3552 | Subs: 2
I believe it is the win rates of the top 200 players, and they don't play each other all the time
The "Extras" will be the times they were playing opposition from outside the top 200 and won, which means the win shows in the stats but the loss doesn't
The "Extras" will be the times they were playing opposition from outside the top 200 and won, which means the win shows in the stats but the loss doesn't
5 Nov 2015, 12:19 PM
#3
Posts: 1225
Hi all,
So, coh2 charts is a big source of data for balance discussions, specifically win rates.
I don't understand one thing though: why is the average win rate 66%? I mean, correct me if I am wrong here, but for any number of winners, there has to be an equal number of losers correct? So why are the win rates always 0.50+ at least? And average is 0.66?
What exactly is the algorithm they use?
(Mods: I put this in the balance section because it seems pertinent, but move it to somewhere else as you deem fit)
The average win rate exclusively refers to the top 150 of players (ie. the stratum of the playerbase that has - by COH standards - a somewhat advanced understanding of the game mechanics and overall skillset).
Of course, these players will also face non-top 150 players, accounting for the asymetric win ratio.
Edit: Ja well, ninjaed.
5 Nov 2015, 12:19 PM
#4
Posts: 1604 | Subs: 3
The control group only consists of the top 150 players. Those don't face exclusively other top 150 members all the time. This means the players that a worse than the top 150 get the loses and percentages you look for.
5 Nov 2015, 16:19 PM
#5
Posts: 509
Ah! Ok THAT explains a lot. Ty guys. Atleast now I can use that winrate data in any balance arguments with that much greater confidence.
5 Nov 2015, 16:24 PM
#6
Posts: 1166 | Subs: 1
Ah! Ok THAT explains a lot. Ty guys. Atleast now I can use that winrate data in any balance arguments with that much greater confidence.
Really you shouldn't be using any data other than 1v1 data in your arguments. Any of the larger game modes (2v2 and up) are using random and not AT, which gives you random results.
5 Nov 2015, 20:17 PM
#7
Posts: 8154 | Subs: 2
Really you shouldn't be using any data other than 1v1 data in your arguments. Any of the larger game modes (2v2 and up) are using random and not AT, which gives you random results.
+1
You should only take extreme examples (+10% W/L) that goes through quite some time (discarding anomalies) for such modes. See what we have now on 2v2 or what used to happened on 3v3+ till before UKF.
6 Nov 2015, 19:35 PM
#8
Posts: 509
Really you shouldn't be using any data other than 1v1 data in your arguments. Any of the larger game modes (2v2 and up) are using random and not AT, which gives you random results.
I am aware, and I dont use any apart from 1v1 when I use it. But ty
8 Nov 2015, 08:06 AM
#9
Posts: 503
Really you shouldn't be using any data other than 1v1 data in your arguments. Any of the larger game modes (2v2 and up) are using random and not AT, which gives you random results.
Was just thinking same thing. Until/if ever they segregate AT and random, you will have AT facing randoms in team games giving big advantage to AT.
8 Nov 2015, 09:48 AM
#10
Posts: 17914 | Subs: 8
The control group only consists of the top 150 players. Those don't face exclusively other top 150 members all the time. This means the players that a worse than the top 150 get the loses and percentages you look for.
This.
Its far from perfect, but it as close as we can get without relic providing their own data.
PAGES (1)
1 user is browsing this thread:
1 guest
Livestreams
62 | |||||
14 | |||||
9 | |||||
8 | |||||
5 | |||||
137 | |||||
25 | |||||
12 | |||||
5 | |||||
5 |
Ladders Top 10
-
#Steam AliasWL%Streak
- 1.831222.789+37
- 2.600215.736+15
- 3.34957.860+14
- 4.1107614.643+8
- 5.305114.728+1
- 6.916405.693-2
- 7.273108.717+24
- 8.722440.621+4
- 9.1041674.607-2
- 10.17146.788+1
Replay highlight
VS
- cblanco ★
- 보드카 중대
- VonManteuffel
- Heartless Jäger
Einhoven Country
Honor it
9
Download
1235
Board Info
948 users are online:
948 guests
0 post in the last 24h
8 posts in the last week
34 posts in the last month
8 posts in the last week
34 posts in the last month
Registered members: 49107
Welcome our newest member, Falac851
Most online: 2043 users on 29 Oct 2023, 01:04 AM
Welcome our newest member, Falac851
Most online: 2043 users on 29 Oct 2023, 01:04 AM