DUST Search - [FEEDBACK]: Mu and afk farming

Nothing Certain wrote:

Ryme Intrinseca wrote:

CCP Rattati wrote:

couldn't agree more

So what to do about it?

Let's say me and Kero are equally good players. But I never AFK and he AFKs, say, half his games. His Mu will be radically lower than mine, so he'll get much easier games.

I think you could deal with this by first ascertaining if the distribution of WP/S (99% correlated with mu, remember) is bimodal (in layman's terms, peaks at either extreme separated by a trough). If it is, then for mu purposes simply discount all the games from the low WP/S peak.

So in the example I gave before, suppose I played a thousand games (average 1.5 WP/S), while Kero played a thousand games (also 1.5 WP/S on average) and AFKed a thousand games (near zero WP/S). Using the system I just described, his AFKing would get excluded as located at a low WP/S peak. So our mu would get calculated just on the basis of the thousand games each that we actually played.

It seems unlikely that his AFKing would give him easier games. His Mu would be lower but the result would only be that he would be seeded lower on the team rankings, with all the players averaged it isn't going to affect him very much in team selection. Systemically this might be a problem but for any individual there is no real incentive to deliberately lower your Mu.

AFK is bad simply because it rewards doing nothing, which hurts everybody else.


Author	Thread Statistics \| Show CCP posts - 1 post(s)
Ryme Intrinseca Fatal Absolution General Tso's Alliance 1875	Posted - 2014.09.17 20:12:00 - [1] - Quote CCP Rattati wrote: couldn't agree more So what to do about it? Let's say me and Kero are equally good players. But I never AFK and he AFKs, say, half his games. His Mu will be radically lower than mine, so he'll get much easier games. I think you could deal with this by first ascertaining if the distribution of WP/S (99% correlated with mu, remember) is bimodal (in layman's terms, peaks at either extreme separated by a trough). If it is, then for mu purposes simply discount all the games from the low WP/S peak. So in the example I gave before, suppose I played a thousand games (average 1.5 WP/S), while Kero played a thousand games (also 1.5 WP/S on average) and AFKed a thousand games (near zero WP/S). Using the system I just described, his AFKing would get excluded as located at a low WP/S peak. So our mu would get calculated just on the basis of the thousand games each that we actually played.
Ryme Intrinseca Fatal Absolution General Tso's Alliance 1892	Posted - 2014.09.18 11:23:00 - [2] - Quote Nothing Certain wrote: Ryme Intrinseca wrote: CCP Rattati wrote: couldn't agree more So what to do about it? Let's say me and Kero are equally good players. But I never AFK and he AFKs, say, half his games. His Mu will be radically lower than mine, so he'll get much easier games. I think you could deal with this by first ascertaining if the distribution of WP/S (99% correlated with mu, remember) is bimodal (in layman's terms, peaks at either extreme separated by a trough). If it is, then for mu purposes simply discount all the games from the low WP/S peak. So in the example I gave before, suppose I played a thousand games (average 1.5 WP/S), while Kero played a thousand games (also 1.5 WP/S on average) and AFKed a thousand games (near zero WP/S). Using the system I just described, his AFKing would get excluded as located at a low WP/S peak. So our mu would get calculated just on the basis of the thousand games each that we actually played. It seems unlikely that his AFKing would give him easier games. His Mu would be lower but the result would only be that he would be seeded lower on the team rankings, with all the players averaged it isn't going to affect him very much in team selection. Systemically this might be a problem but for any individual there is no real incentive to deliberately lower your Mu. AFK is bad simply because it rewards doing nothing, which hurts everybody else. My W/L is a touch over 2. If I AFKed half my games, you'd expect my win/loss to go down to about 1.4 (because my W/L when AFKing is a bit below 1). There would be a similar effect on Mu. Suppose the best five players in a game (other than me ) have Mus like this (obviously I don't know the scale, so I will just use W/L-type numbers; also I assume the other 13 players on each side are constant): 2, 1.7, 1.7, 1.4, 1.4 Without AFKing, my Mu is 2. So the teams will probably get arranged like this: Blue A - 2 (me), 1.7, 1.4 Red A - 2, 1.7, 1.4 With AFKing, the teams are probably like this: Blue B - 1.7, 1.7, 1.4 Red B - 2, 1.4 (me), 1.4 Of course, those teams, with actual, non-manipulated Mu, look like this: Blue C - 1.7, 1.7, 1.4 Red C - 2, 2 (me), 1.4 Red C is clearly a stronger team than Red A. Likewise, Blue C is clearly a weaker team than Blue A. I think it is clear therefore that lowering Mu via AFKing does give you easier games, because the matchmaker will then make your side stronger and the other side weaker, as in this example.