Author Topic: CAN YOU FUCKING FIX THIS ALREADY ? (Read 5758 times)

ClaraSchuman · « **Reply #15 on:** May 07, 2016, 12:58 »

Quote from: Hagarath on May 07, 2016, 12:48

Quote from: CoMMoN1337 on May 06, 2016, 21:34
A 2 YEAR OLD COULD CREATE A BETTER POINT SYSTEM THAN THIS ONE

IT IS UNACCEPTABLE FOR A 50%/50% GAME TO AWARD 0 PSR AT WIN AND 0 PSR AT LOSS, AND THE SAME GAME WITH THE SAME PLAYERS REARRANGED FOR 88%/12% TO AWARD +1 AT WIN FOR THE 88%

ABSOLUTELY UNACCEPTABLE

CEN SHOULD ASK HIMSELF HOW MANY PLAYERS DOES HE EXPECT TO HAVE IN 1 MONTH FROM NOW, 2 MONTHS FROM NOW, AND SO ON, IF HE DOES NOTHING TO RECTIFY HUGE MISTAKES AND IMPROVE THINGS

THE ANSWER IS 0

IF THAT'S WHAT YOU WANT, THEN SO BE IT

CLOSE THIS THREAD, AND ENJOY YOUR SERVER'S SLOW AND PAINFUL DEATH

GOODBYE

Anger won’t solve anything. Remember that there will always be tournaments for those with higher PSR, they get the most out of the rewards. I appreciate your concerns on the PSR problem, having a few "high" ranked players active it is a problem indeed to rise up the ladder because other ones aren’t as active.

Kind Regards,
Hagarath

rep+ u are right my friend..

cen · « **Reply #16 on:** May 08, 2016, 13:24 »

1. I did not code the current PSR system.
2. PSR code is 400 lines of code and works by ELO rating https://en.wikipedia.org/wiki/Elo_rating_system. Which is supposedly a fair system because a lot of serious games use it.
3. That being said, I did some calculations and online ELO calculators were showing that you should get more +/- in the ELO system. But in our bots there is an interesting twist added:

Quote

if(distancefromteamavgpsr > 100 )
   distancefromteamavgpsr = 100;

if( distancefromteamavgpsr > 0 && distancefromteamavgpsr <= 100 )
{
   double implayingwithnewbiesfactor = ( 100.0 - distancefromteamavgpsr ) / 100;
   m_team1kfactors = implayingwithnewbiesfactor * m_team1kfactors;
}

This is a safety mechanism which mun? added so top rated players can't abuse the system. The more you differ from the average the more you will get fucked by lowering your K factor. In fact, if the difference is >100 your K factor will drop to almost 0.

Without this mechanism you can calculate what you would actually get: http://ratings.fide.com/calculator_rtd.phtml (K=20, opponent 1637) which turns out to be 4.4.

And now I guess we can start a holy war if this is fair or not. Some will say this is good so top players can't farm on lowskills, others will say that it's hard to find top opponents in a game anyway and you can't get any points once you get a high rating. I'm sure this was probably discussed somewhere in private forums when it was added (or maybe not, I don't have time to dig it up).

Defining na minimum K factor (so it can't drop below a certain number) would probably be a good solution. Probably K=10 since that is used for professionals in chess. Based on online calculator that would be +2.

CoMMoN1337 · « **Reply #17 on:** May 08, 2016, 13:45 »

Quote from: cen on May 08, 2016, 13:24

1. I did not code the current PSR system.
2. PSR code is 400 lines of code and works by ELO rating https://en.wikipedia.org/wiki/Elo_rating_system. Which is supposedly a fair system because a lot of serious games use it.
3. That being said, I did some calculations and online ELO calculators were showing that you should get more +/- in the ELO system. But in our bots there is an interesting twist added:

Quote
if(distancefromteamavgpsr > 100 )
   distancefromteamavgpsr = 100;

if( distancefromteamavgpsr > 0 && distancefromteamavgpsr <= 100 )
{
   double implayingwithnewbiesfactor = ( 100.0 - distancefromteamavgpsr ) / 100;
   m_team1kfactors = implayingwithnewbiesfactor * m_team1kfactors;
}

This is a safety mechanism which mun? added so top rated players can't abuse the system. The more you differ from the average the more you will get fucked by lowering your K factor. In fact, if the difference is >100 your K factor will drop to almost 0.

Without this mechanism you can calculate what you would actually get: http://ratings.fide.com/calculator_rtd.phtml (K=20, opponent 1637) which turns out to be 4.4.

And now I guess we can start a holy war if this is fair or not. Some will say this is good so top players can't farm on lowskills, others will say that it's hard to find top opponents in a game anyway and you can't get any points once you get a high rating. I'm sure this was probably discussed somewhere in private forums when it was added (or maybe not, I don't have time to dig it up).

Defining na minimum K factor (so it can't drop below a certain number) would probably be a good solution. Probably K=10 since that is used for professionals in chess. Based on online calculator that would be +2.

This system is used for chess, chess is a 1v1 last I checked and not a 5v5 game.

Here you have !balance command, which takes into account the psr of everyone and balances the game.

That means if you have 1800 and all enemies have 1700 and if it's 50/50 balance, then allies of the 1800 will overall have less psr than the enemies, so that "top player feeds on lowskill" is valid for the others too not just the 1800, the 1700's are going to feed on the lowskill allies of the 1800 too because the allies of the 1800 are far worse than them.

So, taking that into account, how does it make any sense for it to be +0 in a 50/50 game ? No matter how high or low your psr is, if it's 50/50 that means the others in your team compensate for you being high/low by being low/high themselves.

So that +0 condition created and added by whoever, makes absolutely no sense, do you understand ?
This is a real problem and it must be fixed ...

cen · « **Reply #18 on:** May 08, 2016, 13:59 »

50/50 does not really matter, the output of that is deceiving. As you can see from my post, even if it's 50/50 it still calculates the factors based on your PSR in comparison to average. So if you are 100+ points above average in current system it will lower you.

I am willing to minimize this punishment as I already said, I think it's fair. Current ELO uses too small K factors which is wrong. There must be a reason why chess uses 10 as minimum.

cen · « **Reply #19 on:** May 11, 2016, 18:14 »

I think this could be done next week. Will update with more info.

CoMMoN1337 · « **Reply #20 on:** May 13, 2016, 00:02 »

Quote from: cen on May 11, 2016, 18:14

I think this could be done next week. Will update with more info.

That's good to hear, it's really sad to not be able to play games at all almost all day because it's +0 no matter what, only from like 18:00 til like 22:00 there is chance for 1650+ games, 1600+ games are +0 now and even some 1650+ games are +0.