Page 4 of 6
Re: Ballistic Pro Ladder
Posted: Wed 19 Jun , 2013 2:59 pm
by iZumo
I haven't finished the patch, but old data will gradually decay, based on new data (not time). There will also be fully recoverable inactivity decay to purge inactive players. It will be the final patch to the skill system.
Re: Ballistic Pro Ladder
Posted: Wed 19 Jun , 2013 3:21 pm
by iRobot
One patch only?
That's worse support than EA
FML
Re: Ballistic Pro Ladder
Posted: Wed 19 Jun , 2013 3:36 pm
by iZumo
Why lie? There was also EFR patch, thawing patch ...
Re: Ballistic Pro Ladder
Posted: Wed 19 Jun , 2013 3:41 pm
by iRobot
Those were almost basically day one DLC that should have been in there originally
Re: Ballistic Pro Ladder
Posted: Wed 19 Jun , 2013 3:42 pm
by iZumo
Yes, there is day one DLC in there. My abuse tools
Re: Ballistic Pro Ladder
Posted: Fri 05 Jul , 2013 7:31 pm
by Butcher
I made it on Matlab code to test it fast, but if you want I can write it in .uc code too.
The skill is calculated from the end of the match Kills+Thaw:Deaths ratio -> gives you a "new score" after every match.
You accumulate them for an amount of times:
Code: Select all
function [ New_Skill_Vect ] = Insert_New_score( New_score,Skill_Vect )
%First in Last Out function
if New_score>0 % for not adding when you just entered to a match
New_Skill_Vect=Skill_Vect*0;
for iii=2:length(Skill_Vect)
New_Skill_Vect(iii)=Skill_Vect(iii-1);
end
New_Skill_Vect(1)=New_score;
else
New_Skill_Vect=Skill_Vect;
end
end
Then you average the vector over the length of it (30 matches for example)
Ladder Skill caculation:
Code: Select all
function [ New_skill ] = Calc_vect_Skill( Skill_vect )
% Calculates average in the vector
New_skill=sum(Skill_vect)/length(Skill_vect);
end
and you have something like this (10 sample test)
Code: Select all
Player_Skill_Vect =
5.1000 0 0 0 0 0 0 0 0 0
New_skill =
0.5100
Player_Skill_Vect =
4.5000 5.1000 0 0 0 0 0 0 0 0
New_skill =
0.9600
Player_Skill_Vect =
8.3000 4.5000 5.1000 0 0 0 0 0 0 0
New_skill =
1.7900
Player_Skill_Vect =
8.7000 8.3000 4.5000 5.1000 0 0 0 0 0 0
New_skill =
2.6600
Player_Skill_Vect =
12.1000 8.7000 8.3000 4.5000 5.1000 0 0 0 0 0
New_skill =
3.8700
Player_Skill_Vect =
3.1000 12.1000 8.7000 8.3000 4.5000 5.1000 0 0 0 0
New_skill =
4.1800
Player_Skill_Vect =
8.1000 3.1000 12.1000 8.7000 8.3000 4.5000 5.1000 0 0 0
New_skill =
4.9900
Player_Skill_Vect =
12.5000 8.1000 3.1000 12.1000 8.7000 8.3000 4.5000 5.1000 0 0
New_skill =
6.2400
Player_Skill_Vect =
5.7000 12.5000 8.1000 3.1000 12.1000 8.7000 8.3000 4.5000 5.1000 0
New_skill =
6.8100
Player_Skill_Vect =
4.0000 5.7000 12.5000 8.1000 3.1000 12.1000 8.7000 8.3000 4.5000 5.1000
New_skill =
7.2100
Player_Skill_Vect =
18.3000 4.0000 5.7000 12.5000 8.1000 3.1000 12.1000 8.7000 8.3000 4.5000
New_skill =
8.5300
Player_Skill_Vect =
8.7000 18.3000 4.0000 5.7000 12.5000 8.1000 3.1000 12.1000 8.7000 8.3000
New_skill =
8.9500
Player_Skill_Vect =
13.1000 8.7000 18.3000 4.0000 5.7000 12.5000 8.1000 3.1000 12.1000 8.7000
New_skill =
9.4300
Player_Skill_Vect =
8.1000 13.1000 8.7000 18.3000 4.0000 5.7000 12.5000 8.1000 3.1000 12.1000
New_skill =
9.3700
Player_Skill_Vect =
8.5000 8.1000 13.1000 8.7000 18.3000 4.0000 5.7000 12.5000 8.1000 3.1000
New_skill =
9.0100
Player_Skill_Vect =
13.5000 8.5000 8.1000 13.1000 8.7000 18.3000 4.0000 5.7000 12.5000 8.1000
New_skill =
10.0500
Re: Ballistic Pro Ladder
Posted: Fri 05 Jul , 2013 7:38 pm
by iRobot
Butcher wrote:The skill is calculated from the end of the match Kills+Thaw:Deaths ratio -> gives you a "new score" after every match.
Butcher wrote:from the end of the match Kills+Thaw:Deaths ratio -> gives you a "new score"
Butcher wrote:end of the match Kills+Thaw:Deaths ratio
Butcher wrote:Kills+Thaweaths
Re: Ballistic Pro Ladder
Posted: Fri 05 Jul , 2013 8:26 pm
by iZumo
There is a bit more complex math behind it. It's a Bayesian estimate over the collected data to weight the amount and we're indeed using efficiency which is: ( kills + (thaws / thaws_to_kill_ratio) ) / deaths. Furthermore, I'm about to make a decay system to cover player inactivity and decay old data (too little time now).
Re: Ballistic Pro Ladder
Posted: Sat 06 Jul , 2013 4:11 am
by Butcher
I see I got your attention...
Izumo_CZ wrote:There is a bit more complex math behind it. It's a Bayesian estimate over the collected data to weight the amount and we're indeed using efficiency which is: ( kills + (thaws / thaws_to_kill_ratio) ) / deaths. Furthermore, I'm about to make a decay system to cover player inactivity and decay old data (too little time now).
Make it simple or complex, but implement please !!
Re: Ballistic Pro Ladder
Posted: Tue 23 Jul , 2013 9:55 am
by Butcher
Izumo_CZ wrote:There is a bit more complex math behind it. It's a Bayesian estimate over the collected data to weight the amount and we're indeed using efficiency which is: ( kills + (thaws / thaws_to_kill_ratio) ) / deaths. Furthermore, I'm about to make a decay system to cover player inactivity and decay old data (too little time now).
Izumo_CZ wrote:
The "inactive delay" starts to kick after 14 days of inactivity with expected play time 30 min / day and makes your skill equal to the average skill after 28 days of pure inactivity, but it's fully recoverable if you start playing again. Old IDs will now be merged with much greater ease.
Could you tell us how the skill is exactly calculated now?
Skill=f(Thaws,kills,Deaths,time) where f=????