Territory War Matchmaker

"CG_SBCrumb;d-182579" wrote:
Hi folks,

We are aware that some Guilds are battling in Territory Wars against higher GP Guilds with a deeper roster and can feel impossible to beat. Currently, we don’t see any egregious issues in the data but, after looking at the posts on the forums, it’s clear that we need to investigate further. Territory War Matchmaker is something we have been monitoring closely for a while but we still need to gather more info about the current situation.

I am combing through the forum for past GP mismatches for my report but I could use your help. Please post here with the following info if you have experienced a large mismatch in Guild GP during Territory War in the last couple months (last change to TW matchmaker was in March 2018):

• Date of TW:
  
• Your Guild’s GP
  
• Opposing team’s GP:
  
• # of my guild members who joined:
  
• # of opposing guild members who joined:
  

Attach any screenshots of the above info if you have them.

The problem is caused by the matching algorithm. What is happening is it is matching guilds purely on GP. The problem arises when you have a guild with say only 30 members and one with 50 and both have the same GP. The guild with 50 members is disadvantaged because overall the individual teams are weaker.

If we take a look at 2 guilds with 60M GP; one with 30 members and one with 60 then the average power of each member of the 30 guild is 2M while the average of the 50 guild is just 1.2M. The weaker teams are wiped out in an instant and there just aren't enough heavy hitters left to win.

A quick dirty fix would be to calculate the mean GP of the guild (although this would slightly disadvantage guilds that are short of members it would not be as bad as using gross guild power). You could then tweak it a little by weighting it according to the number of members and get somewhere near.

If you want to do it properly you could simulate the battles on the server and use a bit of ML, but obviously this would take a little longer to code and may be a little on heavy on the processors.

If I was doing it I would create a vector of the guild members individual power and find the the mean using a for loop (schoolboy stuff). I would then take the gross guild power and take the average. I would then use the average to weight the mean and use the weighted mean to match guilds.

You are always going to have winning and losing streaks, but this is a clear identifiable problem (it might not be the only problem, but it certainly is one).

This may very well create a mismatch on Gross GP, but this is irrelevant by virtue of the nature of the battles and the fact that each team only gets one go.