If you want game quality and Best team composition inside FA

[Ze_PilOt]

I'm not expecting someone doing it, but you can convert Trueskill this to LUA. This is the only way.

Here is the source code in python :
https://bitbucket.org/thepilot/python-trueskill/src

If you decide to do it, you can ask any question about the code here (like, where to start ?

[FunkOff]

Trueskill is THAT many files?

[Ze_PilOt]

There is a 1v1 version without the factor graph that I'm not using. (give the same result with or without, and the speed gain is not significant).
So that file doesn't need to be converted : https://bitbucket.org/thepilot/python-t ... culator.py

There is a also a PHP version if you are more at easy with PHP.

Yes, it's a lot of code. Microsoft sell it for a reason

[Ze_PilOt]

Hm actually I don't need to convert trueskill entirely, just the part that compute game balance, and that doesn't requires the factor graph, so it can be quickly done..

If someone want to know what is precisely needed, it's all the functions used here :

Code: Select all

    def calculateMatchQuality(self, gameInfo, teams) :

#        // We need to create the A matrix which is the player team assigments.
        teamAssignmentsList = teams



        skillsMatrix = self.getPlayerCovarianceMatrix(teamAssignmentsList)
       
       
        meanVector = self.getPlayerMeansVector(teamAssignmentsList)

        meanVectorTranspose = meanVector.getTranspose()

        playerTeamAssignmentsMatrix = self.createPlayerTeamAssignmentMatrix(teamAssignmentsList, meanVector.getRowCount())
       

       
        playerTeamAssignmentsMatrixTranspose = playerTeamAssignmentsMatrix.getTranspose()


        betaSquared = square(gameInfo.getBeta())

        start = Matrix.multiply(meanVectorTranspose, playerTeamAssignmentsMatrix)
       

       
        aTa = Matrix.multiply(Matrix.scalarMultiply(betaSquared,
                                               playerTeamAssignmentsMatrixTranspose),
                                                playerTeamAssignmentsMatrix)
        aTSA = Matrix.multiply(
                    Matrix.multiply(playerTeamAssignmentsMatrixTranspose, skillsMatrix),
                    playerTeamAssignmentsMatrix)



        middle = Matrix.add(aTa, aTSA)


        middleInverse = middle.getInverse()

        end = Matrix.multiply(playerTeamAssignmentsMatrixTranspose, meanVector)

        expPartMatrix = Matrix.scalarMultiply(-0.5, (Matrix.multiply(Matrix.multiply(start, middleInverse), end)))
        expPart = expPartMatrix.getDeterminant()

        sqrtPartNumerator = aTa.getDeterminant()
        sqrtPartDenominator = middle.getDeterminant()
        sqrtPart = sqrtPartNumerator / sqrtPartDenominator

        result = exp(expPart) * sqrt(sqrtPart)

        return result

So, basically, coding some matrices functions in lua.

[Ze_PilOt]

And of course, it's done already :

http://lua-users.org/wiki/LuaMatrix

[Ze_PilOt]

Oh god, okay, I will do it

[Gowerly]

<3

[Pavese]

i'm not sure if this is the best place for feedback:

when the people ingame become an uneven Number (7/8 i.e.) the game quality drops to 0% levels allthough all players inside are fairly equal and are just waiting for someone to join.

The low number might irritate people about the actual game balance.

Is it possible to create a "phantom" player, when the numbers are uneven, that has the average rating of all players inside? This way your formula will always calculate from an even numbers, or how it will be if another evenly matched player would join.

[Ze_PilOt]

It's about how you think people are dumb.

If they see 7 players on a 4vs4, they can assume that the low rating is due to a player missing.

If you don't believe that anybody is capable of understanding that, do you REALLY want them in your game ?

[Gowerly]

Possibly instead of that, remove the average of all players on the larger team (so the game assumes there's one fewer player on the larger team), which will then be compared against the opposing team.

However, a 4v3 should never be 0% anyway. It's unlikely to be THAT unbalanced. Even in HoN, 5v4s are always like 83 - 17 for unmatched sides.