??? Won but Lost ???

[kentuequi]

I(265) have just won an easy match against Ryod77(131), but my new rating fall to 264 ?!? Can somebody explain me ?

[Fiend]

I(265) have just won an easy match against Ryod77(131), but my new rating fall to 264 ?!? Can somebody explain me ?

Maybe the difference was so big that SE have decided to just decrease rank. Lol

[rumsey]

Check out this great thread by shannong on why the ratings system sometimes acts strangely.

[kentuequi]

Check out this great thread by shannong on why the ratings system sometimes acts strangely.

This great thread give me a headache everytime i try to read it... Ok i understand, but for my 10th victory in a row, saying that it's the system don't bring me any compensation... There's definitly something wrong in that system. Next time i'll play against a newbie i'll make him suffer ! That will be my compensation ! Lol

[bob21218]

I(265) have just won an easy match against Ryod77(131), but my new rating fall to 264 ?!? Can somebody explain me ?

The general notion of the TrueSkill algorithm is that for each game you win, your estimated strength increases and the uncertainty in that strength decreases. Some TrueSkill implementations have a constant "tau" which is used to keep things a bit more dynamic (so it does not become so convinced it knows your score you have trouble going up or down). If Kyle has included this factor, then something like this could happen:

your data before the match
mu = 294.5001
sigma = 10
rating = 264.5001 (mu - 3*sigma)
displayed rating = 265 (just rounded)

your data after the match:
mu = 294.559 (this increased as expected)
sigma = 10.024 (this INCREASED a bit due to tau and the weakness of your opponent)
rating = 264.49 (mu - 3*sigma)
displayed rating = 264 (rounded off and now one less than before)

If your next game is a stronger opponent and you win, you will probably see the rating point return.

[kentuequi]

The general notion of the TrueSkill algorithm is that for each game you win, your estimated strength increases and the uncertainty in that strength decreases. Some TrueSkill implementations have a constant "tau" which is used to keep things a bit more dynamic (so it does not become so convinced it knows your score you have trouble going up or down). If Kyle has included this factor, then something like this could happen:

your data before the match
mu = 294.5001
sigma = 10
rating = 264.5001 (mu - 3*sigma)
displayed rating = 265 (just rounded)

your data after the match:
mu = 294.559 (this increased as expected)
sigma = 10.024 (this INCREASED a bit due to tau and the weakness of your opponent)
rating = 264.49 (mu - 3*sigma)
displayed rating = 264 (rounded off and now one less than before)

If your next game is a stronger opponent and you win, you will probably see the rating point return.

Back to school... Thank you for this easy explanation, now i really understand, and thanks rumsey too, and Friend... Stop torturing cats !!!

[Fiend]

Back to school... Thank you for this easy explanation, now i really understand, and thanks rumsey too, and Friend... Stop torturing cats !!!

The whole point of the lolcat picture was that you would look at it and think that YOU made kitty scared. xD but thanks for noticing.