Alpha wrote:
I would suggest >25. G-rating works this way for individual boards. You are not assigned a g-rating until you win which is strange and possibly a bug.
Wouldn't this just be the necessity of acquiring both a numerator and denominator before doing the calculation?
here are the correct H-ratings:Alpha: .6052
Hugh: .7073
ASM: .6771
Norseman: .7586
Yertle: .6851
Poloquebec: .7460
BlackDog: .7773
Waldo: .7592
Any stat that has me behind Yertle must be fatally flawed.
asm - I don't like being at the bottom of the list and almost throw out the post as a result.
mongrel - My incorrect calculations initially had me at the top of the list which is why I was in favor of it.
M57 - Two more stats
BlackDog: .7773
Waldo: .7592
Norseman: .7586
Poloquebec: .7460
Mongrel: .7190
Hugh: .7073
Yertle: .6851
ASM: .6771
M57: .6714
Alpha: .6052
I am still at the bottom.
Multiply by two and call it a good stat.
Alpha wrote: Multiply by two and call it a good stat.
This was also my instinctual answer to the 'how to tell the difference between a good and bad score' question.
So what's the score for a completely average player? Is it still 1 as per G-rating?
Also what's the actual calculation for this Hugh / Alpha?
For each player there is an array of game sizes : number of wins at this game size to work with, is this sufficient?
Completely average would be .5
Win them all = 1
What you have is sufficient to calculate. Hugh' post #11 in this thread gives how, basically:
Sigma [ ((players per game - 1) * wins) / (((players per game - 1) * wins) + losses)]
If you want to give more spread to the scores you could multiply by something. 2 is OK. I think 10,000 might be better, then you could drop the decimal places and it would seem like more of a difference between 7773 and 7592.
Ok now I'm confused... so how come Alpha's calcs give people with <1 if winning them all gets you a rating of 1?
I think a rating which is higher if you're a better player is preferable - can we just invert it using 1/x? I also like the idea of a 10,000 multiplier.
I agree with tom it should be inverted using 1/x. Also I do not feel like I better than the 8 people below me, if anything I don't deserve top ten I just get lucky and win HUGE games.
With that said we cannot really rank people who play different style games in the same category.
I agree with tom it should be inverted using 1/x. Also I do not feel like I better than the 8 people below me, if anything I don't deserve top ten I just get lucky and win HUGE games.
With that said we cannot really rank people who play different style games in the same category.
1/(1-x)
That looks like it could be one of those silly faces that Vataro's a fan of.
>:(
1/(1-x)
Example Calculation:
(mine above did not include the multiply by two. If this is done then the stat will be centered at 1 like g-rating:
<1 less than average
1 average
>1 greater than average)
Alpha:
2 Player Games 77 wins 140 games
3 Player Games 6 wins 19 games
4 Player Games 36 wins 107 games
5 Player Games 16 wins 43 games
6 Player Games 7 wins 27 games
7 Player Games 1 win 10 games
8 Player Games 2 wins 7 games
9 Player Games 0 wins 3 games
10 Player Games 1 win 4 games
sum wins times game size - 1 =
77(2-1) + 6(3-1) + 36(4-1) + 16(5-1) + 7(6-1) + 1(7-1) + 2(8-1) + 0(9-1) + 1(10-1) = 325
This is the numerator.
sum number of games losed
63 + 13 + 71 + 27 + 20 + 9 + 5 + 3 + 3 = 214
Hugh's Rating = (325) / (325+214) = 325 / 539 = .6030
My modification (multiply by 2)
rating: 1.2059
Hopefully there are no error above, will check later.
(X-3)
I like x2 rather than x2000. This H or G rating (whatever we call it) would pretty seamlessly replace the old one. Most people wouldn't bat an eye.
Apart from a low score is better than a high score which is the inverse.
In my originally suggested calculation a high score is better than a low score. It is a ratio - multiply by 100 and you get a winning percentage normalized to the 2-player case.
tom wrote: Ok now I'm confused... so how come Alpha's calcs give people with <1 if winning them all gets you a rating of 1?
I think a rating which is higher if you're a better player is preferable - can we just invert it using 1/x? I also like the idea of a 10,000 multiplier.
Ah, this is what led to the inversion posts :) Winning them all gets you a rating of 1.0, or 100%. None of us win them all, so our scores are lower than 100% ;) Still, truly truly, higher percentages are better than lower percentages.