Pseudo Complete 1v1 Stats: Stat Site Evolving. 98% Complete

[rus|Mike]

I'll be responsible for monthly updates of DEMO games

[Ace Rimmer]

By the way, is the game list up to date? Or does it only include recent games from the duel server where the volunteers played?

Shouldn't we devise some method to ensure the list was kept up to date? Like every member being responsible for a week or something?

Isn't there a better way? I always like to have things automated. Couldn't Pox's site run a check against the duel server listings and search out games that are "missing" from the game log? Then, we could sort out how to delegate those remaining games.

[rus|Mike]

Gee.

[rus|Mike]

I'm happy to hereby report that I finished logging in DEMO games and made a total of 200 games worth of data entered

I'm going to do monthly updates, every last date of each month in a worst case. In the best- each week or so

I'm now going to log in ASW tournament games.

EDIT:
Pox, we definatelly need measures against (MOR) vs White Rabbit matches bumping (MOR)'s rating 47:4 wtf??! I think I'll stand with my previous suggestion if anyone remembers it.

EDIT2:
Ace, got links of another ASW games?

[Ace Rimmer]

Nope, I tried to find more but unless somebody directly asks xander, I think those are all of them.

As far as (MOR) vs White Rabbit, who was it again that put all those games in there in the first place? My suggestion is to separate Demo from Duel. Skim the cream off the top, that's all I care about anyway.

Besides, I think those games were "training" games anyway. So they skew the stats in two ways, not just one. If nothing else, we can all just pick a player and play that person 50+ times to "remove" (MOR)'s artificial win ratio.

[rus|Mike]

who was it again that put all those games in there in the first place?

I can void all of them besides those won by WR anytime

[Ace Rimmer]

who was it again that put all those games in there in the first place?

I can void all of them besides those won by WR anytime

That wouldn't do any good, then WR would have an artificially inflated win ratio.

[rus|Mike]

Right ASW is entered. And I'm bored. Give some part of Duels Anyone?

[Ace Rimmer]

You can have the rest of mine if you want them. I stopped on the 19th, so Aug 20- on (for the rest of my assigned games, see first post) are available if you want them.

[rus|Mike]

I'll start with them tomorrow evening and prolly will finish tomorrow evening also

[Xocrates]

In the mean time I would like to make yet another suggestion: A predict-o-mat

Basically you pick two players and two continents and it will tell you what is the likely result.

I'm still trying to figure out the math to do that though

[rus|Mike]

Ok I'm done with Ace's games (Aug 20-31, Jan 1-7, Feb 8-14)

[Xocrates]

I am an Idiot! After days trying to figure out a way to calculate some pseudo win probabilities through inordinately complex formulae involving corrective factors and even enzyme speed (don't ask), I've remembered an easy way that involved calculate the square root of a squared number (because I wanted to avoid the use of functions like SQRT or ABS).

Without further delay, I present you the mathematical method that allows you to predict the outcome of a game (divided into several parts for easier comprehension):

W=CWR+X+Y

X=(1-CWR)*DW*F1

F1=(1+DW/((DW^2)^0.5+10E-12))/2

Y=CWR*DW*F2

F2=(1-DW/((DW^2)^0.5+10E-12))/2


With:

W: chances of the player to win

CWR: Continent win ratio for the selected setup

DW: difference between the two players overall win ratios

Example:

Ace Vs Mike on NA Vs Russia

Ace:

CWR: 0.49

DW: 0.72-0.74 = -0.02

W: 0.40 or 40% chances

Mike:

CWR: 0.51

DW: 0.74-0.72 = 0.02

W: 0.60 or 60% chances


Additionally, you can use it to see the overall chance of a player to be another player by setting CWR to 0.5. So, taking rus|Mike) vs Hyperion for instances we get.

W rus|Mike: 57% (real 58%)

W Hyperion: 43% (real 42%)

Which seems pretty accurate.


And while I'm at it, I would like to suggest a small change to the corrected win ratio formula:

win/played*(1-1/played^(1.2+played/100)

That way it will compensate faster in higher game counts so we don't get so many players losing 1% or 2% despite having over 50 games.

[rus|Mike]

Jesuis.

[Xocrates]

Jesuis.

Just be glad the other ones fell through.

Really glad