How random is Einstein W.N.?

Hi, just for fun I decided to look at the game record of the strongest EWN player and see how he does.

I looked at 444 games of its. 431 of those are three point matches, so you can basically interpret this as the win rate for 3 point matches.

I made this graph

This shows the win rate of that player, stratified by the strengths of its opponents.

If people want it, here's the

table and the R code .

Niall :)

P.S. Apologies if my attempt at hypertext goes all wrong! I can't see a 'preview' button :)

I'm permantly controlling the stats of the best 60-80 players on LG. I have that curve for all - but a bit more fine - I'm separating the stats in steps of 10 points. For some players with a low number of played games it's not so precise but for others very interesting. If you would be interested in the curves/stats please send me a private message with your e-mail adress. You would need excel to view the stats.

As you suspected, the links above don't work.

Here are the first two:

http://wittelab.ucsf.edu/~niall/personal/topEWNplayer_ratingStratifiedWinRates.pdf

http://wittelab.ucsf.edu/~niall/personal/topEWNplayerResults.txt

I can't figure out what THE R CODE link is. Nothing I try works.

Ahh rats.

The other one is:

http://wittelab.ucsf.edu/~niall/personal/lookAtEWNwins.r

(previously there was a spurious space)

(the code is not well written, but just in case it might be useful)

YHW - that sounds interesting. I think it'd be nice to have the data in flat-text files, though perhaps the Open Office I have here can read your files. I'll send you a message.

I will look up how to do hypertext properly.

> I can't figure out what THE R CODE link is. Nothing I try works.

The R code is a program written for R.

http://www.r-project.org/

A very nice and free environment for statistical analysis, although not very easy to start with.

Yes, thanks. I am aware of what R is, but Niall's original link he provided didn't match the file location. That's why I said I couldn't figure out what his intended LINK was.

When you are interested in a rather abstract view

on randomness of a game and of Ewn in special,

you might find the doctoral dissertation of Jakob

Erdmann to be of interest. It has the title

"Chanciness: Towards a Characterization of Chance in Games"

and can be downloaded from this site on

http://www.althofer.de/3-hirn-grant--erdmann.html

Jakob Erdmann.

In its experimental part you find a study on Ewn and

some Ewn variants.

Ingo.

There's a beautiful paradox - if there's any chance at all in the game, once you've got 2 *perfect* players, they game becomes one entirely of chance!

Well, if there is no chance at all in a game, once you've got 2 *perfect* players the game becomes one entirely of boredom! :-)

@FatPhil: If I think about it, this is exactly my impression in EWN. If two weak players play EWN the result depends more on good or bad moves and less on chance. If two strong players play EWN the result is much more influenced by luck.

I have to read Jakob's thesis more detailed. Up to now it is not clear to me, how a deterministic player fits in the concept of "chanciness".

Understanding how a deterministic player fits in is easier when you think about

a special case of deterministic players: the perfect player (actually a perfect

player need not be deterministic. She may chose at random between equally

perfect moves).

You may think of a non-perfect deterministic player as one who only thinks she's perfect and thus never reconsiders her moves. For her the game is pure boredom as well. She can only wait and see how chance or the antics of her co-players turn out.

However, it's still a border case since actual humans are seldomly deterministic

at the games they play. After all, why should they play in that case?