EinStein Ratings Einstein forum
8 replies. Last post: 2006-04-08
Reply to this topic Return to forum-
Ingo Althofer at 2006-04-02
New EinStein players on LittleGolem start
with a neutral rating of 1500. Currently
the best players are slightly/somewhat
above 1700, and the bottom is about 1300+.
This is well in accordance with EinStein
ratings on the other online server
http://www.inetplay.de, where
start rating is 2000 (instead of 1500 on LG)
and stable ratings are between 2200+ and
about 1900.
It is clear that in a game with a rather luck
factor ratings cannot climb (and fall) arbitrarily.
Ingo Althofer.
PS: Currently, the ratings on LG are more volatile
than on inetplay because of a larger influence factor
for the single results. -
Theo van der Storm at 2006-04-05
Summarising Ingo Althöfer,
if I get his spelling with o-umlaut right:
No surprises in the ratings and games of (some) chance have a limited rating scale.
I wish to replace his clarity (1) with mine (2):
1. IMHO at the low end of the scale there is plenty of room
to fall arbitrarily low, so your clarity is not mine.
I don’t want to consider all-knowing worst play.
At the top end I agree:
Surely an all-knowing entity could not beat Opmp by more than 77.7% :-)
2. In a competition with free (i.e. player determined) pairing (e.g. InetPlay)
the scale of EWN ratings can be stretched significantly, so that the rating
difference does not reflect the win-chance by rating system design of two
arbitrary players in a structural way. On LG this is only a very minor
problem.
It can be prevented by an automatic pairing system, which doesn’t show
the people registering for a game or tournament. For accuracy purposes
it could block people registering with a very big rating difference to
those already registered.
And then there are the rating tournaments to fix the damage.
By the way. Is anybody reading this?
Theo -
Theo van der Storm at 2006-04-05
Here is an exercise for a maths professor:
Based on his playing strength player A has chance 60% to beat player B
in a single point game.
What is player A’s chance to beat B in a best-of-5 (3 point) game?
... and in the funny situation, that this chance is bigger than 60%,
wouldn’t that have an effect on the rating scale :-)?
Actually, I’m quite sure he will be passing it on to a student.
I’m a fan of Deep Thought in more than one meaning. -
richyfourtytwo ★ at 2006-04-05
Here’s the math if I’m not mistake (and have typedcorrectlyinto the calculator).
0,6^5 + 5*0,6^4*0,4 + 10*0,6^3*0,4^2 = 0,68256
So yes, as one would intuitively expect the chances are better than 60%. Hence best of 5 ratings should give a rating distribution with slightly greater width.
Cheers
richyfourtytwo -
Theo van der Storm at 2006-04-05
You have calculated the chance that the stronger player makes 5, 4 or 3 points out of 5 games. The game stops after 3 won games!
I calculate the chance the stronger player makes 3 points in 3, 4 or 5 games
Here’s my math if I’m not mistaken:
0,6^3 + 3*0,6^3*0,4 + 6*0,6^3*0,4^2 = 0,68256 (exact number)
Of course. The same result, but fitting reality more closely :-)
52% -> 53,7%
54% -> 57,5%
56% -> 61,1%
58% -> 64,7%
60% -> 68,3%
65% -> 76,5%
70% -> 83,7%
76% -> 90,7%
85% -> 97,3%
Proost,
Theo
PS: 3 permutations to win in 4 (6 to win in 5),
because the final win is a “given”.