General > Rankings, meeples and achievements

Measuring skill/luck ratio in board games

(1/3) > >>

Jimmy V.:
Hi all,

I have just published a short study about measuring skill versus luck in games on Reddit.

The study uses stats coming from Happy Meeple platform so you will certainly find it interesting.

An up vote on the Reddit thread is appreciated.

Feel free to react there or here!

Gues(s) T.:
Very interesting topic and very interestic article. Worth a second, slower reading (plus commentaries). Thanks a lot for posting this hint.

Mike M.:
Indeed very interesting.  I have thought on this topic quite a bit and have my own opinions (I call them this because I have very little evidence to support them) which are much too lengthy for me to post today. 

Great read!

Mike A.:
Cool post, though I must admit I don't understand much of it. 😛
Here's a question--at one point deep into the thread you said you can think of two reasons why a coefficient (or some term) is used to allow quick movement in ratings. What are those? I understand why that would be the case for games like chess where no luck is involved. An inferior player should rocket upwards when they beat a superior. For games like Lost Cities, however, it seems like the gain should be much, much smaller.

Jimmy V.:
Hi Mike A.,

I said on the Reddit original post that I could see at least two good reasons for having fast evolving ratings (big K coefficient rather than small). And later I explained more or less my thoughts.

Basically, noone wants new players to be discouraged. So their rating needs to evolve quickly if they are somewhat good. Actually, we use an even bigger K for new players to make this even truer. The other reason why I think a (too) big K is better than a (too) small K is that it brings more variety, more fun to the game. Even if rating variations are sometimes a bit too harsh or a bit too positive, at least they give life to your experience on the platform. Highs and lows are good for fun I think, as long as it is not going too far.

The ideal coefficient K is one that optimizes the prediction value of the rating. In other words, the best K is the one that leads to a rating that can predict future outcomes the best. If K is too small or too big, players ratings are less accurate, either because they don't reflect players progress fast enough or because they take latest results too much into consideration. In Chess for instance, I saw a document explaining that the ideal K was 24 when the one they used was 15 or 20.

I hope this answers your questions. Rating systems are a fascinating world.

PS: the FIFA (football) ratings are absolutely awful. Instead of using a ELO system or something similar they devised a completely stupid independent system that has so many flaws it is hard to believe.

Navigation

[0] Message Index

[#] Next page