Given a special case of WVG (Weighted Voting Game) of $a$ 1s and $b$ 2s and a quota q, $ [q:1,1,1,1..1,2,2,..2] $. I need help with calculating the Shapley value of a player with a weight of $2$ and a player with a weight of $1$ as a function of $a$,$b$ and $q$.

I know how to calculate the Shapely value in general, but I would like to get a simple closed form for the result. I tried to think about what happens if there are only players with weights of $1$ (simple scenario $ a=2k-1,b=0,q=k $ I think you will agree that $ \phi_{1}(v)=\frac{1}{a}=\frac{1}{2k-1} $. the simple calculation is $ \binom{2k-2}{k-1}(k-1)!(k-1)!=\frac{1}{2k-1}=\frac{1}{a} $ which is $ \frac{1}{2k-1} $. The calculation for a player of weight 1 (when there are 1s and 2s) would be a sum of choosing i players of weight 1 and $ \frac{q-i}{2} $ players of weight 2. similar calculation for a player of weight 2.

I tagged binomial coefficients because I thought they could be useful for counting the occurrences.

Thanks in advance, Mati

also posted here: https://mathoverflow.net/questions/107507/calculating-the-shapely-value-in-a-weighted-voting-game

  • $\begingroup$ Joriki, please refer to the edit in my question. Thanks. $\endgroup$ – Mati Bot Sep 19 '12 at 14:06
  • $\begingroup$ Thanks for the quick response. I've removed my comments and downvote (and in fact upvoted). I doubt that there's a nice closed form, though. The result will depend on the quota if the quota is close to $0$ or $a+2b$, but for small $b$ there will be a plateau of intermediate quota where the result doesn't change with the quota. $\endgroup$ – joriki Sep 19 '12 at 14:23

Note that the Shapley value for this game is the probability that a player is pivotal in a random ordering. Now the probability that a 1 player is pivotal is the probability that the sum is q-1 when she is reached. While for a 2 player it is the probability that it is q-1 or q-2. This should simplify the analysis, but I haven't worked out the details.

However a simple conclusion is that the Shapley value for 1 players is approximately 1/2 that for 2 players for reasonable size values and away from extremes, so the values are approximately 1/(a+2b) and 2/(a+2b) respectively. In particular this holds in the limit where a,b go to infinity and q=y(a+2b) for 05 and 2/a


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.