data(doubles)There are four players, $p1$ to $p4$. These players play doubles tennis matches with the following results:
| match |
| score |
| ${p1,p2}$ vs ${p3,p4}$ |
| 9-2 |
| ${p1,p3}$ vs ${p2,p4}$ |
| 4-4 |
| ${p1,p4}$ vs ${p2,p3}$ |
| 6-7 |
| ${p1}$ vs ${p3}$ |
| 10-14 |
| ${p2}$ vs ${p3}$ |
| 12-14 |
| ${p1}$ vs ${p4}$ |
| 10-14 |
| ${p2}$ vs ${p4}$ |
| 11-10 |
| ${p3}$ vs ${p4}$ |
| 13-13 |
It is suspected that $p1$ and $p2$ have some form of team cohesion and play better when paired than when either solo or with other players. As the scores show, each player and, apart from p1-p2, each doubles partnership, is of approximately the same strength.
Dataset doubles_noghost gives the appropriate likelihood function
for the players' strengths; and dataset doubles gives the
appropriate likelihood function if the extra strength due to team
cohesion of ${p1,p2}$ is represented by a
ghost player.