With two twins ("A") of strength $a$ and a singleton ("B") of strength $b=1-a$, we race them and have three possible order statistics: AAB, ABA, BAA. These have maximum likelihood estimates for $a$ of 1, $\sqrt{2}-1$ and 0 respectively. We can work out lots of interesting things such as the bias of the ML estimator and the MSE.
With two twins ("A") of strength $a$ and a singleton ("B") of strength $b=1-a$, we race them and have three possible order statistics: AAB, ABA, BAA. These have maximum likelihood estimates for $a$ of 1, $\sqrt{2}-1$ and 0 respectively. We can work out lots of interesting things such as the bias of the ML estimator and the MSE.