Better than flipping a coin?

From Round 6 to Round 10, the Bradley-Terry model correctly tipped 27 of the 40 games played. Is this any better than simply flipping a coin to decide who wins or loses?

While it is certainly possible to get 27 (or more) games correct out of the 40 games just by flipping a coin, the chance of this happening is approximately 1.9%, or 2 in 100.

In other words, if the 40 games between Rounds 6 and 10 were to be re-played over and over again 100 times, then on 2 of these occasions a coin would be able to correctly predict as many as the 27 winners predicted by the Bradley-Terry model. There is therefore moderate evidence that the Bradley-Terry model is in fact doing better than just flipping a coin.

The above probability of 1.9% is an example of a p-value which is often used in hypothesis testing to make decisions.