Originally Posted by B_OR:
“What this model shows, extremely well, is someone like Ed at the bottom will be facing incredible odds every week just to survive which is the result of so comparitively few permutations allowing him to survive. Those number of permutations (as a percentage) which will give him the boot will increase week by week to a ridiculous level in the semi when he will be thrown off the show in 5 of the 6 occassions when he is still very popular with the public when he is second with the public. He will even be given the boot in 1 in 6 occassions when he is first with the public. That is wrong.
From that point of view this model has been very accurate this year, and for many years previously.”
Your model shows that these people should go out by the semi-final, agreed, however that's not the problem. Your model shows these people should be out by week 6 when they actually last until week 10!
Each year, the 'survivor' has comparatively few permutations allowing him/her to survive, according to your model. The number of permutations (percentage) decrease week by week. Yet they regularly survive for many weeks; Ed Balls, Anne Widdicombe, John Sergeant, and so on. Your model is inconsistent with this historical results.
The reason why they survive is that they do not have an entirely random vote each week. They necessarily have a strong vote to avoid the dance off and that vote often carries over the following week as well, so they avoid the dance off again. The exact vote is unknown but you are still discarding historical voting information in your model.
