Note: Updated DI and VF numbers
The methodology for the semi-finals model is described here. The model is 90% accurate on ranking, with a margin of error of about 3 percentage points. Since the mid-line probability is about 58%, the model cannot decide between Charlie Askew and Paul Jolley. It is reasonably sure that Vince, Nick, Elijah, and Cortez will not advance, and it is fairly certain Lazaro, Burnell, Devin, and Curtis will advance.
The model finds the most probable Top 5 men are Lazaro, Burnell, Devin, Curtis, and
Charlie Paul (by a hair’s breadth). (Updated)
Nick, Elijah, and Cortez are below the probability where nobody has ever advanced. That being said, probabilities being what they are, such an event should eventually happen; In my estimation, those three have very little chance. Vince Powell is an edge case. His advancement would not be shocking at all, but he would be the fifth least likely advancement in history. Paul MacDonald and Karen Rodriguez, both in Season 10, had probabilities that low, and advanced.
It may be a bit surprising that Charlie Askew is even in the uncertain category. His highly inadvisable performance garnered a 9 out of 100 on WhatNotToSing, but Dialidol shows people voting for him, and Votefair indicates that he was popular coming into the semi-final rounds. The model is skeptical of those two indicators this early in the contest, but it does not totally discount them.
If you asked me which of these calls I’m most worried about, it would have to be Elijah Liu. The praise he got from the judges, his possible appeal to Asian Americans, I feel like the model might not be picking up on something. Perhaps Dialidol will show an effect as the west coast begins to vote. I will update in the morning.
As a technical note, for anyone interested, it may seem a bit strange that more than 5 people have a greater than 50% chance of advancing, when clearly only half of them can go through. This is partially an artifact of the normalization process (the sum of all probabilities in this case should equal 5). It’s also due to the incredibly low probability of the 3 lowest-ranked contestants. Comparing all years, Paul Jolley would have been likely to advance relative to everyone else. If he did advance this year, it would not be strange at all. He’s simply not quite as likely as the others.