When you name the first actor in Typecast, the day is won. Then the game shows you the second actor's board and asks a question: take your points and stop, or carry on and risk them. Carry on and fail, and you go bust. The win stands but the points go to nothing.
I've got 38 days of people answering that question, from 13 October to 19 November. They nearly all give the same answer.
(This names actors from past rounds, so skip it if you're catching up.)
Of the players who name the first actor, 85% take on the second. Only 15% bank.
Of the ones who take it on, 79% name the second actor and 21% go bust.
Then the same question comes round again for the third, and people are even keener: 90% carry on and 10% bank. This time 78.5% get there and 21.5% go bust.
So the bet comes off about four times in five, both times, and people take it nearly every time it's offered.
Follow everyone who names the first actor through to the end of the day and it comes out like this:
Nearly a third of the people who'd won the day finish it on zero. That's the price of it, and it's a steep one. Of the players who take on the second actor, 56% end the day with all three.
The part I didn't expect is this. You get to look at the next board before you decide, so you'd think more people would bank on the days the next actor is a hard one.
They don't. There's almost no link between how many people bank on a day and how many of the rest go bust.
On 24 October the second actor was Thomas Kretschmann. 14% of players banked, which is a normal day, and of the ones who carried on, 52% went bust. That's the worst the second actor has ever been, and the banking figure gives no sign that anyone saw it coming.
On 29 October the second actor was Michael J. Fox. A quarter of players banked, the most on any day. Of the ones who carried on, only 17% went bust.
The share who bank after the first actor has never been lower than 8% or higher than 25%, whatever's on the board. It looks like something people decide about themselves, not about the round.
The third actor is meant to be the hard one, and it pays double: 500 against 250. But the bust rate is the same, 21.5% against 21%.
That's probably because the people who get that far are the ones who are good at this. It still means that if you've just named the second actor without much trouble, the numbers say the third is as fair a bet as the second was.
It swings a lot by day, though. On the worst days 35% of the players who tried the third actor went bust. On 31 October, when it was Neve Campbell, hardly anyone did.
Look at the board before you press the button. That's what the pause is for, and the figures suggest most people aren't using it.
If the rows are films you half-recognise, carry on: four in five do get there. If you look at it and nothing's familiar, that's the day to be one of the 15%.