yes the player could make 99,999 losing bets in a row but as it is a bet of 1 in ~30k he could still win the outer with his first few bets or even more than once in his first few bets. why should a casino take such a high and risky gamble?
Because it maximises the expected growth of the logarithm of their bankroll.
IMO a casino should never take such a high risk gamble with a possibility to get wiped out in one bet.
And they never do. The only time you risk 100% of your bankroll on a single event is if the house edge is 100% - and in that case the player is guaranteed to lose.
as I posted before I don't know any land based or online casino beside MP that is taking such a dangerous risk.
Then there are lots of places leaving money on the table.
to me it looks like that this high risk KC handling is only for plinko or does MP have other games with such high risk payouts? maybe the previous MP owner could answer this question.
If somebody wants to bet with a 99% house edge against them on a dice game, MP will happily risk 99% of its bankroll on that bet.
An example of such a bet: you roll a number in the range 0.00 to 99.99. If it's less than 5.00 you double your money. Otherwise you lose.
You have a 1 in 20 chance of doubling your money, and a 19 in 20 chance of losing it.
Those are horrible odds for the player, and great odds for the house. The house has a 90% house edge.
What percentage of your bankroll would you risk taking such a bet, if you were the house?
The correct "Kelly" amount is to risk 90% of your bankroll, since the house edge is 90%. (RTP = 5 * 2 = 10%)
I ran a simulation, having the house risk different percentages of its bankroll from 0% to 99% while the player max-bets against them 1000 times in a row. Here's a chart of the average of the log of the house bankroll for each percentage risked:
Notice how the house does best when it risks 90% of its bankroll per bet? That's the point.
Edit: I re-ran the simulation, but for a 1% house edge bet. The variance is a lot higher, due to the smaller house edge, but the basic features of the curve are clear enough: risking 1% is optimal. risking 2% gives you a zero expected growth of log(bankroll), and higher than 2% risk is actually bad for business:
