Deal or No Deal

[ouyangl]

Has anyone watched this show before??

To me, it seems like it really doesn't matter what case you choose, but how much money you can get for a bank offer. I tried finding a pattern between the banks offer and the expected value of the prizes but there seems to be no correlation. Well there is in the first few rounds but after that nothing seems to make sense. Also the models for the first few rounds that I came up with don't match with the models for other contestents.

To me, it seems like if you can predict the bank offer, you can thus calculate EV's on Deals or No deals.

Lets say you have 3 prizes left, $100, $10,000, and $100,0000. The bank is offering $30,000. 2/3 of the time you'll pick a case that increases the bank offer while 1/3 of the time you'll decrease the bank offer. But what I really need to know though is how much will the bank offer change for each case I choose.

Any thoughts?

[Kerth]

Use the search.

[Moneyfisher]

i did watch the show, and i didn't follow it closely, but in later offers it's quite obvious to me that the offer is roughly the average of the cases on the board. it is sometimes also adjusted up or down a small percentage based on like you said, the number of cases above or below the offer. i think it's pretty straight forward probability really

[KenProspero]

actually, it's not straight forward at all.

Let's say there are three cases left -- one with $1million and two with crap. Now, if you were Bill Gates, it would be clear, you keep going for any offer less than $333,333 ish.

However, most of us aren't Bill Gates. Say you're offered $150,000 (which will pay off the mortgage, pay off back debts or otherwise generally make a difference in your life).

Even though $150k is -ev if we're counting dollars. We're talking about a one in three chance in hitting the jackpot, vs a 100% chance of really changing your life. Plus the fact that unlike poker, you don't get to replay the situation endless times.

The decision, in other words revolves around much more than straight $ ev.

Aside from that psychological aspect, the game is both straight-forward and assinine

[Nottom]

The show is pretty straight-foward really. They usually offer you a deal worth somewhere between 50-60% of the true value. As you get down to less suitcases the offers tend to get closer to the actual value but usually never exceed 80-90%.

If you want to see what sort of things can happen I'd suggest playing around with the web game.

[drbst]

Everybody must of course decline all the "offers" with a lower EV. What is the idea of that retarded show, to prove that people don't know what an EV is? [img]/images/graemlins/laugh.gif[/img]

[DoomSlice]

Or to prove that risk premiums and diminishing marginal utility do exist.

[SossMan]

[ QUOTE ]
Everybody must of course decline all the "offers" with a lower EV. What is the idea of that retarded show, to prove that people don't know what an EV is? [img]/images/graemlins/laugh.gif[/img]

[/ QUOTE ]

I know what EV is. Sometimes, for lifechanging money and the inability to run the situation multiple times, I would be willing to sacrafice value for reduced variance.

Extreme examples always help:

case 1: 100 bajillion trillion gagillion dollars
case 2: 100 million
case 3: a free item from the McDonalds value menu

Bank Offer: 90 million

I take the 90 million even though the EV is somewhere in the 80 bajillion trillion range.

[ouyangl]

I dont think the EV of what the case actually has really matters. Because lets say you have 3 offers left:
$1,000,000
$500,000
$1,000

The average is $500,333.33
If you get an offer of $400,000, your getting a negative EV.

But thats only true if when you decline the offer, you randomly select a case and take it home with you.

What I really think matters is being able to predict the bank's offer. I've tried to relate it to the average of the cases and its hard to find one model that fits well to it. Like I said, in the early rounds it may fit to a logarithmic model or power model, and then go completely bezerk after a certain round.

If you could predict the offer for each given case you select, making a decision would be a lot easier.

Lets say the formula for the offer is offer=4/5avg.

Same 3 cases left so the bank's offer is $400,266.66

If you eliminate the case containing $1,000, the average shoots up to $750,000 and the offer increases by $199,733.34 to $600,000

If you eliminate the case containing $500,000, the average increases by $133.34 to $400,400

If you eliminate the case containing $1,000,000, the average declines by $199,866.66 to $200,400.

Thus the real EV of whether of the decision is
1/3(-199,866.66)+1/3(133.34)+1/3(199,733.34)=0.00667

Giving up an expectation of 2/3rds of a cent and taking the money right now would be my choice.

So you see what I mean? I don't really think the average of the cases matter at all. What really matters to me is the offer

[ouyangl]

Thanks Nottom,

its actually pretty entertaining. I'm reading a biography of John Nash and this reminds me of the games he use to study.

Do you by chance know how to get into the flash file and figure out how the bank calculates the offer?