"For those who are wondering, I convinced my interviewees that the bet was not a scam: they could inspect the coin, flip it themselves, use their own coin etc. I explained that the experiment was intended to explore their approach to risk. It was fear of losing $10, not distrust, that led them to decline the bet."

And here I was just about to point out that any bloke on the street offering me a similar bet is a con artist by default. Sometimes it is good to check YT comments first.

I wonder how lack of truly understanding the math comes into play.

It also needs to be factored in the nature of the game.

A coin flip isn't a) very entertaining nor b) skill based.

That's why games at casinos like poker and blackjack are quite popular, because they have either skill or the illusion of skill involved while also being entertaining.

That's how you get someone to agree to a bet like this - added value of the event itself.

He misses out on cost of opportunity. If it was an open standing offer it would be different than just needing to choose to possibly absorb the loss that moment.

Video has 1m+ views on youtube... looks like it worked out for him.

Leaving aside the mechanics of the coin toss; for me, it's about the level of risk. I'd absolutely risk $10 on a whim because I can afford to lose it. Risking my car or my house would be a completely different proposition.

I don't get how it's not obvious to people that over a large sample size this is a good deal. Isn't it just intuitive?

It is. The thing is can you afford the risk. $10 most people can afford the risk without going homeless, but maybe they would have to skip pizza that night and make KD. Just increase the bet, If you could bet your car, today against somebody's porsche in a coin toss, it's a great bet. You also could very reasonably choose not to because the loss of your car is a greater disadvantage to you than gaining a new porsche. It's cost of opportunity, and for a wealthy person, risking a cheap car for a much better one at 50/50 is a bet they can afford to take. For a pizza delivery driver supporting their family it's a choice between maybe coming home with a porsche or coming home without a car, without a job and no means to buy a new car any time soon either.

Your car example has a sample of 1 though, I'm specifically talking about large samples.

I think you're missing the point, when you increase the sample size of a favorable bet, eventually the probability of you losing money in the series of bets becomes negligible.

Take the video's example of say risking a coin toss of $10 for a potential $15.

You only need to win 40% of the time to break even ($15 * 0.4 - $10 * 0.6 = 0).

To work out the probability of this for different sample sizes we just look at the cumulative binomial distribution:

For 10 samples, 40% success or more occurs 82.8% of the time.
For 100 samples, 40% success or more occurs 98.2% of the time.
For 1000 samples, 40% success or more approximates 100% of the time.

If you want to work it out yourself or visualize it, you can use a tool like this:

http://homepage.stat.uiowa.edu/~mbognar/applets/bin.html

If you can take the same odds more than once yes, that changes it drastically. If the other party will just keep going and you can afford to lose a couple times and keep going then it's just free money for you. Casinos run an exactly the same math.