This assumes that people operate only on expected value and don't care about variance at all, which is incorrect empirically, theoretically, and intuitively. The premium you pay for reducing variance is called a "risk premium", and it follows pretty trivially from the diminishing marginal utility of money (though there are many other ways to derive/explain it). This is the entire reason that insurance exists as a product: it's not designed solely to cover existential risks or to prey on the irrational.

Well, if you pay me more then $0.05 for the $0.6/-$0.5 coin flip, I will happily sell you some. How much do you want?

I think GP disagreed with your assumption that a $1.1/-$1 coin flip was worth $0.05, so they'd pay more for the $0.6/-$0.5 coin flip than for the $1.1/-$1 flip, but still less than $0.05.

If that is true, then I will happily buy $1.1/-$1 flips for $0.049 from them!

They would pay a premium to avoid uncertainty, so they wouldn't accept that offer either. The uncertainty is the same whether you sell or buy the flips. I don't know you, but I guess you wouldn't pay $499K for a $1M/$0 flip, and you wouldn't sell such a flip for $499K either. This is the same phenomenon on a much smaller scale.

    I guess you wouldn't pay $499K for a $1M/$0 flip

No. But the discussion here is about shares. And they are not $499k a piece.

We are discussing if the share price of a company that publishes numbers in line with expectations should go up.

What do you mean by “expectations”? If Amazon trades a $2000, is the expectation that it will stay at $2000 forever? Or is the expectation that will go up because it will surpass expectations?

If by expectations we mean the numbers published by sell-side analysts they may or may not be close to the actual market expectations.

Say you can pay $10k and get (50/50) either $0 or $22k. The expected gain is $1k.

A second game also costs $10k to enter, but the (50/50) payouts are $9k and $13k.

Do you find then equally attractive? Most people wouldn’t.

If you do, what if you could instead “play” a game where you pay $10,000 and get $10,999 straight away? It’s a worse proposition if you’re risk-neutral, isn’t it?

And while it’s true that if markets were efficient you wouldn’t be compensated for diversifiable risk it’s also true that a) markets are not efficient and b) uncertainty in the results of companies is correlated with the market and not completely diversifiable.