Hacker News Reader

Comment by ricardobeat | Hacker News Reader

I'm struggling to see how the advice applies to their own examples. Take Jane, who took a great offer but was hit by a recession:

- potential reward of the decision: checks out

- probability of getting it: 95% if you already have the offer in hand

- resources (time/effort/money) to bet to get the reward: not significant

Then there is 'imagining a bet' and 'analysing your past failures'. The first one might lead you to the 'what if the company goes bust?' question, but would that really help in the decision making or be a reason to not accept the better job? The unlucky event is a recession, not this particular company doing badly (i.e. no signals to see).

If a recession hits, there is no reason to believe the consumer goods co. would fare any better and her old job more secure. I'd really like to see the 'poker player' mindset at play but this just feels like a complete miss.

View on HN

I think like a poker player, I skimmed the article. What they seem to fail to mention is that variance is important as well. In poker you need to play at least 10000 hands in order for variance to average down a bit. 10000 hands is still very little.

So, if you have a 95% chance of going 10x that's great, but the question is: can you handle a 5% chance of having being potentially ruined for at least a decade? If not, then you cannot take the bet despite the fact that EV is super high.

Put more extremely, if I offer you a 99% chance to take make one billion dollars (legally) and a 1% chance of being killed, would you take it? I definitely wouldn't.

How about the following? I offer you an unfair coin flip advantage of 51% versus 49%, we're going to flip 1 million times, each time the bet is 1 dollar. Would you take it? I would.

The heuristic that poker teaches is: take a lot of small bets with huge upside with little to no downside. It doesn't matter how slim the chance of winning is. Slightly more nuanced (still a heuristic): the chance should be about as big as your volume of chance-taking. Say, you take 10000 chances, then the slimmest chance you can take (on average) is 1/10000.

Congrats, that's also thinking like a poker player.

https://en.m.wikipedia.org/wiki/Kelly_criterion

This is off-topic, but the mobile wikipedia link you posted is surprisingly nice to read on desktop. Seems a lot more pleasant than the wide lines on the desktop site.

This comment comes up repeatedly on HN. While I somewhat agree, I would also suggest to not maximize your browser window by default, to avoid wide lines. I usually have my browser window square-ish.

I love the Kelly criterion! I should reread it in order to see if I can make a better heuristic on optimal bet-sizing.

> if I offer you a 99% chance to take make one billion dollars (legally) and a 1% chance of being killed, would you take it?

I'd take it in a heartbeat. I've done things that have a 1% chance of death just to have a good story to tell.

> Put more extremely, if I offer you a 99% chance to take make one billion dollars (legally) and a 1% chance of being killed, would you take it? I definitely wouldn't.

I would even if it offered 50+% of getting killed. The small chance you strike a billion is worth the risk.

I wouldn't do it for 50%, I can see how some heuristics would get you there.

- Right now we spend 33% of our life working, nd the stress of work impacts another 33%.

- A billion is also way more than out life expected income, but there's a big marginal reduction of value to money. But you also get it now when you're younger rather than parsing it out.

- As a random 25 year old male, you would only have a 1 in 16 chance of surviving to retirement age at 60 on your own.

I personally would only take the deal for 2.5% chance of death though.

I find that fascinating. I'm not even a millionaire, yet money is rarely a constraint for me and life is great.

The only reason I can fathom desiring a billion dollars is the leverage it would give me to influence the world in what I think is a positive direction. But when it comes to money for my own sake, $10M is about the most I can imagine having use for.

It's a fun thought experiment under the many-worlds interpretation of quantum mechanics. Even at an infinitesimally low chance of getting the billion dollars, if the many worlds interpretation of quantum mechanics is true, you could argue that it is worth it from a selfish perspective.

In all the worlds where you get the money, your life is improved. In all the worlds where you don't, well, you're dead, so you don't really care.

Obviously, this doesn't take into account the impacts to everyone else in the worlds where you died.

How much money would the gain have to be for it to not be worth a 50% chance of getting killed?

I don't think that would make a difference for me if someone offers more money when my life situation is kind of nice.

What would make a difference would be my life situation - if it would be worse or totally shitty I could take even 25% for a billion.

That was kind of theme of "Squid Games".

Sure but at some point the money gets too low to risk it even if you are in the dumps, no? Would you take 1 million for 50%? 10k for 50%? 10 bucks for 1%? 1 dollar for 1%?

At some point it goes from a life changing amount of money to just "meh" money.

I would guess the uptake will fall sharply somewhere around 10 million, which is still a life-changing amount of money for 99% of the population.

Huh, it's very odd to me that you'd take an even chance of dying to win $10m... My amount would be much, much higher, potentially infinite.

Any job that earns money has an implicit acceptable $ for death risk, because all jobs have a risk of death (proxy measurement of injury using ACC rates in New Zealand).

Your sum risk might be reduced (e.g. your job pays health insurance), but many jobs increase your risk e.g. just driving to work is dangerous.

Of course your primary finite resource is usually time (years of life), and your object function usually relates to quality of life (not $) . . . Win $ for risk of death is a poor metric.

I actually wrote a post about this, using a metric for expected outcomes instead of ev probability

https://tompollak.me/blog/ev-v-eg/

> How about the following? I offer you an unfair coin flip advantage of 51% versus 49%

Isn't that just baccarat without ties?

I wouldn't know, I don't know many casino games.

I don't think the advice doesn't apply (they still claim Jane made a good decision, even though it led to a bad outcome: their point is that Jane should keep making good decisions, and not worry about bad outcomes because you can't control luck — if you mostly make good decisions, luck won't matter in the long run), it's just that it's pretty much useless because

> "Probability of getting it — you cannot know the exact probability, but estimate it the best way you can (remember the bet with a friend)"

is the hard part. Jane's case was clear (good decision), but what about Mike? Is there a proportion of his savings he should have invested that would have turned his "bad decision" into a "good one"? We have no signal to tell us what the probability of any cryptocurrency going up or down is over, say, 6 months (they've all gone up and down completely arbitrarily). Basically, finding the probability of any particular "success" (or actually, a failure) will easily tell you how much you should risk.

And that's what it's like with the most of life decisions as well: we don't know the probability — not even the ballpark range. Likelihood that any one person you meet is going to be your life-long partner is basically nil: but we still invest in building relationships before fully committing to either decide they are not, or increase the probability that they will be. But we still do that relationship building investment based on very few signals (appearance, short chats and potentially what social circle someone is from if from a shared acquaintance).

There is no probability to know in the sense you mean -- probability is what we call our personal assessment of the situation, incorporating the (little) knowledge we have.

So sure, there is your probability. Will you make money knowing it? That's the question you're alluding to.

(As a concrete example: there is no "true" probability of rain six days from now. Either it rains or it doesn't. But a skilled meteorologist can give you a probability that would probably make you a little money over the climate average.

Both the climate average and the meteorologist's assessments are correct probabilities. They just reflect different amounts of information.)

I was not referring to discreet outcomes at all (the old joke about everything having a 50% chance: it either happens or it does not).

My point is that with the available information when making most life decisions, we usually have no idea at all of even the ballpark chances of something turning one way or another.

Basically, the advice from the article is to incorporate luck into your estimations (both good and bad), and to use that to determine if a decision is good or bad, and then only go with good decisions (like a poker player would). And don't stress over bad (or good) outcomes if they were mostly due to really bad luck (i.e. something improbable has happened).

But if you don't know if chances of something happening are 10% or 90%, how do you incorporate that into your decision making?

What I am saying is that in life, we implicitely work to reduce the range (eg for a romantic partner, get to know them much better), but we've already decided to invest that much time with very slim chances of them turning out to be our lifelong partners.

Reply on news.ycombinator.com