Hacker News Reader

Comment by buboard | Hacker News Reader

> We rarely appreciate just how much inconsistency there is in the judgments we and others make, argues Prof Kahneman. It can hardly be a surprise, then, if past performance is no guarantee of future success.

Past performance can often predict future success that's why geniuses often are serial geniuses, sometimes in more fields than one. But we often forget that before they became known as geniuses they passed through a string of failures as is necessary. Im not sure if it's a problem with the geniuses themselves, or a problem of perception from our side, as we become intolerant to their failures.

View on HN

“future success” does not have to be preceded by past “string of failures”.

Like, at all. That’s just classic gambler fallacy.

If you want an example of reverse gambler fallacy, see here - https://news.ycombinator.com/reply?id=20231181&goto=threads%...

There the author says “10k heads have shown up, so 10k tails are waiting to show up”

which is equally rubbish.

There’s this weird thing where people on HN keep falling prey to Gambler’s fallacy. Maybe it’s like a programmer’s blind spot ?

If you want an extreme example of this reasoning: I always take a bomb with me on a plane, because what's the chance that two guys with a bomb show up, right?

i agree. but i suppose even einstein had a few failures . It's just that failures are magnified with notoriety, even if these failure might just be the prelude to following success.

> Maybe it’s like a programmer’s blind spot ?

Nah this must come from entrepreneurs , which is a lot more like gambling than science etc.

You could turn it around and say, assuming I eventually succeed, what's the distribution of failures I'd have to endure? You might get lucky, but it's definitely possible to quantify how long it would most probably take to succeed.

If you have a bunch of people tossing dice and stopping when they roll a 1, only ~1/6 will succeed immediately. Most people will take a few tries; a few will take many tries; there's a nonzero chance you will wait forever without a success.

That's not to say that a success gets more likely with time (the Gambler's Fallacy) but on the whole, you'd expect most successes to have some prior failures, unless the dice are rigged.

> it’s definitely possible to quantify how long it would most probably take to succeed

yes. if success means to roll a 1, and anything else is a failure, you will on average roll the die 6 times before you succeed. E(x) = 1/p for a geometric with p = 1/6. your variance is q/p^2, so 30.

so then, my rational decision as an investor is to quit rolling the die after 17 rolls ( mu + 2 sd = 6 + 2*sqrt(30) = 17 ). I wouldn’t wait until heat death of universe.

It's not a gambler fallacy, it's the idea that failure is an important part of learning to be good at something.

Buboard may be assuming that performance is greatly influenced by skill and that skill is learned rather than innate, leaving a 'string of failure' in the learning process. Probabilistically, this is roughly that skill-based success is a coin with highly biased initial weight for failure, and learning changes the coin's bias. I do not care to speculate on how success actually works; I'm quite sure it's more complicated than I can fit in a comment.

If there's one thing I've learned over the years is that there is no correlation between being a programmer and following the scientific method or not being prone to fallacies. A lot of programmers go with their hunches when it's inappropriate.

Reply on news.ycombinator.com