To expand on your model, I think our ability to predict success and our ability to predict failure are separate things not on a continuum with each other but orthogonal metrics. We need to count false positives and false negatives here.
I predict that HN is no better at predicting success than average. It might be better than average at predicting failure but given the distribution that might not be that hard: if you dismiss every idea AND most ideas fail, then you always look like you are getting it right with a few occasional wrongs. Think about it in the inverse: if most startups succeeded spectacularly but a few failed miserably and you always were bullish on all startups, people would just say that you are almost always right.
The only real difference between predicting success and the result being failure and the opposite is that in this world success is unbounded (your valuation isn’t constrained since you can create a whole new market segment), but failure is always bounded. Therefore getting a success wrong sucks more than getting a failure wrong: you could have made a lot of money by correctly predicting a unicorn.
> I predict that HN is no better at predicting success than average
I can't fully agree on this. In a fair game (coin toss) your chances are just that, average. In a game where your experience, network and wealth can influence the outcome, things can be different. This is common in the enterprise market.
In the consumer market (which isn't more fair, but just works differently), that model fails miserably, that's where we start seeing the failures of the predictions and bets made.
What I see happening, is the vested parties (VC, Founders, investors) trying to cherry pick data points (from both markets), and then bragging about how they are good at predicting in both markets, but when you add in more data you see, that is not the case.
I predict that HN is no better at predicting success than average. It might be better than average at predicting failure but given the distribution that might not be that hard: if you dismiss every idea AND most ideas fail, then you always look like you are getting it right with a few occasional wrongs. Think about it in the inverse: if most startups succeeded spectacularly but a few failed miserably and you always were bullish on all startups, people would just say that you are almost always right.
The only real difference between predicting success and the result being failure and the opposite is that in this world success is unbounded (your valuation isn’t constrained since you can create a whole new market segment), but failure is always bounded. Therefore getting a success wrong sucks more than getting a failure wrong: you could have made a lot of money by correctly predicting a unicorn.