Hacker News Reader

Comment by pg | Hacker News Reader

"There are already signs of paring back. In December, Y Combinator’s Graham wrote about his accelerator’s decision to reduce its class size, from 84 companies in the summer 2012 class to less than 50 in the current session."

This was nothing to do with any overall trend. It was just because YC's then structure couldn't deal with so many startups.

The other thing this guy doesn't seem to grasp is that we didn't decrease the size of something homogenous. This is not like a restaurant cutting back from 84 tables to 46. Essentially we picked the best 46 out of 84. And in any group of 84 startups, the top 46 will have 100% of the success.

There are so many of these things now that I wouldn't be surprised if a lot of them die off. But if that happened I don't think it would affect us.

View on HN

If you had the ability to pick the top 46, why did you fund 84 in the first place?

(Not doubting, honestly curious.)

Between batches we spent a lot of time systematically trying to identify the predictors of failure. Previously we only tried to look for predictors of success. And they are not simply complements.

     they are not simply complements.

Which argues against this change, right? If you think the next Dropbox could have some red flags along with its positive signs, you'd want to mostly ignore red flags.

Somewhat. Depends how good the predictors of failures are. If they're strong enough, it doesn't matter how many promising signs there are.

But you're right that this is a perilous move. We only did it because we had to, and I'd like to edge back toward our old approach now that we can scale better. We'll probably always have a tighter filter than we used to, though, if only because some of the techniques we learned for predicting failure seem very useful.

Is there an essay coming on your new findings? Not to pry but "How Not to Die" is one of my all-time favorites.

OK, glad to hear it.

This is a pretty bold assumption that the 46 you selected where the top 46.

We were already claiming we could pick the top 84 out of ~3000. It's not that much harder to narrow it down to 46.

I guarantee you that pg isn't claiming he knows which the "Top 46" of 84 are. All he can do is pick the best as he see's them, and hope for the best.

I think he's said in many article, that, in fact, the vast majority of the returns each session will come from only one or two startups, and he certainly isn't able to identify those two in advance.

You refer to the structure in the past tense. Are we going to see bigger class sizes again with S13?

Probably. Even if we kept the new, tighter filter we'd still accept more groups, because the number of applications continues to rise.

Reply on news.ycombinator.com