4 ms·
I hope somebody runs the data in the future to see if there was a statistical difference in outcomes between the companies intended to be accepted vs. not :-)
by seanlinehan 8y ago
I hope somebody runs the data in the future to see if there was a statistical difference in outcomes between the companies intended to be accepted vs. not :-)
- snowmaker 8y agoThat is a good idea! We conducted an inadvertent A/B test :)
- samstave 8y agoWhen the BOFH gives you segfaults, make self healing systems???
- bigwheeler 8y agoDon't forget to share the results with us!
- seanlinehan 8y agoThat's what I was thinking!
- solveit 8y agoIs there any chance you can blind yourselves to any extent? Or does everyone involved already know too much about which companies were initially accepted?
- DoreenMichele 8y agoany chance you can blind yourselves I wouldn't worry too much about that angle, personally. The human mind can follow a community of about 150. This is two orders of magnitude larger. I think they will be lucky if they can even remember names of folks they are dealing with.
- samstave 8y agoThis would be a seriously missed opportunity for YC if they dont: Immediately start tracking all the applicants and what stages they are in, how much funding/traction/etc they have and which segment vertical they are in and then seeing how well they all do. This will be a very interesting experiment.
- kcorbitt 8y agoYep. We do a little bit of that already with our normal batch applicants, but with the wide net of Startup School there may be more more data available that would be interesting to correlate with long-term success. I think it would be so valuable to have a rigorous study of the factors that correlate with startup success – letting founders know what they should optimize for is an important part of the value we want to provide!
- samstave 8y agoHow can I get started early in helping with this data? Serious question. Is there anything I can start doing/capturing that would help?
- tedmiston 8y agoIt would be interesting to write a bot that tracked fundraising event data from Crunchbase for the two groups in this cohort.
- pcmaffey 8y agoeg. 40% of VCs went to Harvard or Stanford, so therefore if you want to be a VC or know VCs, you should optimize for going to Harvard or Stanford. /s - because much of the data you'd have on such early stage co's is identity-based, and therefor chock full of bias.
- dang 8y agoSome of it for sure, but there are a lot of other things to look at. And of course the whole idea of startup school is to try to cast a much wider net, orders of magnitude wider.
- robocat 8y agoUnfortunately there is a huge self selection bias: the majority of applicants would know it is hard to get accepted, so they will have tried really hard to get in. In the next round if they said they would accept anyone, then presumably there would be a lot of low quality applicants (100's of Facebook clones etc).
- edanm 8y agoThis would still be an interesting experiment of YC's ability to choose the best startups out of a much larger set of good ones.
- k__ 8y agoMaybe they find different reasons to select. They know who they "would" have taken. And somewhere in the future they know who was successful. If that are the companies they wanted anyway, all is good. If not, they can check what was different and what spoke for and against them
- tgb 8y agoThe idea is to compare this round of applicants between the ones that were originally accepted and those that were "accidentally" accepted.