My girlfriend runs a mini student incubator at college, and so occasionally she'll let me read startup applications to get my thoughts. Scanning through them, I sometimes couldn't help but be overly critical. "Applying machine learning to this? Moving onto the cloud what?" My first instinct would be trying to poke holes wherever I could, beyond what would have been productive, my skepticism probably originating from my own numerous startup failures. Nothing like a sore loser, am I right?
"Look at this one," I would say, "using VR to cure insomnia? Come on, that's not realistic at all. Who would wear a headset to sleep? How can you see the screen if you close your eyes???" (not a real example but you get the point)
She would normally give me a frown and ignore me, but one day she probably had enough and responded, "If you assume they're all going to fail, of course you're almost always going to be right just by probability."
More than 90% startups fail. I'd wager that number is closer to 99% if you factor in the half-hearted efforts. As a result, if you assume failure, you're almost always going to be right, but that doesn't mean you're insightful. This is very much a case of "a broken clock is right twice a day" except here the broken clock is right 99% of the day which lead to my flawed mental model: I'd see a startup, I'd think it would fail, I'd watch it fail, I'd feel reaffirmed in my assessment ability, and thus I'd become more disillusioned.
This is an insanely gross oversimplification, but if you buy a stock, the probability it goes up or down is each about 50%. So if you make the "right" decision as often as random chance would, you net even. It takes skill to push the needle above the statistical average. An investor who makes the right decision 60% of the time is a wealthy genius. A baseball batter with a 10% above average on base percentage is a hall of famer. Thus, to be a good assessor of startup ideas, I shouldn't always assume failure even though I'd be mostly right. Assuming failure is easy. Finding that special kernel with above average accuracy is really hard, and to do it, I need to be more open minded and humble.
This insight probably is obvious to most, but for me it's a reminder not to let my unfair skepticism feed my ego. This doesn't mean I should throw away my skepticism (the flip side is becoming the gullible fool), but I should approach assessment with more objectivity and an understanding of the risk. Anyway, here's Wonderwall.