5 ms·

Paul Buchheit: How to be right 90% of the time, and why I'd rather be wrong.

24 points by paul 20y ago

JMiao 20y ago
I recently started thinking similarly about the "right 90% of the time" logic, so I'm glad I read this. Thanks for another interesting entry, Paul.
myoung8 20y ago
But surely you can throw some logic in there too to increase your odds...there are some very legitimate reasons for thinking some ventures will fail. I liked Paul's post, but it comes off as encouraging people to fly by the seat of their pants so to speak. Entrepreneurship is about taking risks, certainly. But not just any risks. It's about taking calculated risks. I don't pretend to be the know-all experienced entrepreneur, but that came from Reid Hoffman and if I trust anyone I trust him.
- paul 20y ago
  Yes, absolutely you should try to improve the odds. I hope it doesn't seem like I'm suggesting that we should be completely random. My point is more that new things are all inherently risky -- I had very legitimate reasons for thinking that Google would fail, for example. As you say, this is about taking calculated risks. In my example, I tried to show how just a 20% chance of success could be very profitable (vs no profit from the 10% chance of success that you might get from being more random with your investments). Let me know if you have a suggestion for how this should be made clearer.
  - rfrey 20y ago
    Another way of looking at this (actually just a rephrasing of what you've said) is that the mathematical expectation of a strategy is the sum of (the profit of each event multiplied by the probability of it happening.) That is expec = sum(prob(x)x) For an $ investment, $expect = sum(prob(x)x - invest(x)) Say we have two strategies: One invests nothing and is right 90% of the time, but pays 0 when it's right; the other invests $50,000 and is right 10% of the time. When it is right it pays 10x, or $500,000. exp(strategy 1) = prob(x)x - 0x = (0.9)0 - 0 = 0 exp(strategy 2) = prob(x)x - x = (0.1)10 -1 = 0 So if we only have a 10% success rate and a 10x payoff, we might as well be sipping tequila on the beach. But if we are smart, or hard working, and increase our probability to 20%: exp(strategy 2) = prob(x)x - x = (0.2)10 -1 = 1 which is a positive expectation, which is a good thing. Increasing the probability OR increasing the payoff increases the expectation.
rfrey 20y ago
I second Paul's recommendation to read Taleb's Fooled by Randomness; however, one should be aware that Taleb has been on "sabbatical" from his fund for well over a year. His analysis was good, but the markets apparently didn't read his book and he went under.
- paul 20y ago
  Where do get that info about Taleb? I searched around a bit, and while there seems to be a lot of speculation, I couldn't find any real evidence that his fund ran into trouble. An older version of his web page (cached on Google) says that he's done with his sabbatical and book writing (new book out next month) and back to work: http://72.14.253.104/search?q=cache:1nVmMgBra40J:www.fooledbyrandomness.com/%3Fref%3DSevSevil.Com+taleb+sabbatical&hl=en&ct=clnk&cd=15&gl=us http://72.14.253.104/search?q=cache:1nVmMgBra40J:www.fooledbyrandomness.com/%3Fref%3DSevSevil.Com+taleb+sabbatical&hl=en&ct=clnk&cd=15&gl=us "I took a one year sabbatical to finish The Black Swan (which lasted 2 1/2 years). Now I am back but will not discuss my business activities -- only my literary and technical ideas on the philosophy of chance uncertainty, & probability. In a way I am back in business taking a rest sabbatical from thinking."
  - rfrey 20y ago
    Hi, Paul. Gosh, I can't find a link anywhere either. I didn't realize it was anything but wide open when I wrote that. But I know it's true through personal conversations with several involved parties. Everyone else will have to settle for "I heard from some guy on a forum"... Sorry - next time I'll be more careful about shooting off my mouth. <edit>After putting out some queries, I'm told that it's all available on the web via online legal filings. I'm too busy building the next FaceBook (kidding!) to spend a day searching, but I'm adding this trailhead in case anyone is very interested.</edit>

24 points by paul 20y ago

JMiao 20y ago

I recently started thinking similarly about the "right 90% of the time" logic, so I'm glad I read this. Thanks for another interesting entry, Paul.

myoung8 20y ago

But surely you can throw some logic in there too to increase your odds...there are some very legitimate reasons for thinking some ventures will fail. I liked Paul's post, but it comes off as encouraging people to fly by the seat of their pants so to speak. Entrepreneurship is about taking risks, certainly. But not just any risks. It's about taking calculated risks. I don't pretend to be the know-all experienced entrepreneur, but that came from Reid Hoffman and if I trust anyone I trust him.

paul 20y ago
Yes, absolutely you should try to improve the odds. I hope it doesn't seem like I'm suggesting that we should be completely random. My point is more that new things are all inherently risky -- I had very legitimate reasons for thinking that Google would fail, for example. As you say, this is about taking calculated risks. In my example, I tried to show how just a 20% chance of success could be very profitable (vs no profit from the 10% chance of success that you might get from being more random with your investments). Let me know if you have a suggestion for how this should be made clearer.
- rfrey 20y ago
  Another way of looking at this (actually just a rephrasing of what you've said) is that the mathematical expectation of a strategy is the sum of (the profit of each event multiplied by the probability of it happening.) That is expec = sum(prob(x)x) For an $ investment, $expect = sum(prob(x)x - invest(x)) Say we have two strategies: One invests nothing and is right 90% of the time, but pays 0 when it's right; the other invests $50,000 and is right 10% of the time. When it is right it pays 10x, or $500,000. exp(strategy 1) = prob(x)x - 0x = (0.9)0 - 0 = 0 exp(strategy 2) = prob(x)x - x = (0.1)10 -1 = 0 So if we only have a 10% success rate and a 10x payoff, we might as well be sipping tequila on the beach. But if we are smart, or hard working, and increase our probability to 20%: exp(strategy 2) = prob(x)x - x = (0.2)10 -1 = 1 which is a positive expectation, which is a good thing. Increasing the probability OR increasing the payoff increases the expectation.

rfrey 20y ago

I second Paul's recommendation to read Taleb's Fooled by Randomness; however, one should be aware that Taleb has been on "sabbatical" from his fund for well over a year. His analysis was good, but the markets apparently didn't read his book and he went under.

paul 20y ago
Where do get that info about Taleb? I searched around a bit, and while there seems to be a lot of speculation, I couldn't find any real evidence that his fund ran into trouble. An older version of his web page (cached on Google) says that he's done with his sabbatical and book writing (new book out next month) and back to work: http://72.14.253.104/search?q=cache:1nVmMgBra40J:www.fooledbyrandomness.com/%3Fref%3DSevSevil.Com+taleb+sabbatical&hl=en&ct=clnk&cd=15&gl=us http://72.14.253.104/search?q=cache:1nVmMgBra40J:www.fooledbyrandomness.com/%3Fref%3DSevSevil.Com+taleb+sabbatical&hl=en&ct=clnk&cd=15&gl=us "I took a one year sabbatical to finish The Black Swan (which lasted 2 1/2 years). Now I am back but will not discuss my business activities -- only my literary and technical ideas on the philosophy of chance uncertainty, & probability. In a way I am back in business taking a rest sabbatical from thinking."
- rfrey 20y ago
  Hi, Paul. Gosh, I can't find a link anywhere either. I didn't realize it was anything but wide open when I wrote that. But I know it's true through personal conversations with several involved parties. Everyone else will have to settle for "I heard from some guy on a forum"... Sorry - next time I'll be more careful about shooting off my mouth. <edit>After putting out some queries, I'm told that it's all available on the web via online legal filings. I'm too busy building the next FaceBook (kidding!) to spend a day searching, but I'm adding this trailhead in case anyone is very interested.</edit>