Really good explanation of it. One assumption you would have to include in this is that the consequences to your choices are ergodic and don't contain ruin events?
In this case, if you walked a bit too far in one direction at night you never get the chance to turn around because you get eaten by wolves or some such problem.
Excellent point, and the wolves analogy really drives it home. The absence of ruin events is indeed an implicit assumption here.
I think it's a legit assumption in the context of startup experiments and startups in general, because they're the canonical (haha, pun fully intended) example of low investment, high upside. It's a similar risk profile to buying lots of out-of-the-money options -- low cost, positive convexity, very Talebian. (Like Taleb, I'm a former derivs trader so I like to cast everything into that language.)
And you're absolutely right, I'd be very wary of using SOWH to explore solution spaces with unbounded downside!
What other areas would you say this SOWH framework fits or doesn't fit well with? I'd definitely agree it is well suited to Andreessen's particular software-first approach to VC. Would be harder to justify for a hard-tech company, or much less a government.
if one gets the information 'after the fact' (taking one of the roads) > with the example of 2 roads, what decision is there to make but to simply take the other road. with the example of only 2 roads; it is still decision making under uncertainity; kind of retrospective distortion
- SOSH; highly doubt if anything can be determistic, it is mostly probabilistic and expectation (=probability x payoff). Even if it works, it is simply the role of luck
- WOSH; is probably the case of blindly copying what others do (like say; the competition) without even putting that into the context and appropriate situations
- WOWH; is typically the playbook/best proactices culture without any case by case thinking as required/necessary. this could also be the case of bad/not data-driven, inspite of having the data/information (essentially lazy thinking)
Heard this repeatedly at Amazon. One of the huge benefits of SOWH is that you are constantly looking for data / evidence and you follow the evidence. Its a simple concept to explain, but extremely difficult to implement consistently and at scale, something that Amazon excelled at.
Really good explanation of it. One assumption you would have to include in this is that the consequences to your choices are ergodic and don't contain ruin events?
In this case, if you walked a bit too far in one direction at night you never get the chance to turn around because you get eaten by wolves or some such problem.
https://medium.com/incerto/the-logic-of-risk-taking-107bf41029d3
Excellent point, and the wolves analogy really drives it home. The absence of ruin events is indeed an implicit assumption here.
I think it's a legit assumption in the context of startup experiments and startups in general, because they're the canonical (haha, pun fully intended) example of low investment, high upside. It's a similar risk profile to buying lots of out-of-the-money options -- low cost, positive convexity, very Talebian. (Like Taleb, I'm a former derivs trader so I like to cast everything into that language.)
And you're absolutely right, I'd be very wary of using SOWH to explore solution spaces with unbounded downside!
What other areas would you say this SOWH framework fits or doesn't fit well with? I'd definitely agree it is well suited to Andreessen's particular software-first approach to VC. Would be harder to justify for a hard-tech company, or much less a government.
Great article! I can say from experience that it applies equally to scientists too.
if one gets the information 'after the fact' (taking one of the roads) > with the example of 2 roads, what decision is there to make but to simply take the other road. with the example of only 2 roads; it is still decision making under uncertainity; kind of retrospective distortion
- SOSH; highly doubt if anything can be determistic, it is mostly probabilistic and expectation (=probability x payoff). Even if it works, it is simply the role of luck
- WOSH; is probably the case of blindly copying what others do (like say; the competition) without even putting that into the context and appropriate situations
- WOWH; is typically the playbook/best proactices culture without any case by case thinking as required/necessary. this could also be the case of bad/not data-driven, inspite of having the data/information (essentially lazy thinking)
Heard this repeatedly at Amazon. One of the huge benefits of SOWH is that you are constantly looking for data / evidence and you follow the evidence. Its a simple concept to explain, but extremely difficult to implement consistently and at scale, something that Amazon excelled at.