4 ms·

Presumably they've tried to estimate that probability and factored it into the 20% probability of not succeeding. (I don't know though, since I haven't seen the

by smeyer 11y ago

Presumably they've tried to estimate that probability and factored it into the 20% probability of not succeeding. (I don't know though, since I haven't seen the calculation.)

tsotha 11y ago
Still... you can't assign a probability to something that is, by definition, unknown. "This is just a complete guess..." <-- reasonable "Bayesian analysis tells us..." <-- smells of ass
- TeMPOraL 11y ago
  But you shouldn't also pretend you have zero information. They hit the barge last time, engineering principles are sound, so those two things alone give you an estimate of "quite likely". Yes, 80% is kind of arbitrary, but it tells you the same story as "quite likely" with an added benefit of being able to plug it into some math that will yield you better results than going with just words. See "If It’s Worth Doing, It’s Worth Doing With Made-Up Statistics": http://slatestarcodex.com/2013/05/02/if-its-worth-doing-its-worth-doing-with-made-up-statistics/ http://slatestarcodex.com/2013/05/02/if-its-worth-doing-its-....
  - jaggederest 11y ago
    As a counterargument: http://lesswrong.com/lw/sg/when_not_to_use_probabilities/ http://lesswrong.com/lw/sg/when_not_to_use_probabilities/ (Note that I think that starting with a good 50/50 coinflip and revising it based on evidence in this case is reasonable, I just think it's nice to have a reasoned opposing opinion sometimes.)
  - tsotha 11y ago
    >But you shouldn't also pretend you have zero information. If you've never done something before you really have no idea if it's going to work, and in rocket launch everything that occurs after the last failure is a complete unknown in terms of failure modes. Now, if he'd made successful flyback stages before, or it was a relatively routine thing for the industry he might have a good enough feel to assign a ballpark number. But that's not the situation. In the link you provided he uses computer failure as an example, where a person who is familiar with computers does have some information - I've owned my current rig for three years and it's never failed. If you told me you thought it has a 50% chance of failure next month I can pretty confidently say that's an overestimate. But it's useless to pull numbers out of the air for something as complicated as SpaceX is trying to do.
    - FeepingCreature 11y ago
      If you have an expectation, you have a probability. Not a numeric one, but a probability nonetheless. Nobody has ever landed a rocket before. But we don't expect the rocket to turn into an alarm clock or suddenly develop antigravity. In fact, we have quite reasonable expectations on what behaviors the rocket will exhibit. We might, for instance, expect the rocket to crash more than we expect it to land. What else does that say rather than p(crash) > p(land)? (Of course, this may well come down to "probability as ratio" vs. "probability as anticipation", which is probably a matter of preference.)

by smeyer 11y ago

Presumably they've tried to estimate that probability and factored it into the 20% probability of not succeeding. (I don't know though, since I haven't seen the calculation.)

tsotha 11y ago

Still... you can't assign a probability to something that is, by definition, unknown. "This is just a complete guess..." <-- reasonable "Bayesian analysis tells us..." <-- smells of ass

TeMPOraL 11y ago
But you shouldn't also pretend you have zero information. They hit the barge last time, engineering principles are sound, so those two things alone give you an estimate of "quite likely". Yes, 80% is kind of arbitrary, but it tells you the same story as "quite likely" with an added benefit of being able to plug it into some math that will yield you better results than going with just words. See "If It’s Worth Doing, It’s Worth Doing With Made-Up Statistics": http://slatestarcodex.com/2013/05/02/if-its-worth-doing-its-worth-doing-with-made-up-statistics/ http://slatestarcodex.com/2013/05/02/if-its-worth-doing-its-....
- jaggederest 11y ago
  As a counterargument: http://lesswrong.com/lw/sg/when_not_to_use_probabilities/ http://lesswrong.com/lw/sg/when_not_to_use_probabilities/ (Note that I think that starting with a good 50/50 coinflip and revising it based on evidence in this case is reasonable, I just think it's nice to have a reasoned opposing opinion sometimes.)
- tsotha 11y ago
  >But you shouldn't also pretend you have zero information. If you've never done something before you really have no idea if it's going to work, and in rocket launch everything that occurs after the last failure is a complete unknown in terms of failure modes. Now, if he'd made successful flyback stages before, or it was a relatively routine thing for the industry he might have a good enough feel to assign a ballpark number. But that's not the situation. In the link you provided he uses computer failure as an example, where a person who is familiar with computers does have some information - I've owned my current rig for three years and it's never failed. If you told me you thought it has a 50% chance of failure next month I can pretty confidently say that's an overestimate. But it's useless to pull numbers out of the air for something as complicated as SpaceX is trying to do.
  - FeepingCreature 11y ago
    If you have an expectation, you have a probability. Not a numeric one, but a probability nonetheless. Nobody has ever landed a rocket before. But we don't expect the rocket to turn into an alarm clock or suddenly develop antigravity. In fact, we have quite reasonable expectations on what behaviors the rocket will exhibit. We might, for instance, expect the rocket to crash more than we expect it to land. What else does that say rather than p(crash) > p(land)? (Of course, this may well come down to "probability as ratio" vs. "probability as anticipation", which is probably a matter of preference.)