There’s a front-page story today. As I noted in a comment there, I found the final sentence interesting:
The rocket has just two successful test launches.
While true, there are other ways to phrase it. They could have left out the “just,” which implies that the number is both low, and bad. There is also an implication that there have been unsuccessful launches. It would have been just as accurate, and more favorable to the company, to write, “The rocket has had two successful launches, with no failures.” They could have even pointed out that the capsule performed successfully on its first and only flight.
>> there are other ways to phrase it.
.. but is there a reason to ?
2 launches is statistically insignificant, and with their current launch rate ( avg of 1.5 a year ? ) its going to take a long time to build up a serious reliability record.
.. but is there a reason to ?
To not mislead the reader?
2 launches is a terrible small statistical sample, but 2 successes and no failures is more information than 2 successes and no mention of total attempts or failures. As phrased it vaguely suggests there were failed launch attempts.
Why would the Times denigrate a successful local employer? Reminds me of the Churchill speech…”adamant for drift.” The Times is solidly for employment, except when someone is hiring.
A very neutral phrasing might go something like:
The Falcon X has just 2 successful launches in two tries, compared to the longer record of Atlas V which is 26 of 27
1. Space launch reliability is not (yet) established by conducting a statistically significant number of launches. Launcher reliability is established by a combination of mathematical analysis, handwaving, and/or wishful thinking to reliability statistics at the component and subsystem level. There may or may not be a small number of test launches as a final check before the vehicle is declared operational, but that’s not the basis for declaring it safe or reliable.
2. The Saturn V was used to launch men to Lunar orbit after just two launches, only one of them successful.
3. The L.A. Times is still a print newspaper, which means deadlines matter and bandwidth matters. Maybe they should have left out the “just”, but really – Mr. Hennigan had on maybe eight minutes to figure out the right eight words to describe the Falcon 9’s track record to the audience. He got it mostly right, in an article that was mostly positive, so I’m inclined to cut him some slack in the nitpicking department.
“He got it mostly right, in an article that was mostly positive, so I’m inclined to cut him some slack in the nitpicking department.”
I agree but it makes for good comments here 🙂
How many successful launches would be considered statistically significant and how could any company afford that many test launches?
2/2 is no less significant than 26/27. They both suffer from the fallacy of small numbers.
Call me crazy, but I’m going to have more confidence in a vehicle that has flown successfully twice with no failures than one that had failed twice with no successes, or not flown at all.
That’s a good point about the Saturn V and a good question about how many flights would be considered statistically significant.
They flew the first Shuttle mission with men on board. Before that, it’d only had 5 glide tests and the last one of those encountered Pilot Induced Oscillations (PIOs). Without any additional glide tests, they launched Young and Crippen on a brand new design. The individual components of the Shuttle had been tested many times but never all of them at once. It almost didn’t end well.
So, how much testing is sufficient? Well, this week Boeing announced the completion of 787 flight testing (at least for the Rolls Royce powered version).
Flight tests for certification of the initial Rolls-powered 787-8 follows 20 months of testing starting with the maiden flight of the first aircraft on Dec. 15, 2009. The company originally planned to complete tests in nine months and 2,430 flight hr., but was forced by several test and production-related issues to more than double the overall certification period while simultaneously extending flight time for the Rolls-fleet to around 3,790 hr.
So, for an airliner, a test program consisting of thousands of flights and hours and involving multiple planes is required to certify it ready for commercial service. Would such a test program even be possible for spacecraft, even if they were reusable? I don’t think so. The Russian Soyuz booster in its various versions (SL-04 Soyuz and SL-06 Molynia) has a track record of over 1200 flights spanning more than 40 years. It has a proven reliability of about 96%. However, when they were first flying the design, they had a rash of flight failures. Instead of abandoning the design, they fixed the problems and went on to launch over 50 of them a year until the Soviet Union collapsed. No one, not even SpaceX, is planning on anything like that many flights per year. So how many flights are necessary to say the system has a demonstrated statistically valid reliability record? Good question.
how many flights are necessary
Let the people paying and riding decide. Then in comments over there…
the shuttle is doing more than re-supply
Suggesting it’s a more capable craft than the Dragon which happens not to be the case (book and it’s cover anyone? Those wings are sexy but do they allow it to remain on station for two years or land on the moon or mars?)
So Dragon is cheaper AND more capable. Soon it will land on it’s own rockets… take that shuttle.
The ultimate answer to the Saturn V – and the reason why it should never even be brought up in a discussion about safety – is that it was *important* like nothing we do today is when it comes to space. The only reason we even care about safety today is because vehicles blowing up is a good enough reason to shut down the funding.
They both suffer from the fallacy of small numbers?
Which makes statistics the wrong tool for this problem. Imagine if we needed statistical support for all of our actions. There wouldn’t be any since everything starts small.
So you do what you can and apply quality control by any reasonable means. One of the things that impressed me with SpaceX early on was their approach to quality controls. It doesn’t have to be costly to do that right.
They’ve got 1500+ employees now which makes it an entirely different company from when they started. But it’s not too big yet, not to be manageable. So failures will come, but by then they should have established themselves pretty well to overcome them.
Apparently someone at the Los Angeles Times listened. When I received my copy of the paper Thursday morning, 18 August 2011, the last sentence/paragraph was:
The rocket has two successful test launches.
Rand rules!
So failures will come, but by then they should have established themselves pretty well to overcome them.
I had an interesting conversation with some SpaceX employees last April at the Space Symposium. When I complimented them on their fixing the unexpected roll of liftoff of the first Falcon 9, they said that one of the advantages of building almost everything in-house is that there’s no room for finger-pointing. You didn’t have a bunch of subcontractors all trying to blame the others. Instead, they knew it was a SpaceX problem and immediately set about solving it. While SpaceX is growing and that can make things less manageable, as long as they maintain the sense of company accountability, I think they’ll do well even in the event of setbacks. They learned those lessons on the less expensive Falcon 1. Even if they never fly another Falcon 1 (and there aren’t many scheduled flights), building that rocket was money well spent if for nothing else than to learn some valuable lessons and gain operational experience.
>>The rocket has two successful test launches.
>>Rand rules!
I bet the chain reaction was Rand->SpaceX PR department->LAT Editor
I still dont think saying “just two” was misleading the reader.
I would use “just launches” with any operational rocket, with the single exception of R-7 family perhaps, which has over a thousand launches combined.
But even the venerable Soyuz-U has never had a statistically significant flight rate in its history.
Actuallyi take that back : in 1979 Soyuz 11A511U flew 47 times.
Thats close to a flight a week. For an orbital launch system, thats actually significant.
Quiz for the Reporter: How many launch attempts (for argumentation purposes, defined here as hold-down restraints being let go and the vehicle lifting off the pad) did the National Space Transportation System have before the third successful one?