SpaceX Beer Bet Tracker

[mikelepage]

Stirring the pot - Will the next human being to land on the moon do so before the 50th anniversary of the first man on the moon, July 21 2019?  Before the 50th anniversary of the *last* man on the moon, December 7, 2022?

Will a private enterprise beat the Chinese space agency to the Moon?

I was going to add the moon anniversary idea if no one else did

But back on topic, I tend to agree with Jim though that number of launches is the thing that matters (recycling stages should make that easier, but it's only a means to an end, and trying to count them will complicate matters).

I'd rather have a running pool - a kind of "guess the number of jellybeans in the jar" style competition.  Everyone puts in their guesses and then when it actually happens we can decide who was closest and arrange for them to win a prize (I dunno - maybe Chris can pull some strings and arrange for them to meet Elon Musk ).

So with that in mind, I'd phrase it this way: Guess (at to-the-day level accuracy) when SpaceX will first perform 48 successful orbital (or beyond) launches within a running 365 day period.  Only one person per nominated date.  To make things interesting: You may only renew your guess once the date you nominate comes to pass without success, for a day not nearer than 2 years in the future.  Obviously if you don't think it will ever happen, you only win when SpaceX goes bust, in which case I doubt Elon would want to meet you

My guess is July 1st 2022.

EDIT: I suppose we also need a caveat that to win you must have nominated the date at least 2 years in advance.

[majormajor42]

So with that in mind, I'd phrase it this way: Guess (at to-the-day level accuracy) when SpaceX will first perform 48 successful orbital (or beyond) launches within a running 365 day period. 

My guess is July 1st 2022.

It's hard for me to imagine two things: an upcoming need for 48 launches by one company in a year. What do you think the spread of those 48 launches will be?
a few commercial station supply
a few commercial station personnel ferries
ISS supply and ferries
commercial satellites
military satellites
NASA satellites & probes
It's an interesting question because it just doesn't imagine SpaceX being able to do it but also imagines a need for it to be done.
By 2022, will there be some sort of large station or spacecraft construction project that requires many launches?
So this is an economic issue bigger than just SpaceX's ability to do it.

Then there is also the frequency of range closures for these flights.
What is the current total annual high for orbital flights out of the two main launch ranges? Then add the number you think they will be able to get out of Texas. (they are initially being limited). Do range closures for rocket launches have economic impact on the maritime and aviation industries? Is an increase in range closures something they would protest? Will 1st stage recovery increase the impact on other users of that space if it requires the range to remain closed for a longer period? I'm not sure how long it has taken in the past for civil airplanes to be allowed to traverse the area after a rocket launch. Extending this till the stage comes back...what is this, an extra ten...twenty minutes?
Maybe, as reliability and launch frequency increases, the rules will change.
So this is a bureaucratic issue bigger than just SpaceX's ability to do it.

That all said, one way to think about the answer it to work up to it instead of just jumping to it.
2014...6, 2015...9, 2016...14, I realize now I'm just increasing launch rate by 50% each year...2017...21, 2018...32, .... yeah, so I'm even more optimistic than Mike, despite my comments above. So "before this decade is out"? Nah, I think I throw in another half year and say July 1st, 2020.

[Robotbeat]

2022 for 48 orbital launches in a year?? I sincerely doubt it. If we DO get there, no question we'd be operating in a totally different paradigm.

[Lar]

I think 48 a year would probably be about 1/2 or more propellant supply launches (Once Elon gives in on the need for depots EVERYWHERE and agrees... ) so it could come earlier than we might think. I'd say when MCT has 2 units in operation

Of course, the BFR increases the size of the tankerload, cutting the number of flights.

I hope I live to see the day. But I doubt 2022... even me, crazy optimist that I am..

[mikelepage]

My reasoning may be a little touchy-feely for all you engineer types, but for me it's probably best expressed as a question which highlights a massive pent-up desire in the western world:

Is there any other industry where there is a such a large gap between the number of people who *want* to participate and the number of people who *can* participate?

(EDIT: My point being, once all the pieces are in place, there's no reason not to expect it to expand at an exponential rate.  I don't see the pieces being in place until next year or the year after, but then I think an expansion such as Major predicted is entirely reasonable.)

Let me just throw out one idea which becomes more viable once the cost of LEO isn't so insanely expensive: televised zero-gee sports.  As an example: Zero-gee table tennis.  Four playing surfaces (transparent perspex or some such), regulation height nets, bats and balls.  You'd also need a fixed ring of foot-holds on either end of the table so each player can stand at any orientation.  I think it would probably be as good or better an exercise regimen as what astronauts currently have to do anyway.

Honestly I think if Bigelow or anyone else can put up habitats at a reasonable pace, there will be more than enough people willing to pay (or be sponsored) for a visit.

[QuantumG]

Is there any other industry where there is a such a large gap between the number of people who *want* to participate and the number of people who *can* participate?

Sports stars, movie stars, rock stars. They all have their amateur subcultures, just like rocketry.

[mikelepage]

Is there any other industry where there is a such a large gap between the number of people who *want* to participate and the number of people who *can* participate?

Sports stars, movie stars, rock stars. They all have their amateur subcultures, just like rocketry.

All of those things require there to be an audience.  Plenty of people go to a rock concert without wanting to be onstage.

[Robert Thompson]

I converted SpaceX's launch dates so far into Unix time, plugged those into Excel, got a quadratic polynomial fit of: y = 6 * 10^(-16) * x^2 - 9 * 10^(-9) * x + 1.4313. Polynomial had higher R-squared than exponential. Keeping it as tantamount to exponential because BIC. Then I plugged that into a graphing calculator and traced to y = 11, 12 and 13, etc, to evaluate x (Unix time). Then I converted from Unix time back to date. If SpaceX maintains its present acceleration of cadence, the next launches will be roughly around

Tue, 02 Sep 2014
Sun, 16 Nov 2014
Mon, 26 Jan 2015
Fri, 03 Apr 2015
Mon, 08 Jun 2015
Mon, 10 Aug 2015
Sat, 10 Oct 2015

SpaceX cannot maintain present acceleration of cadence.

[Lar]

I take it you think it's not accelerating fast enough given the curve fitting you did? Did you use everything back to F1? What if you toss some data points, do you get different results? What if you start at F9 V1.1

Just wondering, this is interesting.

[dcporter]

I take it you think it's not accelerating fast enough

I read Hernault's post to say that their schedule is picking up speed exponentially and that they won't be able to maintain that forever (of course). Hernault is that what you meant?

[Lar]

I take it you think it's not accelerating fast enough

I read Hernault's post to say that their schedule is picking up speed exponentially and that they won't be able to maintain that forever (of course). Hernault is that what you meant?

Most phenomena can't maintain exponential acceleration indefinitely (or even polynomial for powers greater than 1 and I thought he said it was a better poly fit that exp, but used exp "because I can" (BIC) )

My read of those dates is that they don't meet the manifest... not enough launches in out years and the backlog grows (or customers leave)

[Coastal Ron]

I converted SpaceX's launch dates so far...

Of course that assumes no other factors change from the past, including design improvements, production rate, customer readiness, number of customers, number of launch facilities, improvements in launch facilities, improvements in launch capabilities, a learning curve, etc.

I'm not a math whiz, but if those new factors increase faster than the past launch history, and could be factored in, wouldn't that change the predicted outcome?

[Lar]

Of course, but this was just a curve fit... not an actual model

[Robert Thompson]

Lar: If I drop first and second launches, to erase that huge gap between second and third launches, it's exponential. I'll post that projection later. If I do Falcon 9 V1.1 alone, there's fewer data, so larger error, but an unambiguous negative quadratic (decelerating) launch rate. {y = - 3 * 10 ^ -15 x^2 + 8*10^-7*x.}

(BIC is http://en.wikipedia.org/wiki/Bayesian_information_criterion. It means, 'Don't use so many parameters to achieve a perfect fit that it looks airbrushed. Your model will pay for that in credibility.' I'm sticking with a 2nd order polynomial because customers/investors don't want to hear about a service provider's 4th order oil in the capsule or 5th order helium leak. They want a simple, robust, predictable prospect.)

Coastal Ron: Useful objection. Moore's Law affects most but not all the factors you mention. If SpaceX's launch rate reflected only factors pegged to Moore's Law, it should increase exponentially. But number of facilities for one will not grow exponentially, so that bottleneck is a linear subtraction. In that light, the present increase is realistic (sub-exponential), rather than pessimistic (only-quadratic).

dcporter: The launch rate to date is only accelerating at a rate less than quadratic, therefore less than exponential. The difference between these fits is the difference between an R^2 of .98 over .97, .96, .95 or so. The curve *looks exponential. I think, and I think I have numerical indicators, that SpaceX Must step it up just to meet the demands of its back log. Most commercial customers can probably bear with a service provider that gives consistent evidence of progress. E.g. SpaceX has lost no primary payload.

[Hyperion5]

Lar: If I drop first and second launches, to erase that huge gap between second and third launches, it's exponential. I'll post that projection later. If I do Falcon 9 V1.1 alone, there's fewer data, so larger error, but an unambiguous negative quadratic (decelerating) launch rate. {y = - 3 * 10 ^ -15 x^2 + 8*10^-7*x.}

[/quote]

Hernalt, you may want to wait a year or two before applying that kind of mathematics to the Falcon 9 V1.1's launch rate.  The sample size is simply too small to be very valuable.  When we've gotten at least 10 launches, then you'll have a halfway decent gauge of launch rate.  Statistically speaking however I'd prefer if we didn't cast too much judgment till we've got a truly adequate sample size (20 launches at least).  Otherwise the chances of errors being caused by small sample size will be simply too large. 

[skybum]

It's hard for me to imagine two things: an upcoming need for 48 launches by one company in a year. What do you think the spread of those 48 launches will be?

The operational model for commercial stations such as Bigelow proposes will be very different from the ISS. Short-duration, high turnover. I can easily imagine a single Bigelow station having a requirement for 12-24 passenger/cargo launches per year. By 2022 I expect that there will be more than one such station -- so yeah, the industry will be very very different than what we see today, and 48 will hopefully seem like a conservatively small number then.

In the short-term, I don't foresee SpaceX getting more than 6 launches this year, and 9 next year... but my hunch is that they'll be able to continue that rate of growth for a very long time.

[AJW]

If you look at the number of days between the first 10 Atlas V launches, they were staggered at 266, 64, 519, 84, 154, 160, 91, 323, and 98 days, so an average of 195 days between launches.  This will tell you very little about launch rates just a few years later.  If you want to test your methodology of prediction the future, start to see if it can accurately predict events in the past.

Launches are dependent on a myriad of other factors, payload readiness, weather, range, ISS VV schedule.  In addition, the first units off any assembly line often have issues that need to be resolved, and many efficiencies are only found after running production for some time.  As an example, the Tesla factory here is currently shut down for improvements that are expected to increase production efficiency by as much as 25%.

The next two months of SpaceX launches may tell us more about their capabilities going into the future, than any amount of studying of the past four years of launches.

[Robert Thompson]

Hyperion5: You're not wrong.

AJW: I want to look at Atlas next. Primary competitor.

Lar: I did the V1.1 launches slowly from scratch. Not exponential. A quadratic with a positive linear term. y = 9 * 10^-16 * x^2 + 4 * 10 ^08 * x. The schedule slides to the right early because of ignoring the first two Falcon launches. It overtakes the previous fit on the 11th launch of V1.1. Then it picks up because of the positive linear term.

Sun, 21 Sep 2014 9th  launch of V1.1 (11th launch of Falcon)
Tue, 25 Nov 2014
Mon, 26 Jan 2015
Thu, 26 Mar 2015
Thu, 21 May 2015 13th launch of V1.1 (15th launch of Falcon)
Tue, 14 Jul 2015
Fri, 04 Sep 2015 15th launch of V1.1 (17th launch of Falcon)

[TrevorMonty]

It's hard for me to imagine two things: an upcoming need for 48 launches by one company in a year. What do you think the spread of those 48 launches will be?

The operational model for commercial stations such as Bigelow proposes will be very different from the ISS. Short-duration, high turnover. I can easily imagine a single Bigelow station having a requirement for 12-24 passenger/cargo launches per year. By 2022 I expect that there will be more than one such station -- so yeah, the industry will be very very different than what we see today, and 48 will hopefully seem like a conservatively small number then.

In the short-term, I don't foresee SpaceX getting more than 6 launches this year, and 9 next year... but my hunch is that they'll be able to continue that rate of growth for a very long time.

A 2 module Bigelow research station with 12 staff on 3 month cycle would require 8 passenger flights a year and 6+ cargo flights.

I to can see there being a lot more just 2 BA330 modules.

[Robert Thompson]

Prior to Asia Sat 8, I am lodging my c code results for Lar's optimistic choice of fit for Falcon 9 v1.1, alone. At its current increase of flight rate, Falcon 9 v1.1 reaches 21 launches in 2023, 23 launches in 2024 (<15> days apart), 25 launches in 2025, 26 launches in 2026 (the 26th falls on December 18), 29 launches in 2027 (so 2026 should really have 27). Unix time for 32 bit stops at 2038 - the number of fights per year in 2037 is 46 (<7> days apart). I might try Musk's statement about one thousand launches per year doing manual calculation or upgrading for 64 bit.

[editing 8/1 15:16]

Atlas V:

Atlas V launches start 8/21/2002, with highly variable cadence up until the 15th launch on 4/4/2009  (wikipedia).
The polynomial for the _entire_ sequence of all Atlas V launches is:
Launch no. = 3*10^-16*sec^2 + 2*10^-9*sec + 2.4766, R^2 = 0.99671.
A sequence of low variability cadence missions starts June 19, 2009.
The polynomial for the _recent, low variability_ sequence of Atlas V launches is:
Launch no. = 4*10^-16*sec^2 + 1*10^-7*sec + 2.0243, R^2 = 0.99607.
ULA formed in December 2006. I cannot determine any reason for the year long gap between launch 14, April 14, 2008, and launch 15, April 4, 2009. No failures, nothing on the wiki page, and I have no knowledge of political effects at the LM/ULA level.
The fit for the entire launch history skews forward in time because of the high-variability low-cadence in the first few years.
The fit for the recent low-variability high-cadence string skews forward in time due to the 4 recent instances that ULA achieved whiplash turnaround, 8/30/12 + 9/13/12, 1/31/13 + 2/11/13, 11/18/13 + 12/6/13, 4/3/14 + 4/10/14. These whiplash instances "make" the rest of the low-variability high-cadence launches "look" like they're standing still. The last one, with one-week turnaround, causes the most skew forward. An analogous effect is seen going from the entire Falcon 9 sequence to the Falcon 9 v1.1 alone. So that's assured launch for ya. In comparing annual launch count, the recent string of Atlas V immediately overtakes the entire string.

The following sequence of future circa launch cadence reflects the entire lifespan of the Atlas V:

Launch No.  45.000000090052    No. for Year:   4    Circa: 2014 06 18 17:47:32    Unix: 1403138852    Delta_t Secs:  4453016    Delta_t Days:  51
*Launch No.  46.000000089251    No. for Year:   5    Circa: 2014 08 08 16:16:53    Unix: 1407539813    Delta_t Secs:  4400961    Delta_t Days:  50
Launch No.  47.000000072161    No. for Year:   6    Circa: 2014 09 28 00:48:23    Unix: 1411890503    Delta_t Secs:  4350690    Delta_t Days:  50
Launch No.  48.000000197148    No. for Year:   7    Circa: 2014 11 16 19:50:07    Unix: 1416192607    Delta_t Secs:  4302104    Delta_t Days:  49

2014, 4 (skews dates forward less but under-reports for annual count)
2015, 5
2023, 11 launches
2024, 12
2025, 13
2026, 14
2027, 14 (last launch on Dec 20, so fudge that to "about 15" launches)
2037, 23 launches.

The following reflects the recent string of high cadence low variability launches (note skew forward):
Launch 32 in this, recent string, is the same physical launch as Launch 46 (above) in the entire lifespan.

Launch No.  31.000000219046    No. for Year:   6    Circa: 2014 09 12 15:34:34    Unix: 1410561274    Delta_t Secs:  4242539    Delta_t Days:  49
*Launch No.  32.000000169339    No. for Year:   7    Circa: 2014 10 31 01:26:48    Unix: 1414744008    Delta_t Secs:  4182734    Delta_t Days:  48
Launch No.  33.000000127751    No. for Year:   8    Circa: 2014 12 17 19:23:18    Unix: 1418869398    Delta_t Secs:  4125390    Delta_t Days:  47
Launch No.  34.000000151481    No. for Year:   1    Circa: 2015 02 02 22:02:20    Unix: 1422939740    Delta_t Secs:  4070342    Delta_t Days:  47
Launch No.  35.000000104198    No. for Year:   2    Circa: 2015 03 21 09:59:40    Unix: 1426957180    Delta_t Secs:  4017440    Delta_t Days:  46
Launch No.  36.000000091333    No. for Year:   3    Circa: 2015 05 06 07:48:49    Unix: 1430923729    Delta_t Secs:  3966549    Delta_t Days:  45
Launch No.  37.000000022953    No. for Year:   4    Circa: 2015 06 20 16:01:13    Unix: 1434841273    Delta_t Secs:  3917544    Delta_t Days:  45

2014, 8 (skews dates forward more but more or less 'correctly' reports for annual count)
2015, 8
2023, 15 launches
2024, 15
2025, 16
2026, 17
2027, 18
2037, 25

Here's to Asia Sat 8.

[Editing Tue 8/5 2:41 Tucson]

I was misusing "V1.1". Incorporating Asia Sat 8, and correctly fitting V1.1 (6! flights), it's a decelerating curve with low R^2, so it's not useful. What I fitted at the beginning of this post, calling it "V1.1", was "Falcon flights since 2012". This string, "Falcon flights since 2012", has higher R^2 than "All Falcon flights" or "V1.1", and after Asia Sat 8 has annual flight rates:

5 for 2014, 7 for 2015, 9 for 2016, 11 for 2017, …, 23 for 2023, 25 for 2024, ..., 51 for 2037.