SpaceX progress towards a 100 launch year

[Robotbeat]

We don’t have statistical power in the data to look month to month, this is just looking for a signal in noise.

The data currently is consistent with them reaching 100 flights this year. Squinting at noisy data is not really a valid analysis method. I’m not saying it “looks exponential,” I’m using the current launches so far this year, comparing them to last year, and finding the rate of improvement necessary for those two numbers to be consistent, and then extrapolating that to the rest of the year. (Note, this would work even if they were getting worse over time.) I’m not setting anything to a certain growth rate, it just comes from long term data. And this growth rate is consistent with the rate from 2022 to 2023 as well.


“But the launch rate in late 2022 isn’t much different than early 2023” yeah, that’s what local linearity means. That’s what you’d expect with a slow compounding improvement. The data will be too noisy to get a clear exponential signal at anything smaller than, say, 4-12 month chunks.

A Falcon launch failure could throw a monkey wrench into all of this. But barring that, ~100 flights seems the most likely outcome at this time. And as time goes on, the odds of being much more or much less than that number decreases.

[oldAtlas_Eguy]

With current assets: ASDS and pads. SpaceX can do a maximum of 137 launches /yr. That is 8 day turnaround per ASDS, 3 ASDS and all launches using ASDS. For pads it is a maximum of 219 launches /yr. That is 5 day turnaround per pad with 3 pads and 82 of the 219 launches being RTLS.

This suggest even now we are not significantly close to the max realistic launch rate for F9.

SpaceX is supposedly in process of building/converting 2 new pads/sites one at CC and one at VSFB.  If they also start building 2 more ASDS, the maximum then jumps for the all ASDS launches to 228. For optimal mix of ASDS and RTLS that becomes 365. One a Day.

But unlikely the F9/FH yearly launch rates will reach much above 110 just because Starship will start eating into the sats that F9 is launching. At a conversion of initially 1 Starship launch for an existing 3 F9. Such that even 5 Starship with Starlinks would reduce F9 launches by 15. The exception is that if Starship has more difficulty getting out of its early development phase and launch rates remain in the 12/ year rate for a couple more years. Starship has the potential to significantly jump in launch rate even with just 2 pads. With each of the 2 existing easily doing a launch a week for ~100 total a YEAR. Starship's limiting launch rate is pad turnaround ONLY. The only exception is weather impacts which is also presents F9 with similar launch rate impacts. Weather impacts could be as high as 30% reduction in the max rates possible for Starship or F9. NOTE that F9 ASDS max rate of 137 when the 30% weather impact is applied is 96. For the Starship in the once a week and 2 pads it becomes 70. With a combined of a reasonable 165.

So even when Starship starts taking over all the Starlink launches to 53 degree or lower inclination orbit launches. The number of Starlink launches are not likely to decrease but increase. The need is to do eventually in the next couple of years ~6000 Starlinks' sats launch per year, which becomes a mix of Starship and west coast F9 launches of >100.Then add to that another bunch of >40 Starship launches for BEO destinations Lunar/Mars. Next other LEO launches by Starship ~10 growing to 50 as this is mostly taken over by Starship from F9/FH. Then all of the remaining but decreasing F9/FH flights starting at 90+ and decreasing each year possibly over less than a decade.

My suggestion is that SpaceX launch rates will not in generalities unless significant launch failures case cause decreases but would increase each year even though more and more of the current payloads is moved from F9/FH to Starship.

Launch rate per year will continue to go up in jumps and spurts. Sometimes slight decreases or flattening then a big increase. This is the pattern we have seen with F9/FH. Such that with a combined Starship and F9/FH it will continue and then once in the only Starship mode would  even then continue upward.

[meekGee]

We don’t have statistical power in the data to look month to month, this is just looking for a signal in noise.

The data currently is consistent with them reaching 100 flights this year. Squinting at noisy data is not really a valid analysis method. I’m not saying it “looks exponential,” I’m using the current launches so far this year, comparing them to last year, and finding the rate of improvement necessary for those two numbers to be consistent, and then extrapolating that to the rest of the year. (Note, this would work even if they were getting worse over time.) I’m not setting anything to a certain growth rate, it just comes from long term data. And this growth rate is consistent with the rate from 2022 to 2023 as well.


“But the launch rate in late 2022 isn’t much different than early 2023” yeah, that’s what local linearity means. That’s what you’d expect with a slow compounding improvement. The data will be too noisy to get a clear exponential signal at anything smaller than, say, 4-12 month chunks.

A Falcon launch failure could throw a monkey wrench into all of this. But barring that, ~100 flights seems the most likely outcome at this time. And as time goes on, the odds of being much more or much less than that number decreases.

And yet you look at the last 3 months, ignore the 6 before it, then (on a different postl look at a few months before those 6.

Again, spaceX yearly launches were 21, 13, 26, 31, 60, and maybe 100.  That's jumps of -40%, 100%, 20%, 100%, 50%.

That's not smooth or exponential by any stretch.  It speaks of discrete jumps in capabilities (reusability, new pads, etc) and very slow increases in between due to optimization.

Without new pads, I agree they'll cross 100, but I don't expect more than 150 F9 ones.

I also don't expect more F9 pads.  I think that's just a hedge in case there's a major problem with SS.

[steveleach]

We don’t have statistical power in the data to look month to month, this is just looking for a signal in noise.

The data currently is consistent with them reaching 100 flights this year. Squinting at noisy data is not really a valid analysis method. I’m not saying it “looks exponential,” I’m using the current launches so far this year, comparing them to last year, and finding the rate of improvement necessary for those two numbers to be consistent, and then extrapolating that to the rest of the year. (Note, this would work even if they were getting worse over time.) I’m not setting anything to a certain growth rate, it just comes from long term data. And this growth rate is consistent with the rate from 2022 to 2023 as well.


“But the launch rate in late 2022 isn’t much different than early 2023” yeah, that’s what local linearity means. That’s what you’d expect with a slow compounding improvement. The data will be too noisy to get a clear exponential signal at anything smaller than, say, 4-12 month chunks.

A Falcon launch failure could throw a monkey wrench into all of this. But barring that, ~100 flights seems the most likely outcome at this time. And as time goes on, the odds of being much more or much less than that number decreases.

And yet you look at the last 3 months, ignore the 6 before it, then (on a different postl look at a few months before those 6.

Again, spaceX yearly launches were 21, 13, 26, 31, 60, and maybe 100.  That's jumps of -40%, 100%, 20%, 100%, 50%.

That's not smooth or exponential by any stretch.  It speaks of discrete jumps in capabilities (reusability, new pads, etc) and very slow increases in between due to optimization.

Without new pads, I agree they'll cross 100, but I don't expect more than 150 F9 ones.

I also don't expect more F9 pads.  I think that's just a hedge in case there's a major problem with SS.

Smooth curves through noisy data don't have smooth jumps between data points.

[meekGee]

Again, spaceX yearly launches were 21, 13, 26, 31, 60, and maybe 100.  That's jumps of -40%, 100%, 20%, 100%, 50%.

That's not smooth or exponential by any stretch.  It speaks of discrete jumps in capabilities (reusability, new pads, etc) and very slow increases in between due to optimization.

Without new pads, I agree they'll cross 100, but I don't expect more than 150 F9 ones.

I also don't expect more F9 pads.  I think that's just a hedge in case there's a major problem with SS.

Smooth curves through noisy data don't have smooth jumps between data points.

Yeah but I'm looking at entire years now, if anything it'll even out noisy months. You don't have to do it on whole year boundaries either - you can do the same exercise at months increments, using a 12 month sliding window.

To help your argument, suppose we discard the old data since it's from before reusability really kicked in, so only look at the mature F9R, the last three years.
We see a giant leap of 100% followed by maybe a 50% leap.  That's not exponential either.

Pick your method.

There's no reason for it to be exponential, and there's no sign of it in the data.

What there is, is continuing growth. People use "exponential" to describe anything that keeps growing, but honestly even linear growth would be astounding for a market described by all the old players as "hard capped".


(Btw linear growth would mean a fixed delta from year to year.  If SpaceX hits 100 this year and 140 next year, that would be linear growth:  Fixed growth by launch count but smaller if looked at as %.  That is why the 31 -- > 60 growth looked so large...  same approximate delta, but in % it looks bigger)

[oldAtlas_Eguy]

As noted by me above. Growth will continue but as the changeover occurs between F9/FH dominated launch numbers to Starship dominated launch numbers. The growth rate will be a complex polynomial with one (F9/FH) showing at first a slowing of increase and then starting a decline toward 0. Which will itself not be a simple curve. The other will be Starship which will have the same early F9 "noisy" rate increases requiring many years. Of 5 or more to even get a value that does not significantly change every new year of launches. Most of that is the Starship itself development but also in adding of additional pads.

So basically it is going to increase ON AVERAGE (when 3 or more years are evaluated) year to year but the rate of change will in itself be a complex function. With fifth or sixth level derivative values. Something on the level of a launch vehicle flight distance traveled modeling during engine burns through multiple stages.

[Zed_Noir]

With current assets: ASDS and pads. SpaceX can do a maximum of 137 launches /yr. That is 8 day turnaround per ASDS, 3 ASDS and all launches using ASDS. For pads it is a maximum of 219 launches /yr. That is 5 day turnaround per pad with 3 pads and 82 of the 219 launches being RTLS.

This suggest even now we are not significantly close to the max realistic launch rate for F9.

SpaceX is supposedly in process of building/converting 2 new pads/sites one at CC and one at VSFB.  If they also start building 2 more ASDS, the maximum then jumps for the all ASDS launches to 228. For optimal mix of ASDS and RTLS that becomes 365. One a Day.
<snip>

SpaceX could expended older Falcon 9's to get about 10 more launches annually without needing ASDS support. Presuming SpaceX limited their F9 fleet to 20 or maybe 30 flights per booster.

[rockets4life97]

I'm thankful folks used their different methods to make predictions at the beginning of the year. Then we'll be able to evaluate which method worked best at the end of the year and settle this debate.

[alugobi]

I don't think it's likely that they'll build more barges, at least not until they've determined that SS is going to work for Starlink.  If SS stalls, then it might be feasible, as that would keep F9 flying longer than for what they now plan.  But if SS works for Starlink, then F9 numbers should drop off dramatically.

[Robotbeat]

On the other hand, they might also be willing to sacrifice more Falcon 9 boosters on expendable launches once Starship works well, so you could get a short term burst.

[redneck]

On the other hand, they might also be willing to sacrifice more Falcon 9 boosters on expendable launches once Starship works well, so you could get a short term burst.

It would be a very short term boost.  Reusability has seriously decreased the number of boosters required for high launch cadence

[meekGee]

I'm thankful folks used their different methods to make predictions at the beginning of the year. Then we'll be able to evaluate which method worked best at the end of the year and settle this debate.

Tragically, both sides of the current sub-faction of a debate predicted 100, since back then we were battling those who said that the best prediction is that it'll continue at 7/month.

What you're witnessing is a ferocious argument over irrelevance.  We both agree that 100 is a good estimate for this year, and next year more than that.

Beyond 2024 will likely never get tested, and nobody will care whether past data followed more of an exponential or more of a linear curve.

Not even us.

[steveleach]

There's no reason for it to be exponential, and there's no sign of it in the data.

There is when I look at it.

I've plotted expected vs actuals for every month from January 2022 to August 2023 (including the 2 planned for later this week), with the expected starting at 4 (same as actual) and growing by a factor of 1.042 per month (which is the same as RobotBeat's 1.64 per year).

When I add an exponential trend line to the actuals I get a curve that tracks the expected values really closely, and obviously the expected growth is exponential by definition with this model.

[meekGee]

There's no reason for it to be exponential, and there's no sign of it in the data.

There is when I look at it.

I've plotted expected vs actuals for every month from January 2022 to August 2023 (including the 2 planned for later this week), with the expected starting at 4 (same as actual) and growing by a factor of 1.042 per month (which is the same as RobotBeat's 1.64 per year).

When I add an exponential trend line to the actuals I get a curve that tracks the expected values really closely, and obviously the expected growth is exponential by definition with this model.

That thing looks like a star constellation... 

You could just as easily have drawn a straight line through that scattering of points, or a swan.

Going from 60 to 100 (which is what you're covering) can be thought of as +55% or +40 flights.  Maybe closer to 50% if they miss 100.

Go back to 2021.  They went from 30 to 60.  That's +100% or +30 flights.

Clearly the % change dropped in half.  That's not noise, since it cumulative over a year.

Which means that the straight line fit is much better over the entire 3 year period.

I'd stop at 2021 since 2020 had a virtual launch stoppage for almost half a year, so is artificially low, lower by a lot than 2019.

[alugobi]

On the other hand, they might also be willing to sacrifice more Falcon 9 boosters on expendable launches once Starship works well, so you could get a short term burst.

I don't think that they'd do that just to get launch numbers, though. But we might see 10xx.21 go into the drink when they get to that point in a booster's history. Or .25.

[Robotbeat]

If the argument is that long term, the trend will not be exponential, I might agree with that. It depends on Starship.

Falcon 9 will necessarily reach some sort of plateau as it’s not fully reusable so it won’t be able to keep dropping in costs.

I kind of expect SpaceX in 2024 to have less than 164 flights, but they may well maintain exponential growth in *mass launched* either way. It depends on how Starship does.

[DanClemmensen]

On the other hand, they might also be willing to sacrifice more Falcon 9 boosters on expendable launches once Starship works well, so you could get a short term burst.

I don't think that they'd do that just to get launch numbers, though. But we might see 10xx.21 go into the drink when they get to that point in a booster's history. Or .25.

If SpaceX achieves its stated goal of making a Starship launch cheaper than an F9 launch for all normal payloads, then F9 will be used only for the remaining contracted Dragon missions. At the point when the number of active boosters exceeds the number of remaining Dragon missions (plus one contingency) they will expend the remaining boosters and retire the ASDSs. Currently there are 7 more contracted CCP, 5 non-CCP crewed Dragon, and 7 contracted CRS Cargo Dragon missions for a total of 19. My guess is that Starship will fully replace all non-Dragon missions in 2027 or 2028. My totally wild guess: the number of remaining Dragon missions in 2028 be three remaining CCP, six remaining CRS, and 2 remaining crewed non-CCP a total of 11. (Adjustments: add three CCP if Starliner becomes unavailable, add three if Cygnus becomes unavailable, subtract some if Starship is allowed to take cargo to ISS).

Now consider FH. FH is NSSL-certified. When will Starship be NSSL-certified? My guess: 2027, so it will EOL before F9.

Therefore, my totally wild guess prediction is that if they still have 12 active boosters in January 2028, They will retire the ASDSs, cease booster production, expend one on each Dragon flight, and put the last one (the contingency reserve) in a museum.

[steveleach]

There's no reason for it to be exponential, and there's no sign of it in the data.

There is when I look at it.

I've plotted expected vs actuals for every month from January 2022 to August 2023 (including the 2 planned for later this week), with the expected starting at 4 (same as actual) and growing by a factor of 1.042 per month (which is the same as RobotBeat's 1.64 per year).

When I add an exponential trend line to the actuals I get a curve that tracks the expected values really closely, and obviously the expected growth is exponential by definition with this model.

That thing looks like a star constellation... 

You could just as easily have drawn a straight line through that scattering of points, or a swan.

Going from 60 to 100 (which is what you're covering) can be thought of as +55% or +40 flights.  Maybe closer to 50% if they miss 100.

Go back to 2021.  They went from 30 to 60.  That's +100% or +30 flights.

Clearly the % change dropped in half.  That's not noise, since it cumulative over a year.

Which means that the straight line fit is much better over the entire 3 year period.

I'd stop at 2021 since 2020 had a virtual launch stoppage for almost half a year, so is artificially low, lower by a lot than 2019.

Yep, of course you can draw any line you want through those points. That doesn't change the fact that a exponential growth at x1.042 per month fits the data quite well, and so probably has at least as much predictive power as anything else. And it's prediction is still around 100 for 2023.

Here's the same thing but projected back a year to the start of 2021...

[meekGee]

Yep, of course you can draw any line you want through those points. That doesn't change the fact that a exponential growth at x1.042 per month fits the data quite well, and so probably has at least as much predictive power as anything else. And it's prediction is still around 100 for 2023.

Here's the same thing but projected back a year to the start of 2021...

You say "quote well", but it doesn't look like it.

Your own numbers:
x1.042 per month means 64% growth per year.  (1.042^12)
Let's look at cumulative years, so we're not thrown off by random months.

2021 -> 2022, 31 launches to 61, so a 96% growth.  Real life growth overshot the exponential prediction by 1.5
2022 -> 2023, 61 launches to 100(?), so 64%.  Bingo!

Guess what?  The x1.042 is simply taking the 2022->2023 growth and taking the 12th root.
And then saying "hey, it fits"

You can do the same trick for the 2021-2023 period - take 100/31, and take the 24th root, and get 1.05x per month.
Again you can say "hey it fits", but you'll find out that you underestimate the first period and overestimate the second.

You know why?  Because it's not exponential.  Exponential growth is constant in it's % growth, so the more you have the more you grow, proportionally. 

The data is not like that. Sadly, this is not what we're seeing.   31->61->100 is not a geometrical series (aka "exponential growth").  It is closer to an arithmetic series. (aka "linear growth")

Shrug.  This is so basic.  Have at it, I can't explain it any better.

[steveleach]

Shrug.  This is so basic.  Have at it, I can't explain it any better.

Yep, we're just arguing past each other now.