[JSR 669, 2012]

Launch Vehicle Statistics
--------------------------

The practice, of which I and others have long been guilty, of evaluating launch
vehicle reliability based on a simple pass/fail value for each launch is rather a
blunt instrument, as the recent Falcon 9 launch shows.

I have revised my launch tables (http://planet4589.org/space/lvdb/) 
to include an attempt at reasonably objective fractional success values for marginal cases. 
For pass/fail purposes I consider a score of 0.75 or less to be a failure; one could argue for 
lowering that boundary a little bit.

For launches with a single payload, or multiple equal-priority payloads, I give:
 - full success                   1.00
 - orbit usable but not nominal   0.75
 - orbit but not a usable one     0.40
 - payload failed to separate     0.25  (even if good orbit)
 - orbit not reached              0.00  (or reentry after circa 1 orbit)

For missions with primary (P) and secondary (S) payloads, a rough scaling
to give the P 3 times the weight of S -
 - full success                   1.00
 - S off-nominal orbit            0.95
 - S unusable orbit               0.85
 - S failed to sep                0.75
 - P off-nominal orbit            0.55
 - P unusable orbit               0.30
 - P failed to separate           0.10

There's still some subjectivity here, and I've allowed myself to assign intermediate values,
e.g. when an orbit is only slightly off-nominal. Now obviously scores of 0.40 or less
are going to mean an unhappy customer, but I think it's still worth distinguishing from complete
failure to orbit a it usually indicates a vehicle which is 'close' to working in contrast
to some vehicles which never make it beyond first or second stage burn (I'm looking at you,
North Korea...). For Earth escape missions stranded in LEO, I've kept a 0 (complete failure) score.

Here are the proposed scores assigned so far that differ from 0.0 and 1.0:
 0.25  96-061 Pegasus
 0.40  63-021 Thor Agena; 67-032 Proton; 76-062,76-088, 80-031, 86-075,90-055 Molniya, 78-119, 95-052 Kosmos,
       84-120, 04-052 Tsiklon, 91-051 Pegasus, 95-U01 Mu-3S-II, 96-048 CZ-3, 80-043 Atlas,
       99-017, 99-023 Titan, 99-024 Delta 3, 06-006, 08-011, 11-045, 12-044 Proton/Briz, 11-005 Rokot
 0.45  04-050 Delta 4H  (primary payload medium-bad orbit, secondary failed to orbit)
 0.50  01-029 Ariane 5/V142
 0.75  97-057 PSLV, 97-066 Ariane 502, 07-027 Atlas V/NROL-30, 09-029 Soyuz/Meridian
 0.80  00-048 Delta 3, 01-015 GSLV  (somewhat off-nominal orbit)
 0.85  12-054 Falcon 9  (primary perfect, secondary unusable)

I haven't done a throrough scrub of the database, particularly the older launches - let me know
what you think. Note that I don't count PAM and IUS payloads on Shuttle as part of the launch vehicle.
There's a whole other discussion to be had about measuring each stage instead of the LV
as a whole, and measuring upper stage and apogee motor reliability - but that is not this discussion:
the question of integrated launch vehicle reliability comes up often enough to be worth doing better.

Orbital Launch Stats 2012 to Date
---------------------------------

Total 60 attempts:
Russia 18, US 14, China 14, France/ESA 7, India 2, Japan 2, Iran 1 + 1 fail, North Korea 1 fail

Sea Launch counted as US, Soyuz/CSG counted as France, so for country of LV manufacture it's
Russia/Ukraine 21, China 14, US 12, Europe 6, India 2, Japan 2, Iran 2, NK 1. Orbital's Antares
and South Korea's Naro will make it even tougher to construct meaningful nation-based statistics,
since they have Russian-built first stages.