After several days of reading various views on the value of the G-N
hypothesis AS FAR AS PROVIDING A GUIDE FOR BUILDING A BOX CAR FLEET, I think the following applies...not unlike Charlie
Vlk's conclusion.

First, a small minority [ including myself ] of modelers on the STMFC have
access to actual frt train consists. There are those that cover the UP from
Laramie to Rawlins in 1938, 1947, 1949, 1951, and for Cheyenne to
Laramie...1956. Given the relatively small number of trains covered...34 during about a month in
the 1949 book...compared to the number operating [ the ratio is about
1/35 ], the consist information simply is too small to reach definitive
conclusions about the other 34 trains in a given day...let alone what
happened during the other 11 months. OTOH, the G-N hypothesis does give a
possible population over the long run if one makes certain assumptions...for
example, that "away" cars were more or less distributed randomly among other RR's. The actual
train consists appear to deviate considerably at times from the G-N
hypothesis [ the infamous SP cars for example but others as well ] but,
again, this is a short run view.

Those that have actual consists might like to model them even with
compression. Uh oh. For those few who don't know [ surely they care <G> ], I have video of a UP frt train pulling 36 SP box cars in a train of 96 cars. That is 37.5% SP box cars. OK...if I assume a max train length [ compression due to layout size ] of 35 cars, the train will need 13 SP box cars and, given that I compress to 8 frt trains down from 35 in a day, I'll probably need about 6 more SP box cars so that they are represented in the other 7 trains. That means that I will need 19 SP box cars in my fleet. However, SP box cars only represent 4% of the total national fleet. So, if I use the G-N hypothesis as a guide I'll need about
475 box cars in order for me to provide the SP cars. However, I
find that an EJ&E box car is in one of the trains in my
frt conductor book. EJ&E box cars represent only .002 of the national fleet. Hence, [ 0.95 cars ] I can have almost one car [ well, I'll leave off one of the cut levers ]. However, there is an FEC box car showing in a video near Hermosa Tunnel in 1953. FEC had 190 box cars...or 0.0002 of the national fleet. My 475 car population will provide me with 0.095 cars. That might get me 4 ribs on an end. Nooooo problem. I'll just increase my box car population to 4750. That will just about get me the car [ again, missing a cut lever ] and frequent visits to a therapist in Orlando.

The G-N theory is interesting and useful because it tells us
that box cars populated the various RR's in proportions somewhat similar to the national fleet over "long" periods of time. It doesn't appear to help much in modeling trains because trains often were "designed"
for specific tasks [ a lumber train, for instance ] and such a train might draw cars from a year's population at rates not similar to other trains. IOW, a model lumber train on the UP
originating from Cal might require 15 SP box cars while a model merchandise train from the northwest might require no SP box cars. Those of us with frt train consists are fortunate IF we are modelers modeling real trains...even those compressed...because we can simply acquire the cars we need [ or try ]. Fortunately for me I only have consists for about 70 trains in two different yrs. Those without such consists are left to use the G-N hypothesis to provide a frt car population. I don't know what one without consists would do to build individual trains...I suppose some manipulation of the G-N data. It can be done, of course, but the frt train consists I have access to seem to show trains of cars NOT in the mode of distribution that G-N proposes. For example, my frt conductor book shows a train of 40 MTY reefers...all but 3 PFE, 2 box cars in the middle of the reefers [ one loaded GM&O going to Butte, MT ], 2 tank cars and 26 box cars loaded with bauxite ore headed to the Northwest...not one UP box car [ or PA ] in the mix. There were 4 CN box cars, however, and 3 Frisco, 3 Q and, of course, the required NP. Not exactly your national fleet. I would also think that some sort of "small number" factor would need to be used with the G-N. How else to get that FEC car?

Mike Brock