Another post? Wow!
While I was researching hopper rosters I was looking at the various ways John Nehrich has broken down the 1949 fleet of freight cars on the RPI site. Among this data was an interesting tidbit - in 1949 there were approximately 60,000 36' cars on the railroads. Of those, nearly 2/3 are CN and CP Fowler Patent single sheathed box cars. Leaving just 20,000 or so 36' double sheathed cars. About 5,000 of those are ACL and SAL ventilated box cars.
So not many 36' double sheathed cars at all.
Chris pointed out that there would have been more in 1947. So I was curious, and rather than keeping this between the two of us (and also so this info is someplace where I can find it later...), here's what I've compiled (with much thanks for John's research).
The percentage of composite to all steel box cars went from 56% in 1940 to 44% in 1945 and 31% in 1949.
This amounts to about 106,000 cars of a total box car roster of about 700,000+. And that's assuming all of those cars were 36 footers.
There were still 26,000+ all wood box cars in 1945, with about 7,500 'other' box cars in 1949, so that reduction was most likely all double sheathed cars since I'm not aware of any single sheathed car being made without a steel underframe. And I'm not aware of many 40' all wood cars.
In 1949 there were 60,000 of 223,0000 composite cars were 36' box cars, or about a 1/4. But of those, 40,000 were Fowler patent single sheathed cars. Leaving about 20,000 double sheathed plus the 7,500 'other' cars. That amounts to about 3% of the total box car fleet being under 40' and double sheathed.
So if all of the difference was double sheathed, and the number of Fowler patent cars remained constant, then 36' double sheathed cars would account for about 120,000 of 160,000 composite cars, plus most of the 'other' cars, which amounts to only 19% of the fleet as 36' cars.
Most likely the drop off accelerated instead of being linear. So it's somewhere between 19% and 3%, so 10% of the entire box car fleet as 36' double sheathed cars sounds reasonable. But I'm guessing it's lower.
In 1949, the roads with more than 1,000 cars (not including the CN and CP since they were all single sheathed) - Southern, L&N, NC&StL, D&H, ERIE, and Nationales de Mexico. Adding local roads that had at least 500 there's BAR and DL&W.
ORER Data from 1947
Comparing a few roads between 1947 and 1949:
SOU - 9,228 - 3,928
L&N - 7,438 - 2,829
D&H - 1,751 - 1,560
ERIE - 1,275 - 1,005
BAR - 1,099 - 903
DL&W - 832 - 538
RDG - 2,135 - 223
LNE - 190 - 183 (but these are 39' 10")
NYC - 752 - 156
In 1947 5,407 of these cars were ACL or SAL ventilated box cars.
So obviously some roads were retiring cars faster than others. But I think that up to 10% of box cars being 36' is reasonable.
So how many 36' double sheathed cars do we need? Well, I think that the NH cars being home road cars don't count. We'll need a couple of them for revenue service, and a few more for MOW service.
I think that specialty cars like the ventilated box cars also don't count, provided we're modeling a season when they would be in use as ventilated box cars. If they are in use as regular box cars then they would be treated as such.
So I think the must have cars are those for direct connections with a decent percentage of 36' double sheathed cars (CV and RDG are both around 25% in 1947). The SOU SU design is also one that was such a large class of cars, and the Southern had the largest number of 36' double sheathed cars even in 1949.
As for other cars I think it depends on whether we identify commodities coming for other roads. Assuming there's a model available, then it's a question of how those cars relate to their larger box car fleet. But these would be low priority cars.
So for Chris' layout this amounts to, well, one 36' double sheathed box car. He's got two trains with 20-25 cars each, and perhaps 30ish cars on the layout. So less than 100 cars.
In my case, I've got 8 trains of 15-25 cars each, plus another 50 or 60 on the layout. So at any given time I'm probably looking at 2-3 36' double sheathed cars in any given session. Of course, on some days there could be more, others less.
F&C has the most 36' double sheathed car models. I've got the CV, NH and NYC
I also have SOU (Westerfield, but F&C makes the same prototype).
The NYC car is a great example of a rare car. Their 1947 roster of 752 36' cars is only 3% of their USRA design steel box cars (25,000+), just one class of steel box cars on the road, or about 1% of their 60,000+ box car fleet.
Roster vs Operating Session
Having said that, I go back to my assertion that there's a difference between your model roster, and the cars you run during a given operating session. I'll use Chris's layout as an example. Until he has an operating Shore Line, he needs about 100 cars for an operating session, probably less. So by our calculations he needs a single 36' double sheathed box car. And that's assuming (as we will for now) that all 100 cars per session are box cars.
So looking at the choices, he decides he likes the CV one. If he has an industry that is directly serviced by an industry on the CV then this is a good choice. I might even argue that it's no longer a rare car, as that industry would have about a 25% of receiving this type of car based on the CV roster (assuming they never use a foreign road car).
Problem is, if this isn't a car in semi-captive service, then this 'rare' car becomes 'common' if we choose to run one 36' car per session and that's the only car available.
So I think that it's worth modeling at least a handful of cars. This is why your roster should be larger than what you'd run in an operating session (or a couple of sessions).
Instead of running one 36' double sheathed box car per 100 cars per session (which would be a statistical anomaly anyway), these cars would simply fall under the 'rare' category. Rare cars could make up 10-20% of the cars used in the session. If we go with 20% then 2 cars of his 100 would be rare. Sometimes one would be a 36' car, sometimes none, sometimes two. In that mix are other rare cars.
One of the main reasons I like this approach is that I like freight cars. It allows me to have more cars, and more unusual cars, than most people would choose to model. But I also think it means that from session to session the mix of cars is good, and individual rare cars continue to be rare.
My initial approach was to model one car per 'x' number of a given car on the prototype. But for large classes of cars (PRR X29 for example) this would require 20+ cars. These are cars that belong in pretty much every ops session. But I can make that happen with far fewer cars. Folks won't remember car numbers from session to session, but they will remember unusual cars.
So now I'm working on finding an upper limit to the number of cars needed for any large class. This will probably vary depending on your prototype (you might need 25 X29s if you model the PRR). It's important to have enough common cars, whether a specific class, or a specific design used by many roads (1937 AAR Standard Box Car).
I suppose that's a next step - figure out what a representative mix of cars is for each session. That's for another post.