Re: Freight car Distribution

Tim O'Connor

Jerry, I think I pointed out before that a very reasonable
approach is simply software simulation that makes car assignments
based on the AAR rules laid down in every ORER. Take an imaginary
fleet of 1,000 box cars with say, 300 of them home road, and then
program in your loads out and loads in to your 'system' and your
destination points (interchanges etc) and let the rules work their
magic. Years ago I wrote switchlist software that did this and
generated train movements, and I ran it hundreds of times observing
train consists. Over time you notice regular patterns but also a
regular occurence of unusual events (like 4 produce reefers sent up
on a branchline local, that occurred once out of scores of trips)
And if your imaginary fleet begins with an "ORER apportionment" by
prototype ownership numbers, then the observations you see will
either support or refute the G-N thesis.

Notice that all simulations of all railroads could start with the
same "imaginary fleet". But the RESULTS for each railroad could be
quite different, because of the loads in/loads out, and connections
(and region of the country).

Anyway, absent data, IMO the only way to operate realistically is
to operate statistically (randomness & rules) and to avoid tedium,
to use switchlist software that embodies it.

Tim O'Connor

I hate to rekindle a dormant debate, but I'm in favor of revisitng the Gilbert and Nelson work on general service cars.

I don't challenge the methodology or the conclusions of their original work. But, the limitation is in exactly what Ben has noted: a sample confined to UP and Southern conductor's books.

I'm certain that many more sources of primary data have been uncovered than we knew to exist only a few years ago. I have some in my own collection. I haven't analyzed it rigorously, but it's apparent that it doesn't follow the same trends that Gilbert and Nelson found in the UP and Southern data.

The question should be: Would the Gilbert and Nelson conclusions hold up under the weight of additional data, or not?

Jerry Dziedzic
Pattenburg, NJ

Join to automatically receive all group messages.