Mike Brock <brockm@...>
Dave Nelson writes:
"I'll repeat myself: Tim and I were doing analysis of railroads, not model
railroads. Our hypothesis on the distribution of ordinary boxcars was for
real world data. That it might have some bearing on what an owner of a
model railroad could do has always been a bit problematic, if, for no other
reason, the huge reduction in the sample size."
One thing that I have been arguing for is that the process one chooses to develop a box car fleet [ other fleets for other types of cars will be needed ] is that such a fleet needs to be capable of producing trains that match...with compression...actual trains. My conception of the N-G theory is that it shows box car populations over a very long time period...perhaps a yr. The problem I have is, will it address train consists?
Consider: I model only a tiny part of Sherman Hill and only about 3 hrs of traffic during one op session. 8 frt trains in all. I compress both the layout and the frt trains...from 75 to 30 cars. About 100 box cars....or 12 per train if they were applied equally which they were not. Using the % of the entire box car fleet, I should have 4 SP box cars. However, in order to model the infamous X4005 train with its 36 SP box cars...very close to 50% of the known train...I would need 15 SP box cars [ 50% of a 30 car train ]. To do that and stay within the N-G envelope of 4% of the national fleet, I would need 375 box cars. But...I will need more SP box cars to populate other trains so I'll go with 20 SP box cars. I now need 500 box cars. The question then is...how do I apply these 500 box cars to my 8 frt trains and their average of 12 per train? In any session the number of box cars should be about 100...leaving 400 in their boxes.
What does this do for the MWR car? Well, nothing. The 1953 ORER shows 100 MWR box cars. For me to have one MWR box car and match the N-G data base, I would need 7400 box cars. Anyone think I'll make it? Incidentally, using the 500 box car data base, I could not use a box car if a RR had less than 1480 box cars. Whew. SP&S barely makes it in. Tucson, Cornelia, and Gila Bend with their 3 cars missed the cut along with the Montana, Wyoming & Southern...which might be over represented due to their close proximity to the UP except for the fact that they had no box cars.
BTW, will I have to run 7399 box cars before the MWR car can show? Of course not. It might be the first...or 640th...car. However, you can also toss 24 7's in a row at a crap table in a Las Vegas Casino except for the fact that, after the 12th such toss, your toss might be a bit off course since your fingers...among other items...will all be broken. Still...there is no reason why...
Yes. Modeling the UP on Sherman Hill over a three hour time period with its 35 frt trains doesn't match well with, for example, a yr's worth of data. Then, compress the 8 trains by 60%...from about 75 cars to 30.
my 100 box cars, I cannot generate a
compressed version of X4005 and its 36 SP box cars. I mean...compressing by
60%, I still need 14 SP box cars and the SP national % of 4% only gives me 4
of my 100 total not allowing that. I would, in fact, need 350 box cars to
give me enough to produce the train in question. And, since I would need
other SP box cars in frt trains...say 20 in total...I would need about 500
box cars. In that case, I would have to spread the cars...do op sessions
with different cars. The trouble with that is that I cannot just randomly
select cars. I still only have 8 frt trains and I can only apply about 1/5
of the fleet to the session.
As is obvious, model RRs compress everything. Compressing the box car fleet
always argued against the inclusion of Canadian marked cars... Or if
included, to set their railroads contribution to distributed fleet at 10% so
as to match the FACT that only 10% of home road cars loaded in Canada were
sent south of the border and that the law required them to be returned to
Canada quite directly.
Not expressed, but IMO a reasonable addition, would be to add something to
take into account the complete ownership of one road by another, such as the
SN by the WP, the T&NO by the SP, and yes, perhaps even the CV by the CN.
All of that slices out of the picture a lot of locations and in some cases,
a lot of cars (i.e., counting CN cars on the CV as home road).
So right off the bat, Jack Burgess and his YV were out of scope. As were
all the PRR boxcars in Enola Yard. I'm not familiar w/ the Ball Line
Route... But I want to ask: Were those cars part of the common carrier fleet
or were they private road cars? One would expect a different movement
pattern for the later.
In recent years I have done futher analysis on the Western Pacific traffic
and seen a very distinct pattern of a large number of loaded boxcars
terminating in the SF Bay Area and correspondingly fewer outbounds.
Slightly more than a 2:1 ratio. Thinking on that fact led me to realize the
WP had little need for storing ordinary, home road boxcars in protective
service in this area as they were "blessed" w/ a a generous supply provided
by everyone else. Which led to the notion that in areas where the
opposite traffic conditions were true... Such, places like, say, Modesto,
CA, the opposite conditions for protective service would also be true. If
true, then there may have been a fair number of SP and ATSF boxcars sitting
aound in and near Modesto... from which some could have been loaned to the
YV when asked for.
Moving on... As for the interesting analysis of how to manage a modeled
freight car fleet... I agree the correct procedure is to reserve some
portion of the foreign road boxcars -- I thought in the 10-20% range of the
owners nominal count of on-layout foreign road cars, for double or tripple
the number of cars that percentage calculated to, but bearing the marks of
smaller roads, cycling them in and out regularly. The only problems I see
with that approach is storage, extra handling, and the bother of having to
do it. But the theory is sound and the variety it provides, both visually
and in the pleasure of collecting them, could still make it worthwhile.
All of which adds to the difficulty of applying the hypothesis.
OTOH, for those few of us doing V-Scale -- that is computer sim railroading
-- we have the opportuity to do urban sites 1:1 (as I happen to be
modeling), to run 60, 70, 80 car freights (as I do), and to run over an
entire division of 100+ miles, with each town represented in full (as I
happen to be modeling). IOW, the difficulties of applying the distribution
to this method of modeling a setting disappears entirely.
All that said, I'll still stick to the premise that, absent historical data
for the site being modeled, the distribution model we offered is a pretty
decent place to start, whether one is modeling in plastic, resin, or
brass... or pixels.