The G-N Distribution Model


Schuyler Larrabee
 

Gize,

 

I wasn’t very interested in the Gilbert-Nelson Distribution Model when it was being developed.  I didn’t have a layout then (still don’t, but maybe . . .) so I satisfied my Freight Car building urges by building kits that struck my eyes as being interesting.  Since those times I’ve settled (more or less) on modeling the ERIE in the 1950-52 time frame, and thus the G-N Model has become of more interest to me.

 

But is the G-N Model neatly summarized anywhere?

 

And where might I find that?

 

Schuyler

 


Bruce Smith
 

Schuyler,


Yes. Repeatedly. In the archives.  But to make it simple, here it is, again:


Boxcars from any given railroad appear on any other railroad at the same frequency as they occur in the national pool.


That's it. Really pretty simple. Note that it does not govern the frequency of home road cars, or the frequency of cars in individual trains, but the overall frequency of foreign road box cars.  Several caveats deal with cars in assigned or pool service, branch lines in the middle of nowhere, yada, yada, yada... but the essence is very pure and simple.


Regards

Bruce 

Bruce Smith

Auburn, AL


From: STMFC@... on behalf of 'Schuyler Larrabee' schuyler.larrabee@... [STMFC]
Sent: Monday, September 4, 2017 5:21 PM
To: STMFC@...
Subject: [STMFC] The G-N Distribution Model
 


Gize,

 

I wasn’t very interested in the Gilbert-Nelson Distribution Model when it was being developed.  I didn’t have a layout then (still don’t, but maybe . . .) so I satisfied my Freight Car building urges by building kits that struck my eyes as being interesting.  Since those times I’ve settled (more or less) on modeling the ERIE in the 1950-52 time frame, and thus the G-N Model has become of more interest to me.

 

But is the G-N Model neatly summarized anywhere?

 

And where might I find that?

 

Schuyler

 




Tim O'Connor
 

Schuyler

A summary??? LOL !!

I have over 1,500 saved emails I could send to you ... :-)

Or just the 578 emails from Tim Gilbert, and about a hundred from Dave Nelson?

My summary for your time frame: box car and flat car and mill gondolas distribution
is proportional to the fleet size of every railroad, i.e. the model is that those cars
roamed freely and randomly. Then you add the spice of traffic for your railroad: reefers,
stock cars, coal cars, coal and stone gondolas, tank cars. Home road cars of the latter
variety (livestock, coal, stone) tended to stay near home.

The email archive includes hundreds of special cases, exceptions, and variations in
time and space - and scores of known consists from the 40's and 50's.

Tim




Gize,
 
I wasn�t very interested in the Gilbert-Nelson Distribution Model when it was being developed.  I didn�t have a layout then (still don�t, but maybe . . .) so I satisfied my Freight Car building urges by building kits that struck my eyes as being interesting.  Since those times I�ve settled (more or less) on modeling the ERIE in the 1950-52 time frame, and thus the G-N Model has become of more interest to me.
 
But is the G-N Model neatly summarized anywhere?
 
And where might I find that?
 
Schuyler


Bruce Smith
 

Tim,


And with all that, you got it wrong? :)  The N-G model does NOT apply to mill gondolas.  They are regional in distribution.  Flat cars do conform to the N-G model, at least for some periods.


Regards,

Bruce

Bruce Smith

Auburn, AL


From: STMFC@... on behalf of Tim O'Connor timboconnor@... [STMFC]
Sent: Monday, September 4, 2017 5:57 PM
To: STMFC@...
Subject: Re: [STMFC] The G-N Distribution Model
 


Schuyler

A summary??? LOL !!

I have over 1,500 saved emails I could send to you ... :-)

Or just the 578 emails from Tim Gilbert, and about a hundred from Dave Nelson?

My summary for your time frame: box car and flat car and mill gondolas distribution
is proportional to the fleet size of every railroad, i.e. the model is that those cars
roamed freely and randomly. Then you add the spice of traffic for your railroad: reefers,
stock cars, coal cars, coal and stone gondolas, tank cars. Home road cars of the latter
variety (livestock, coal, stone) tended to stay near home.

The email archive includes hundreds of special cases, exceptions, and variations in
time and space - and scores of known consists from the 40's and 50's.

Tim




Gize,
 
I wasn?t very interested in the Gilbert-Nelson Distribution Model when it was being developed.  I didn?t have a layout then (still don?t, but maybe . . .) so I satisfied my Freight Car building urges by building kits that struck my eyes as being interesting.  Since those times I?ve settled (more or less) on modeling the ERIE in the 1950-52 time frame, and thus the G-N Model has become of more interest to me.
 
But is the G-N Model neatly summarized anywhere?
 
And where might I find that?
 
Schuyler



Claus Schlund \(HGM\)
 

Hi List Members,

Like Schuyler, I too would be interested in reading a rigorous summary of the N-G distribution model. But it sounds like answering Schuyler's question is not as simple as it might appear to be at first glance.

Perhaps a scholarly work of this level of importance to us should have an official repository of its findings somewhere. I would suggest something like a web page could be appropriate, perhaps.

Furthermore, it would seem to me that providing the contents of this repository perhaps should not be the job of anyone other than the Gilbert-Nelson team members themselves. That way they get to speak for the work they have done. If it is not them, then it is a bunch of other folks interpreting the work of the N-G team. This will inevitably lead to the introduction of errors, inaccuracies, and departures from the intent of the N-G team.

Suggesting that someone seeking a summary might want to inspect "over 1,500 saved emails" or "just the 578 emails from Tim Gilbert, and about a hundred from Dave Nelson" is effectively saying that no such summary exists in a easily digested or accessible form. Is that in fact the case - has the N-G team not published a rigorous summary of their findings in any one place?

Just my humble thoughts on the topic...

Claus Schlund

----- Original Message -----
From: "'Bruce F. Smith' smithbf@auburn.edu [STMFC]" <STMFC@yahoogroups.com>
To: <STMFC@yahoogroups.com>
Sent: Monday, September 04, 2017 4:08 PM
Subject: Re: [STMFC] The G-N Distribution Model


Tim,


And with all that, you got it wrong? :) The N-G model does NOT apply to mill gondolas. They are regional in distribution. Flat cars do conform to the N-G model, at least for some periods.


Regards,

Bruce

Bruce Smith

Auburn, AL

________________________________
From: STMFC@yahoogroups.com <STMFC@yahoogroups.com> on behalf of Tim O'Connor timboconnor@comcast.net [STMFC] <STMFC@yahoogroups.com>
Sent: Monday, September 4, 2017 5:57 PM
To: STMFC@yahoogroups.com
Subject: Re: [STMFC] The G-N Distribution Model



Schuyler

A summary??? LOL !!

I have over 1,500 saved emails I could send to you ... :-)

Or just the 578 emails from Tim Gilbert, and about a hundred from Dave Nelson?

My summary for your time frame: box car and flat car and mill gondolas distribution
is proportional to the fleet size of every railroad, i.e. the model is that those cars
roamed freely and randomly. Then you add the spice of traffic for your railroad: reefers,
stock cars, coal cars, coal and stone gondolas, tank cars. Home road cars of the latter
variety (livestock, coal, stone) tended to stay near home.

The email archive includes hundreds of special cases, exceptions, and variations in
time and space - and scores of known consists from the 40's and 50's.

Tim




Gize,

I wasn?t very interested in the Gilbert-Nelson Distribution Model when it was being developed. I didn?t have a layout then (still don?t, but maybe . . .) so I satisfied my Freight Car building urges by building kits that struck my eyes as being interesting. Since those times I?ve settled (more or less) on modeling the ERIE in the 1950-52 time frame, and thus the G-N Model has become of more interest to me.

But is the G-N Model neatly summarized anywhere?

And where might I find that?

Schuyler


Benjamin Hom
 

Claus Schlund wrote:
"Furthermore, it would seem to me that providing the contents of this repository perhaps should not be the job of anyone other than the Gilbert-Nelson team members themselves. That way they get to speak for the work they have done. If it is not them, then it is a bunch of other folks interpreting the work of the N-G team. This will inevitably lead to the introduction of errors, inaccuracies, and departures from the intent of the N-G team."

Unfortunately, unless you hire a medium, that only means Dave Nelson.  Sadly, Tim Gilbert has passed on.


Ben Hom


Bruce Smith
 

Claus,

It isn't all that hard... I did it in one sentence ;) Why all those posts and text? Well, if you take the time to read them, then it become obvious that:
1) In spite of the simplicity, some folks misapply the model.
2) In spite of significant evidence in favor of the mode, some folks just don't believe it.
3) There are some specific incidences that may not conform to the model.

Regards
Bruce
Bruce Smith
Auburn, AL
________________________________________
From: STMFC@yahoogroups.com <STMFC@yahoogroups.com> on behalf of 'Claus Schlund' claus@hellgatemodels.com [STMFC] <STMFC@yahoogroups.com>
Sent: Monday, September 4, 2017 6:40 PM
To: STMFC@yahoogroups.com
Subject: Re: [STMFC] The G-N Distribution Model

Hi List Members,

Like Schuyler, I too would be interested in reading a rigorous summary of the N-G distribution model. But it sounds like answering
Schuyler's question is not as simple as it might appear to be at first glance.

Perhaps a scholarly work of this level of importance to us should have an official repository of its findings somewhere. I would
suggest something like a web page could be appropriate, perhaps.

Furthermore, it would seem to me that providing the contents of this repository perhaps should not be the job of anyone other than
the Gilbert-Nelson team members themselves. That way they get to speak for the work they have done. If it is not them, then it is a
bunch of other folks interpreting the work of the N-G team. This will inevitably lead to the introduction of errors, inaccuracies,
and departures from the intent of the N-G team.

Suggesting that someone seeking a summary might want to inspect "over 1,500 saved emails" or "just the 578 emails from Tim Gilbert,
and about a hundred from Dave Nelson" is effectively saying that no such summary exists in a easily digested or accessible form. Is
that in fact the case - has the N-G team not published a rigorous summary of their findings in any one place?

Just my humble thoughts on the topic...

Claus Schlund




----- Original Message -----
From: "'Bruce F. Smith' smithbf@auburn.edu [STMFC]" <STMFC@yahoogroups.com>
To: <STMFC@yahoogroups.com>
Sent: Monday, September 04, 2017 4:08 PM
Subject: Re: [STMFC] The G-N Distribution Model


Tim,


And with all that, you got it wrong? :) The N-G model does NOT apply to mill gondolas. They are regional in distribution. Flat
cars do conform to the N-G model, at least for some periods.


Regards,

Bruce

Bruce Smith

Auburn, AL

________________________________
From: STMFC@yahoogroups.com <STMFC@yahoogroups.com> on behalf of Tim O'Connor timboconnor@comcast.net [STMFC]
<STMFC@yahoogroups.com>
Sent: Monday, September 4, 2017 5:57 PM
To: STMFC@yahoogroups.com
Subject: Re: [STMFC] The G-N Distribution Model



Schuyler

A summary??? LOL !!

I have over 1,500 saved emails I could send to you ... :-)

Or just the 578 emails from Tim Gilbert, and about a hundred from Dave Nelson?

My summary for your time frame: box car and flat car and mill gondolas distribution
is proportional to the fleet size of every railroad, i.e. the model is that those cars
roamed freely and randomly. Then you add the spice of traffic for your railroad: reefers,
stock cars, coal cars, coal and stone gondolas, tank cars. Home road cars of the latter
variety (livestock, coal, stone) tended to stay near home.

The email archive includes hundreds of special cases, exceptions, and variations in
time and space - and scores of known consists from the 40's and 50's.

Tim




Gize,

I wasn?t very interested in the Gilbert-Nelson Distribution Model when it was being developed. I didn?t have a layout then (still
don?t, but maybe . . .) so I satisfied my Freight Car building urges by building kits that struck my eyes as being interesting.
Since those times I?ve settled (more or less) on modeling the ERIE in the 1950-52 time frame, and thus the G-N Model has become of
more interest to me.

But is the G-N Model neatly summarized anywhere?

And where might I find that?

Schuyler






------------------------------------
Posted by: "Claus Schlund" <claus@hellgatemodels.com>
------------------------------------


------------------------------------

Yahoo Groups Links


Schuyler Larrabee
 

Well, thank you Bruce. Yes, you did this in one
sentence. And then instantly provided multiple
"excepts."



And while I think this is a logical extension of
the fact that freight cars are free-rolling, can
be assigned anywhere, look at the trek of the
Monon 1 as an example, on any given day, in any
given train, you can see any given railroad's
freight cars . . . I don't quite believe it.
Sorry. Sorry I asked.



This would suggest that in a train on, for
example, the NP you could, on average, over time,
see the sane, or very nearly the same, consist
makeup as say, on the Southern. Or the UP, Or
the D&H. Or the Rutland, the MEC, the NH, the
M&StL, the CGW, the AA, the NKP, the C&O, the SP .
. .



No wonder Brock obsesses about some damned hopper
going over some hill out west. 8^)



Maybe. But I have a bit of a hard time believing
that the consists of freight trains could be that
homogeneous across the country as that. And that
it would be a good way to establish the mix of
cars you have on your model of the XYZ railroad,
because that's what the G-N Distribution Model
says.



I can easily see how quickly that would get skewed
for any one railroad overall, never mind in
specific areas of a particular road. And I
suppose that's just one of the "excepts."



The ERIE had traffic between steel mills and
rolling mills around Youngstown. So if I were to
be modeling the Youngstown area (it would take a
gymnasium to model that!) I'd need hot metal
bottle cars, as well as more gons than I want to
think about. Plus a lot of hoppers, most likely
from eastern roads. Yeah, well, that's an
"except."



The ERIE had a tremendous produce and fruit market
to serve in NYC, so there were reefer blocks,
solid reefer trains, from Chicago to NYC (with a
few diverted to other eastern cities). I'll need
a lot of reefers. Yeah, well, that's a "except."



Maybe other than those "excepts," the composition
of the general freight pool for the rest of the
trains running over the entire line of the ERIE
might conform to the G-N Distribution Model, but I
am not even so sure of that. It seems to me that
there must have been some level of regional
skewing.



I must say that the responses I've seen from
long-time list members that amount to "weren't you
paying attention?" and offers to send me 1500 of
these emails and also 100 or so from someone else
so I can figure it out for myself certainly don't
reflect the usual level of "here, let me help you
with that" that it typical of this list. I think
the Claus is in much the same boat as I am, which
in effect is "what are you guys talking about?
Please educate me, and put this in an easily
understood form that I can go back to and think
about for my own purposes."



In my first post with this subject line, I asked:
"But is the G-N Model neatly summarized anywhere?"
I'm sorry Bruce, but your one sentence summary
doesn't quite get me there.



Schuyler



From: STMFC@yahoogroups.com
[mailto:STMFC@yahoogroups.com]
Sent: Monday, September 04, 2017 6:50 PM
To: STMFC@yahoogroups.com
Subject: Re: [STMFC] The G-N Distribution Model



Schuyler,

Yes. Repeatedly. In the archives. But to make it
simple, here it is, again:

Boxcars from any given railroad appear on any
other railroad at the same frequency as they occur
in the national pool.

That's it. Really pretty simple. Note that it does
not govern the frequency of home road cars, or the
frequency of cars in individual trains, but the
overall frequency of foreign road box cars.
Several caveats deal with cars in assigned or pool
service, branch lines in the middle of nowhere,
yada, yada, yada... but the essence is very pure
and simple.

Regards

Bruce

Bruce Smith

Auburn, AL

_____

From: STMFC@yahoogroups.com
<STMFC@yahoogroups.com> on behalf of 'Schuyler
Larrabee' schuyler.larrabee@verizon.net [STMFC]
<STMFC@yahoogroups.com>
Sent: Monday, September 4, 2017 5:21 PM
To: STMFC@yahoogroups.com
Subject: [STMFC] The G-N Distribution Model





Gize,



I wasn't very interested in the Gilbert-Nelson
Distribution Model when it was being developed. I
didn't have a layout then (still don't, but maybe
. . .) so I satisfied my Freight Car building
urges by building kits that struck my eyes as
being interesting. Since those times I've settled
(more or less) on modeling the ERIE in the 1950-52
time frame, and thus the G-N Model has become of
more interest to me.



But is the G-N Model neatly summarized anywhere?



And where might I find that?



Schuyler


Tim O'Connor
 


Do we HAVE to go through all of this again Schuyler??

Many people seem to need a refresher class in probability and statistics
before they can understand WHY not all trains look alike.

If you remember even a little bit of high school math, you should remember
that a random distribution of coin tosses (heads or tails) produces a
BINOMIAL DISTRIBUTION (aka the Bell Curve). The chance that two random
freight trains of 100 cars each have exactly the same cars in them is
astronomically small! And THAT is why even though ON AVERAGE freight cars
are distributed according to their share of the fleet, on any given day or
in any given train, there may be very different proportions in that train.

To PROPERLY model based on the G-N model you really need at least 10x or 20x
the number of cars that can possibly be used on your layout. That fleet should
be based on the national percentages. Then write yourself a Basic program that
RANDOMLY chooses box cars to enter the layout from your staging yard. You will
notice two things. First, there is an incredible variety of consists. And Second,
over time, you'll notice that the percentages of cars follow the G-N model. QED.

Tim O'Connor



This would suggest that in a train on, for
example, the NP you could, on average, over time,
see the sane, or very nearly the same, consist
makeup as say, on the Southern.  Or the UP,  Or
the D&H.  Or the Rutland, the MEC, the NH, the
M&StL, the CGW, the AA, the NKP, the C&O, the SP .


Tim O'Connor
 

Bruce

You say, but I disagree. I have enough pictures of PRR gondolas in California,
Montana and the west to convince myself that they were found in proportion to
their population. Of course, since there were not many loads available for them
in California, they tended to go home quickly. But that does NOT change the
model at all. Do the math. The Southern Pacific owned a tiny number of mill
gondolas compared to its entire fleet, and you would be very unlikely to see
one in Pennsylvania, simply because they were incredibly rare compared to the
PRR's fleet. Heck, they were rare in SP freight trains too!

Tim O'Connor



And with all that, you got it wrong? :)  The N-G model does NOT apply to mill gondolas.  They are regional in distribution.  Flat cars do conform to the N-G model, at least for some periods.

Bruce Smith


Schuyler Larrabee
 

Tim O'Connor wrote:


Do we HAVE to go through all of this again
Schuyler??

Actually, the point is NOT to go through all this
again. I asked, again, is, the G-N Model neatly
summarized anywhere? I most assuredly do not want
to go through the 1600+ emails you offered.
Bruce's single-sentence summary is helpful but not
entirely convincing. It's TOO simple.

Many people seem to need a refresher class in
probability and statistics
before they can understand WHY not all trains look
alike.

Well, sorry, but my career path didn't require me
to take a class in probability and statistics to
begin with, never mind a refresher. And you know
the quotation that involves statistics. And since
I don't want to assume you know, as you're
assuming I do, the quotation is "Lies, damned
lies, and statistics."

I never said all trains look alike, Of course
not, that's why we loved (in the transition era)
to watch them go by, just about every car was
different, different shape, different roads,
different regions of the country. But the logical
extrapolation of what has been said leads me back
to This would suggest that in a train on, for
example, the NP you could, on average, over time,
see the sane, or very nearly the same, consist
makeup as say, on the Southern. Or the UP, Or
the D&H. Or the Rutland, the MEC, the NH, the
M&StL, the CGW, the AA, the NKP, the C&O, the SP

Notice I didn't say the same train. I said the
same CONSIST MAKEUP, which doesn't even claim that
the assortment of road names represents the same
car types. Hence, a train with all the same
reporting marks could look completely different
than the same set of reporting marks for another
train.

If you remember even a little bit of high school
math, you should remember
that a random distribution of coin tosses (heads
or tails) produces a
BINOMIAL DISTRIBUTION (aka the Bell Curve). The
chance that two random
freight trains of 100 cars each have exactly the
same cars in them is
astronomically small!

Of course it is. Don't be, well, insulting. I
never suggested that. Maybe you are not grasping
what I am saying.

And THAT is why even though ON AVERAGE freight
cars
are distributed according to their share of the
fleet, on any given day or
in any given train, there may be very different
proportions in that train.

I get that. But the G-N Distribution Model has
been getting used as a way to establish the
distribution of cars a modeler should have in ther
complement of cars on their model railroads. It
strikes me that if you were to stand trackside
around the country in the early 50s, for example,
the G-N Model might well be fulfilled, overall, on
average, across the nation. But to extrapolate
(actually, the inverse of extrapolate, a term I
don't know) the Model and use it as a method of
determining the complement of car reporting marks,
car types, and so on for a model railroad
representing a subdivision of a prototype railroad
strikes me a not a solid basis for doing so.


To PROPERLY model based on the G-N model you
really need at least 10x or 20x
the number of cars that can possibly be used on
your layout. That fleet should
be based on the national percentages.

If we were to do that, it's my impression that we
would all have trains made up predominantly of PRR
cars. For most railroads, I think that would be
highly unlikely. Many, sure, but not likely in
proportion to "the national percentages."

Then write yourself a Basic program that
RANDOMLY chooses box cars to enter the layout from
your staging yard. You will
notice two things. First, there is an incredible
variety of consists.

Undoubtedly.

And Second,
over time, you'll notice that the percentages of
cars follow the G-N model. QED.



I don't think that is necessarily true.

Schuyler




Tim O'Connor




This would suggest that in a train on, for
example, the NP you could, on average, over time,
see the sane, or very nearly the same, consist
makeup as say, on the Southern. Or the UP, Or
the D&H. Or the Rutland, the MEC, the NH, the
M&StL, the CGW, the AA, the NKP, the C&O, the SP .


Steve and Barb Hile
 

I think that there are two things under consideration here and it is important to keep them separated.
 
The consist of an individual train is neither random, nor governed by the Gilbert-Nelson model.  It is based on the work that train is intended to do.  Extreme examples include mine runs (loaded and empty) versus a local way freight, versus a through or a drag type freight.  As prototype modelers, we should be considering what our prototype ran for all the various types of freight trains and then using whatever means (car cars and waybills, computer generated switch lists, etc.) to establish the consists.
 
Mike Brock's corollary to the G-N model was that nearly every train should include a Northern Pacific boxcar.
 
Where the Gilbert-Nelson model comes into play is in creating your own pool of freight cars to use in your trains.  Home road cars are exceptions and should constitute a healthy proportion of your collection.  To be realistic, the remaining boxcars (and to some extent flatcars) should be aligned in proportion to the national fleet at the time you are modeling.
 
The G-N model is data based, looking at the consists of multiple trains passing in multiple regions, thanks to preserved conductor's train books.  As much as anything, it was somewhat of a response to the old-time model railroad assumption that suggests about half home road, a good portion of connecting/neighboring roads and a few cars from more distant lines.
 
Personally, I have not yet fully applied the G-N to my fleet, but I'll bet I am not too far off.  I have tried to concentrate on the "signature-type" cars from roads, near and far from my Rock Island theme.
 
Steve Hile



From: STMFC@... [mailto:STMFC@...]
Sent: Monday, September 04, 2017 9:21 PM
To: STMFC@...
Subject: RE: [STMFC] The G-N Distribution Model

 


Do we HAVE to go through all of this again Schuyler??

Many people seem to need a refresher class in probability and statistics
before they can understand WHY not all trains look alike.

If you remember even a little bit of high school math, you should remember
that a random distribution of coin tosses (heads or tails) produces a
BINOMIAL DISTRIBUTION (aka the Bell Curve). The chance that two random
freight trains of 100 cars each have exactly the same cars in them is
astronomically small! And THAT is why even though ON AVERAGE freight cars
are distributed according to their share of the fleet, on any given day or
in any given train, there may be very different proportions in that train.

To PROPERLY model based on the G-N model you really need at least 10x or 20x
the number of cars that can possibly be used on your layout. That fleet should
be based on the national percentages. Then write yourself a Basic program that
RANDOMLY chooses box cars to enter the layout from your staging yard. You will
notice two things. First, there is an incredible variety of consists. And Second,
over time, you'll notice that the percentages of cars follow the G-N model. QED.

Tim O'Connor



This would suggest that in a train on, for
example, the NP you could, on average, over time,
see the sane, or very nearly the same, consist
makeup as say, on the Southern.  Or the UP,  Or
the D&H.  Or the Rutland, the MEC, the NH, the
M&StL, the CGW, the AA, the NKP, the C&O, the SP .


Tony Thompson
 

Schuyler Larrabee wrote:

 

Maybe. But I have a bit of a hard time believing
that the consists of freight trains could be that
homogeneous across the country as that. And that
it would be a good way to establish the mix of
cars you have on your model of the XYZ railroad,
because that's what the G-N Distribution Model
says.


The word "homogeneous" here reveals a deep misunderstanding of statistics. I will be happy to elaborate OFF-LIST, Schuyler, if you wish.

Tony Thompson             Editor, Signature Press, Berkeley, CA
2906 Forest Ave., Berkeley, CA 94705         www.signaturepress.com
(510) 540-6538; e-mail, tony@...
Publishers of books on railroad history






Tony Thompson
 

      I think it's very important to recognize that G-N is not an assertion but is based on data. It certainly works for the cases where it should work, though I sure wish we had lots more data for a more robust conclusion. I found it to work for conductor's time books on the SP Coast Route in 1948-1951, for example (as I reported in my blog).
       As others have observed, OF COURSE it does not work for everything. That's been demonstrated repeatedly. And as for home road car proportions, I would remind the dogmatists that railroads seemed to differ in this area. Conclusions on one railroad may well not apply to another, and of course that can be seasonal as well.
       To me, it's one more tool to help us choose more realistic freight car fleets.

Tony Thompson             Editor, Signature Press, Berkeley, CA
2906 Forest Ave., Berkeley, CA 94705         www.signaturepress.com
(510) 540-6538; e-mail, tony@...
Publishers of books on railroad history






np328
 

     First of all, I can state that I am a great adherent to the N-G model. Everything I have found and studied on my road supported the thesis.

    After studying it awhile though, I have come to find, I cannot say disruptions to the theory, rather if I was talking about music, the existence of back-beats and syncopated rhythms that again, do not render the N-G model invalid, just rather less transparent than a cursory glance might reveal.  

    1) The crops in the US ripen from southwest to northeast across the US. They just do. Not west to east directly, not north to south directly, however in a diagonal across the US. After studying the AAR reports and then USDA reports and charts this jumped out. And from studying for my presentation on Closer to the Railhead, I saw the AAR issuing “home area” car orders in this same manner, about three weeks prior to the expected harvesting time. Car orders would go out to get all Southwestern road cars home first then head across the US in a plow shape towards New England with different states and home roads. This would (I think) gives a ripple in the N-G model as it is human directed input. It certainly is a seasonal deviation.

And another reason Richard Hendrickson’s admonishment that the more we as a modeler narrower the time frame modeled the greater one can apply data correctly is sage advice.

     California in some manner seemed (perhaps) exempt somewhat on a global scale to the AAR car orders ONLY as grain seemed to be the real boxcar driver and crops ripening in CA needed reefers. And that call home seemed to be pretty well handled by the private car fleets like PFE though AAR admonishments are found. Perhaps it is because of the nature of California with large population centers in that loads are terminated and so empties are available in greater proportions than inland. I am open to other commenting on that.

2) My home road, the NP…had a fleet of cars somewhat younger than the national average and/or because of that in better shape. A big commodity on the NP was lumber, dressed lumber and that honestly required cars to be in good shape inside. I submit that this fleet of cars younger and in better shape than the national average made it more likely for those cars to be stolen or kept, human nature being what it is. It helps explain why Mike Brock can state “and an NP boxcar in every photo”. And places another cross ripple in the N-G model. Human interaction (covetousness?) reducing the pure singularity of N-G?   

3) Crops like grain move twice. From the most recent AAR reports I presented on, grain is moved off the farm, and into what the local grain elevators can hold, where it sits until the sale price is acceptable. Then moves again either to port or to major large elevators elswhere. This second move is typically around January/February. Another set of cross currents to contend with. And like point 1, seasonal ones.   

4) Now if I throw in local research, like in my fall 1953 timeframe of traffic between Mpls/St. Paul and Duluth/Superior, I had found that lake coal into the Twin Cities area southbound was the No.1 commodity. I honestly had thought grain north to the ports would overshadow that, but no. Add a steel mill in Duluth with major manufacturing that would use that steel south in the Twin Cities and even more-so in southeastern Wisconsin and upstate Ill, and these all add to layers of complexity on top of the N_G theory.

    After much research, I have a list of the top six southbound commodities and top six northbound commodities for my modeled line in close to my time era. I also have a list of top five commodities for the entire NP Ry, all in my modeled era and these lists are all different. To me this only underlines the value of research.

     It does not however, to me - render the N-G theory invalid. It is a big determinant in how I shop for off-line cars in the LHS, or meets. I can purchase a car with comfort or walk away from a seemingly good deal.

    I do believe that if all points above were stripped away, underneath would be the N-G traffic patterns quietly humming away. However it would be under the layers I listed above.   

Brian J Carlson gave what I thought was a better than decent presentation on how the N-G model was applied to his modeling at CCB back in, it must be over a decade by now. His was a regional overview applying N-G to his situation IIRC, mine I tried building on Brian’s by applying AAR Car Service orders.  

I find nothing wrong in what Bruce wrote, or Steve Hile wrote or Tony wrote.

Again it is how close above the railhead that you are looking down from, which is why after Brian’s presentation, I called mine – Closer to the Railhead.

If in doubt or if one does not care to do a lot of research on your modeled environment, the N-G theory is like a set of good manners your mother asked you to use, it should cover you in many (but not all) situations.  

                                                                                         Jim Dick - Roseville, MN               


Bruce Smith
 

Tim,
 
Photographic evidence, while appropriate for the presence of cars, is not the type of evidence that can be used to support or refute the N-G model because it says nothing about the proportion (with the oft debated exception of yard photos, which may or may not reflect the in service proportions).

The problem with your assumption with respect to gondolas is that, given the PRR’s huge fleet of gondolas, it is absolutely expected that some would arrive on the left coast.  The question would be for example, would you see a ration of RR gondolas to ATSF or UP gondolas that mimicked the national numbers on the SP?  I’d like to see conductors books or train records that confirm this before I believe that gondolas follow the N-G model.

Regards

Bruce


Bruce F. Smith            

Auburn, AL

https://www5.vetmed.auburn.edu/~smithbf/

"Some days you are the bug, some days you are the windshield."



On Sep 4, 2017, at 9:27 PM, Tim O'Connor timboconnor@... [STMFC] <STMFC@...> wrote:



Bruce

You say, but I disagree. I have enough pictures of PRR gondolas in California,
Montana and the west to convince myself that they were found in proportion to
their population. Of course, since there were not many loads available for them
in California, they tended to go home quickly. But that does NOT change the
model at all. Do the math. The Southern Pacific owned a tiny number of mill
gondolas compared to its entire fleet, and you would be very unlikely to see
one in Pennsylvania, simply because they were incredibly rare compared to the
PRR's fleet. Heck, they were rare in SP freight trains too!

Tim O'Connor



And with all that, you got it wrong? :)  The N-G model does NOT apply to mill gondolas.  They are regional in distribution.  Flat cars do conform to the N-G model, at least for some periods.

Bruce Smith




devansprr
 

Schuyler, Claus,

First - I'll take the heat for my erroneous post a few days ago about the Delano 1943 Barstow picture that appears to have started this thread (to the apparent irritation of the more senior members of this group). I got a little over-excited about the solid string of ATSF box cars that could be loads, likely a train out of the War Department's Yermo storage facility, on its way to a WWII port of Embarkation on the west coast.

No way to know for sure if that is the case, unless someone WILL WEIGH IN on what sort of paperwork would be stapled to the tack boards of XM's by the ASTF....

That string of cars, if it was a loaded consist, would be an example of a consist that appears to defy the N-G model, but in fact does not. As Bruce, Tim and Tony have pointed out, the N-G model is appropriate for the overall distribution of Box cars for at least the WWII period (my personal interest) and beyond. Averaging consists over a period of time (undefined), I believe it is accurate for trunk line traffic. For model railroads that are close to, or focus around, specific industries (either individual, or within a region), the diversion from N-G can be significant. There has also been considerable discussion about how the Canadian RR's fit into that model since there were requirements, varying over time, to hustle Canadian cars back towards Canada once unloaded.

I'll take crack at a summary, at least as it pertains to box cars - where there has been general agreement on this list. For flats and gons, there is obviously differences of opinion (and I believe the N-G model never claimed to cover other types of cars - Dave Nelson?) The group consensus is that it is not appropriate for hoppers and tank cars, and may have been appropriate for reefers over only a few years of WWII (Tony? Bruce?).

As you have surmised, many of us are using the N-G model to help create a realistic freight car fleet for our model railroads. That is where the statistical aspect comes into play. Tim is correct that statistical variations will yield significant diversions from N-G theory for individual trains, and even specific days for a major trunk line. For modelers, this translates into trains and op sessions.

The question is do you want the prototype police to cry foul when they operate your layout because they never saw a NP XM, or a PRR X29? (depending on your era - the PRR deployed more X29 class boxcars than the NP owned in 1943... Sorry Mike ;-)

Tim's remark that you will need 10 to 20 times the number of cars for your layout to accurately represent N-G is incorrect. (Although he may be correct if you want to match the distribution for every box car class within a RR's box car fleet, rather than be representative only by RR.) If you were only concerned about an approximate representation of each RR, then you do not even need double the fleet - although it really depends on how many op sessions you are willing to host before someone complains that they never saw a Tennessee Central XM (200 cars in 1943 ORER, out of around 800k XM's in the North American fleet. One car in 4,000!) I actually dug into the statistics a few years back and if you search for messages by member devansprr on 2/4/2009 (wow - long time ago), I reported the results of a statistical analysis I did on this issue. The February 2009 thread on freight car distribution was pretty extensive.

In summary, based on the 1943 ORER (your era will vary - YEWV?), if you were to bring 200 box cars out of staging each op session, and your operators could only recall the box cars that appeared over five op sessions, then to match the statistics of the N-G theory, you would have 178 box cars from 39 roads that would appear every op session. An additional 17 cars per op session would be drawn from a pool of 33 cars (one for each of 33 roads), appearing once to four times out of five op sessions, depending on that road's fleet size (at this point we have accounted for the 72 roads with the largest box car fleets.). The last five cars to appear in each session would be from a pool of 147 roads that had the smallest fleets. Since you are clearing everyone's memories after five op sessions, you would only need 25 of these cars, each appearing only once per every five op sessions, allowing you to avoid building box cars from 122 of the 147 roads with the smallest box car fleets.

In the end, 236 box cars would provide you with a fleet that matched the N-G distribution over five op sessions. To match over more op sessions, a few more cars would need to be added. But you would really only need 36 extra box cars (above the 200 required) to match the N-G model. BTW, of the 178 box cars from the 39 largest fleets, 100 of those cars would be from the 11 roads with the largest fleets. Note that extra cars from the 39 largest fleets would be usable over the five op sessions to provide greater, and perfectly legitimate, statistical variation on a per op session basis (e.g. run 3 extra ATSF box cars in one session, and pull out 3 of the regular ATSF box cars in another session.) And having more than 236 cars for the 200 car session is perfectly acceptable - just don't put out three TC box cars every session - you still want to be representative of the national fleet percentages.

If those 200 cars appear in 20 trains, then only five of the trains will have the rare cars, and a few trains may be heavily biased towards one or a few roads. That matches the data that forms the basis of the N-G theory.

I would note that N-G applies to loads. I diverge a little from the group when it comes to MTs. My personal interest is the PRR mainline west of Harrisburg during WWII, and since the WB traffic was mostly MTs, there will be exceptions to N-G, even over large sample sets. This is because of the routing of foreign cars to home roads. Specific to my era and location of interest, NYCentral cars will be significantly under-represented for the WB traffic - the PRR had specific instructions on where to route MT NYC cars in Harrisburg, and it was NOT towards Altoona. So there may be some adjustments required for MT's depending on the location you model, and where MTs could be returned to foreign roads your road interchanged with. (Conversely, if you are modeling an interchange with a large road, you may need more of that road's cars delivered to the interchange, representing MT's being sent home from across the system you model.)

If you only require a smaller fleet, then the number of cars required still drops. Sticking with the concept that none of your operators will recall what cars they saw after five op sessions, if you only need 50 box cars to come out of staging every op session, you only need about 81 cars - 34 cars from 17 roads that participate in every session, 27 cars from 27 roads that will appear from once to four times out of every five op sessions, and a pool of 20 cars that will contribute 4 cars to each op session, none appearing in more than one of the five op sessions. More cars for the large roads will help create a little more variability (it would be justified) as long as you fiddle them in and out between sessions. Because the 1943 box car fleet was so diverse, the once per five sessions "fiddle pool" would include such well known roads such as CGW, Cotton Belt, D&H, MEC, B&M, WM, KCS, CofG, and several others. You could have more than 20 cars in that pool - but then some of the cars would not even appear every fifth session (which is ok.) And your five session consist averages will match the what one could expect from a random sample of 250 box cars from several days of trains on a trunk line.

There are other ways to slice the N-G data/theory. For a WWII modeler trying to capture the flavor of the WWII fleet, matching the distribution of steel versus wood, overall volume, and length is another consideration. During WWII, 36 foot boxcars outnumbered 50 foot box cars. Ignoring the Canadian 36 foot fleet (due to the exception above), 36 footers were 50% more numerous than 50 footers during WWII - which is why the 36 foot Accurail model is so significant for the WWII modeler. And wood sided cars outnumbered steel sided cars (55% vs. 45%.) Significantly, three of the six largest North American fleets (PRR, NYC, B&O) were about 90% steel - so it turns out that in 1943 66% of the non-PRR, non-NYC, and non-B&O North American box car fleet was wood.)

Modeling WWII is almost like a different scale compared to the modern era - the cube of today's 53 trailer van is about the same as an X29 - and I am driving next to it on the interstate at 70 mph... and now the RR's double stack them. Using N-G helps ensure the sensation of smaller box cars is preserved for a WWII layout.

Therefore certain classes of cars should also probably be considered - the PRR X29/ARA 1923 steel box car and the 1937 ARA designs being two numerous classes during WWII (PRR X29's made up nearly 3% of the national fleet, and there was an equal number of brethren from the other roads, and by 1944 the 1937 ARA fleet, across all roads, totaled nearly 12% of the national fleet.) With so many fine models of the 1937 fleet, it is easy to acquire a representative sample - and also easy to over-represent that fleet - and they were considerably larger (by cube) than over half of the national 40 foot fleet.  Obviously these numbers would change as the national box car fleet changed. Personally I believe that following N-G helps ensure a model RR gives one the correct impression of the RR fleet for the era modeled.

From a railfanning standpoint, the N-G model should also simulate the "diamond in the rough" experience of a WWII prototype railfan on Sherman Hill who, on average, might spot only one box car each from the GB&W, Georgia, TH&B, NWP, and FEC out of every 1000 box cars that passed by. (Each of those roads are in the bottom 147 fleets) To match the N-G theory, even roads like MKT, T&P, CGW, WP would not be represented in every 200 boxcar op session (and I recognize that is a pretty big op session.) Conversely, out of those 1000 box cars appearing in 5 op sessions, in 1943, on average, on a road other than PRR, 90 of those box cars would be PRR - 30 of them class X29 (PRR made up about 9% of the US box car fleet in 1943, so a 236 boxcar fleet should have around 18 or so PRR cars, 6 of them X29s.)

Note that with 200 box cars appearing every op session, you should include 2 NP box cars every session to keep the sheriff happy... (which is why we are all eagerly awaiting Rapido's next freight car - it could be the jail buster.)

With such a fleet, as long as your trains out of staging are not all run-throughs, the psuedo-random distribution of cars should yield what seem like unlikely scenes - yet they will be quite legitimate. There is a classic Delano 1943 picture from Belan, NM with four identifiable box cars in it. All belong to east coast roads, several are unique classes, and one is from a road, that if I recall correctly, owned less than 200 XM's. So in the micro, N-G will never be apparent - in the macro, it should, and if you use it to guide your fleet acquistions, your micro scenes will be entirety defensible to the prototype police.

Dave Evans

PS - I would note that I have not factored in the general agreement that home road cars would be over-represented compared to what N-G would predict. Your home road fleet might impact these numbers a fair amount, or hardly at all, depending on the size of their fleet.


---In STMFC@..., <schuyler.larrabee@...> wrote :

Tim O'Connor wrote:


Do we HAVE to go through all of this again
Schuyler??

Actually, the point is NOT to go through all this
again. I asked, again, is, the G-N Model neatly
summarized anywhere? I most assuredly do not want
to go through the 1600+ emails you offered.
Bruce's single-sentence summary is helpful but not
entirely convincing. It's TOO simple.

Many people seem to need a refresher class in
probability and statistics
before they can understand WHY not all trains look
alike.

Well, sorry, but my career path didn't require me
to take a class in probability and statistics to
begin with, never mind a refresher. And you know
the quotation that involves statistics. And since
I don't want to assume you know, as you're
assuming I do, the quotation is "Lies, damned
lies, and statistics."

I never said all trains look alike, Of course
not, that's why we loved (in the transition era)
to watch them go by, just about every car was
different, different shape, different roads,
different regions of the country. But the logical
extrapolation of what has been said leads me back
to This would suggest that in a train on, for
example, the NP you could, on average, over time,
see the sane, or very nearly the same, consist
makeup as say, on the Southern. Or the UP, Or
the D&H. Or the Rutland, the MEC, the NH, the
M&StL, the CGW, the AA, the NKP, the C&O, the SP

Notice I didn't say the same train. I said the
same CONSIST MAKEUP, which doesn't even claim that
the assortment of road names represents the same
car types. Hence, a train with all the same
reporting marks could look completely different
than the same set of reporting marks for another
train.

If you remember even a little bit of high school
math, you should remember
that a random distribution of coin tosses (heads
or tails) produces a
BINOMIAL DISTRIBUTION (aka the Bell Curve). The
chance that two random
freight trains of 100 cars each have exactly the
same cars in them is
astronomically small!

Of course it is. Don't be, well, insulting. I
never suggested that. Maybe you are not grasping
what I am saying.

And THAT is why even though ON AVERAGE freight
cars
are distributed according to their share of the
fleet, on any given day or
in any given train, there may be very different
proportions in that train.

I get that. But the G-N Distribution Model has
been getting used as a way to establish the
distribution of cars a modeler should have in ther
complement of cars on their model railroads. It
strikes me that if you were to stand trackside
around the country in the early 50s, for example,
the G-N Model might well be fulfilled, overall, on
average, across the nation. But to extrapolate
(actually, the inverse of extrapolate, a term I
don't know) the Model and use it as a method of
determining the complement of car reporting marks,
car types, and so on for a model railroad
representing a subdivision of a prototype railroad
strikes me a not a solid basis for doing so.


To PROPERLY model based on the G-N model you
really need at least 10x or 20x
the number of cars that can possibly be used on
your layout. That fleet should
be based on the national percentages.

If we were to do that, it's my impression that we
would all have trains made up predominantly of PRR
cars. For most railroads, I think that would be
highly unlikely. Many, sure, but not likely in
proportion to "the national percentages."

Then write yourself a Basic program that
RANDOMLY chooses box cars to enter the layout from
your staging yard. You will
notice two things. First, there is an incredible
variety of consists.

Undoubtedly.

And Second,
over time, you'll notice that the percentages of
cars follow the G-N model. QED.



I don't think that is necessarily true.

Schuyler




Tim O'Connor




This would suggest that in a train on, for
example, the NP you could, on average, over time,
see the sane, or very nearly the same, consist
makeup as say, on the Southern. Or the UP, Or
the D&H. Or the Rutland, the MEC, the NH, the
M&StL, the CGW, the AA, the NKP, the C&O, the SP .





[Non-text portions of this message have been removed]


Dave Nelson
 

Tim's dead, I'm alive. Bruce Smith's summary does as good a job at stating
the bottom line as any. If you need extra wrinkles then add WWII (maybe as
early as 1940) into the 50's, mainline trunk routes -- you need to see a lot
of boxcars rolling by to get a proper sample. The dates cited are what they
are only because I had hard data for that period, not because before or
after have been shown to not be relevant.

Dave Nelson

-----Original Message-----
From: STMFC@yahoogroups.com [mailto:STMFC@yahoogroups.com]
Sent: Monday, September 04, 2017 4:41 PM
To: STMFC@yahoogroups.com
Subject: Re: [STMFC] The G-N Distribution Model

Hi List Members,

Like Schuyler, I too would be interested in reading a rigorous summary of
the N-G distribution model. But it sounds like answering Schuyler's question
is not as simple as it might appear to be at first glance.

Perhaps a scholarly work of this level of importance to us should have an
official repository of its findings somewhere. I would suggest something
like a web page could be appropriate, perhaps.

Furthermore, it would seem to me that providing the contents of this
repository perhaps should not be the job of anyone other than the
Gilbert-Nelson team members themselves. That way they get to speak for the
work they have done. If it is not them, then it is a bunch of other folks
interpreting the work of the N-G team. This will inevitably lead to the
introduction of errors, inaccuracies, and departures from the intent of the
N-G team.

Suggesting that someone seeking a summary might want to inspect "over 1,500
saved emails" or "just the 578 emails from Tim Gilbert, and about a hundred
from Dave Nelson" is effectively saying that no such summary exists in a
easily digested or accessible form. Is that in fact the case - has the N-G
team not published a rigorous summary of their findings in any one place?

Just my humble thoughts on the topic...

Claus Schlund




----- Original Message -----
From: "'Bruce F. Smith' smithbf@auburn.edu [STMFC]" <STMFC@yahoogroups.com>
To: <STMFC@yahoogroups.com>
Sent: Monday, September 04, 2017 4:08 PM
Subject: Re: [STMFC] The G-N Distribution Model


Tim,


And with all that, you got it wrong? :) The N-G model does NOT apply to
mill gondolas. They are regional in distribution. Flat
cars do conform to the N-G model, at least for some periods.


Regards,

Bruce

Bruce Smith

Auburn, AL

________________________________
From: STMFC@yahoogroups.com <STMFC@yahoogroups.com> on behalf of Tim
O'Connor timboconnor@comcast.net [STMFC]
<STMFC@yahoogroups.com>
Sent: Monday, September 4, 2017 5:57 PM
To: STMFC@yahoogroups.com
Subject: Re: [STMFC] The G-N Distribution Model



Schuyler

A summary??? LOL !!

I have over 1,500 saved emails I could send to you ... :-)

Or just the 578 emails from Tim Gilbert, and about a hundred from Dave
Nelson?

My summary for your time frame: box car and flat car and mill gondolas
distribution
is proportional to the fleet size of every railroad, i.e. the model is that
those cars
roamed freely and randomly. Then you add the spice of traffic for your
railroad: reefers,
stock cars, coal cars, coal and stone gondolas, tank cars. Home road cars of
the latter
variety (livestock, coal, stone) tended to stay near home.

The email archive includes hundreds of special cases, exceptions, and
variations in
time and space - and scores of known consists from the 40's and 50's.

Tim




Gize,

I wasn?t very interested in the Gilbert-Nelson Distribution Model when it
was being developed. I didn?t have a layout then (still
don?t, but maybe . . .) so I satisfied my Freight Car building urges by
building kits that struck my eyes as being interesting.
Since those times I?ve settled (more or less) on modeling the ERIE in the
1950-52 time frame, and thus the G-N Model has become of
more interest to me.

But is the G-N Model neatly summarized anywhere?

And where might I find that?

Schuyler






------------------------------------
Posted by: "Claus Schlund" <claus@hellgatemodels.com>
------------------------------------


------------------------------------

Yahoo Groups Links


Aley, Jeff A
 

Schuyler,

 

               A couple of points – the G-N model only applies to free-roaming cars (think “plain boxcars”).  Not bottle cars, not hoppers, not reefers.  Not even boxcars in Pool service (auto parts?).  For those, you may develop the “Larrabee Distribution Model”.

 

               The G-N model only applies to Foreign-Road cars.  The home road is a whole ‘nuther thing.

 

               What I conclude from the model (given that I model the UP mainline in Kansas), is that I actually do need X29’s and NYC and C&O boxcars, even though Kansas is a long ways from those RR’s territories.  And the proportions of those cars on my model RR should reflect their proportions in the national fleet.  I shouldn’t have 1 X29 with 37 MKT boxcars – the regional bias just ain’t so.  Maybe a particular train would have that variation, but my overall stash of cars should not.

 

Regards,

 

-Jeff

 

 

 

 

From: STMFC@... [mailto:STMFC@...]
Sent: Monday, September 04, 2017 6:48 PM
To: STMFC@...
Subject: RE: [STMFC] The G-N Distribution Model

 

 

Well, thank you Bruce. Yes, you did this in one
sentence. And then instantly provided multiple
"excepts."

And while I think this is a logical extension of
the fact that freight cars are free-rolling, can
be assigned anywhere, look at the trek of the
Monon 1 as an example, on any given day, in any
given train, you can see any given railroad's
freight cars . . . I don't quite believe it.
Sorry. Sorry I asked.

This would suggest that in a train on, for
example, the NP you could, on average, over time,
see the sane, or very nearly the same, consist
makeup as say, on the Southern. Or the UP, Or
the D&H. Or the Rutland, the MEC, the NH, the
M&StL, the CGW, the AA, the NKP, the C&O, the SP .
. .

No wonder Brock obsesses about some damned hopper
going over some hill out west. 8^)

Maybe. But I have a bit of a hard time believing
that the consists of freight trains could be that
homogeneous across the country as that. And that
it would be a good way to establish the mix of
cars you have on your model of the XYZ railroad,
because that's what the G-N Distribution Model
says.

I can easily see how quickly that would get skewed
for any one railroad overall, never mind in
specific areas of a particular road. And I
suppose that's just one of the "excepts."

The ERIE had traffic between steel mills and
rolling mills around Youngstown. So if I were to
be modeling the Youngstown area (it would take a
gymnasium to model that!) I'd need hot metal
bottle cars, as well as more gons than I want to
think about. Plus a lot of hoppers, most likely
from eastern roads. Yeah, well, that's an
"except."

The ERIE had a tremendous produce and fruit market
to serve in NYC, so there were reefer blocks,
solid reefer trains, from Chicago to NYC (with a
few diverted to other eastern cities). I'll need
a lot of reefers. Yeah, well, that's a "except."

Maybe other than those "excepts," the composition
of the general freight pool for the rest of the
trains running over the entire line of the ERIE
might conform to the G-N Distribution Model, but I
am not even so sure of that. It seems to me that
there must have been some level of regional
skewing.

I must say that the responses I've seen from
long-time list members that amount to "weren't you
paying attention?" and offers to send me 1500 of
these emails and also 100 or so from someone else
so I can figure it out for myself certainly don't
reflect the usual level of "here, let me help you
with that" that it typical of this list. I think
the Claus is in much the same boat as I am, which
in effect is "what are you guys talking about?
Please educate me, and put this in an easily
understood form that I can go back to and think
about for my own purposes."

In my first post with this subject line, I asked:
"But is the G-N Model neatly summarized anywhere?"
I'm sorry Bruce, but your one sentence summary
doesn't quite get me there.

Schuyler

From: STMFC@...
[mailto:STMFC@...]
Sent: Monday, September 04, 2017 6:50 PM
To: STMFC@...
Subject: Re: [STMFC] The G-N Distribution Model

Schuyler,

Yes. Repeatedly. In the archives. But to make it
simple, here it is, again:

Boxcars from any given railroad appear on any
other railroad at the same frequency as they occur
in the national pool.

That's it. Really pretty simple. Note that it does
not govern the frequency of home road cars, or the
frequency of cars in individual trains, but the
overall frequency of foreign road box cars.
Several caveats deal with cars in assigned or pool
service, branch lines in the middle of nowhere,
yada, yada, yada... but the essence is very pure
and simple.

Regards

Bruce

Bruce Smith

Auburn, AL

_____

From: STMFC@...
<STMFC@...> on behalf of 'Schuyler
Larrabee' schuyler.larrabee@... [STMFC]
<STMFC@...>
Sent: Monday, September 4, 2017 5:21 PM
To: STMFC@...
Subject: [STMFC] The G-N Distribution Model

Gize,

I wasn't very interested in the Gilbert-Nelson
Distribution Model when it was being developed. I
didn't have a layout then (still don't, but maybe
. . .) so I satisfied my Freight Car building
urges by building kits that struck my eyes as
being interesting. Since those times I've settled
(more or less) on modeling the ERIE in the 1950-52
time frame, and thus the G-N Model has become of
more interest to me.

But is the G-N Model neatly summarized anywhere?

And where might I find that?

Schuyler

[Non-text portions of this message have been removed]


Dave Nelson
 

One of the problems people seem to have in accepting the theory is in the
failure to understand that average does not mean instance. I recommend a
search in the archives for the word yogurt (possibly Yoplait yoghurt),
There should be two hits, one from me and a reply from Tony. Please find
and read.

Basically a single train will never conform to the average. It takes 20-30
real freight trains to conform to the average. How many times do you move
1200-2000 cars across your model railroad? So let's get rea herel: FIRST,
not everybody on the list is a model railroader moving 12-18 car trains.
Some of us are historians. Me for instance. The theory is about real
railroads, not model railroads. SECOND, A few of us use simulators where
60-80 car consists moving across a 100 mile district are the norm. Me for
instance. Which is to say I can apply the theory quite directly because my
choice of expressing historical railroading has no compression in its
representation whereas physical model railroaders are faced with huge
compression requirements and so cannot do a dozen whole 60-80 car consists.

What physical model railroaders can do is make reference to the theory when
making purchases so they're not too overloaded with the arcane AND give some
consideration in the planning stages of how one might cycle cars on and off
their layout so as to achieve greater variability in the arcane because they
do show up.

Dave Nelson