Alternative Scoring Systems for Avalon Hill's "Dispatcher"
Rediscovering a Classic
I played Dispatcher when I was a grade-schooler, when it was a relatively new game, back then in the early sixties, during my youth in New Jersey. It was a lot of fun to lay it out on a picnic table, outdoors on a fine summer day, and play it with neighborhood friends.
None of us ever really mastered the game, and neither did we question the rules. Game rules, to us in those times, were handed down from above, inviolable and sacrosanct, based on the certainty that the game author had superior knowledge and was likely a superior being. And besides, we were having a lot of fun and didn't care about much else.
Now, some four decades later, I've returned to Dispatcher. By luck I obtained a slightly scuffed used set for the sum of $17, which apparently is a real bargain from a collector's standpoint. (I did need to generate a few missing counters with a desktop drawing program. Times have changed; as kids we would replace a missing game piece with something scrawled on cardboard.) Imbued with the knowledge of age (or at least liking to think so), my view of the author's omniscience has changed. In short, Dis- patcher has failings and shortcomings.
Playing through the game, that sense of realism that I thought was there so long ago fell somewhat short now. To be sure, Dispatcher is a great game, and it is certainly still a great deal of fun. But I'd like to set out some thoughts about what bothers me about the original game, and what I think can be done to square things away.
Global Optimization vs. Competition
Dispatcher's existing scoring system, assigning demerits based on various factors, doesn't seem to work well. It was clearly designed with the needs of a competitive two-player game in mind, and that shows in the implementation.
That's fine as far as it goes, but it's a bit forced, because Dispatcher as a two-player game is a bit forced. Dispatcher is a consummate solitaire game, because the game is really at its best when the goal is not victory over an opponent, but global optimi- zation of the entire system.
To be sure, the two-player game tends toward a certain amount of a certain kind of optimization. But that optimization is done in separate divisions of the system, and "winning" the game often involves tactics which force non-optimal conditions on the oppo- nent. How often have you piled up freight trains at the east- west division at Glen Yard, knowing your opponent won't be able to handle the situation and will earn additional demerits under the existing game rules?
It is far better and a greater challenge to play the game in a solitaire or a cooperative mode and attempt to truly run the railroad. But the existing demerit system really doesn't promote that. Let's take a closer look.
Dispatcher by the Rulebook
First-class trains are considered the most important trains in the game, and running them on schedule is supposed to be the top priority. Those passengers can't be late, and the mail must go through. Next in importance, and sharing a similar demerit scoring system, are the merchandise express trains. (There are only a few of these, but the principles apply nonetheless.) First-class trains and MEs are assigned demerits according to these rules:
1. Each game-turn that one of these priority trains suffers a delay, by not being able to move its full movement allowance, demerits are awarded equal to the full movement allowance of the train. In other words, if the train can't move one space, repre- senting a 15 minute delay, it's treated the same as if it can't move several spaces, representing, in the case of a completely immobile first-class train, a delay of a full hour. For each game-turn that this "short" movement takes place, additional demerits are awarded. This bears some relationship to a train arriving late at its destination, but as we will see, not nearly enough.
2. If a train is mis-routed, no matter how badly, even if it is sent all over the system, it receives one full set of demerits, but no more. As long as it moves its full movement allowance every game-turn, a train can go anywhere it pleases with only one set of demerits.
3. Non-priority trains (the blue "freights") receive one demerit for each game-turn that they rest in their starting station without being moved out onto the railways. They also receive one demerit for each game-turn they rest in the east-west transition area at Glen Yard. According to the original rule- book, the former part of the rule is to penalize players who are very conservative and simply don't move their freights out. The second part of the rule, while having to some extent a similar purpose, seems to be targeted at the two-player "competitive" aspect of the game. One player can pile up freights in the transit yard at the expense of the other player, who will have to deal with the situation.
The sum total of these scoring rules simply doesn't lead to optimization of operations. I'll present a few key points that I believe represent "railroad-like" behaviors, which the rules should encourage. Then, let's see (a) how close the original rules come to this ideal state (or ideal as I see it) and then (b) how the rules might be changed to achieve a closer match-up. And remember, I'm looking at this more from the standpoint of "how to run a railroad" instead of what might make for two-player competition.
How to Run a Railroad
What would a "real" railroad system try to achieve? Let's look at this as it might have been in the early fifties, an era that the game could easily represent. (We certainly won't look at the railroads of today, where, in the United States at least, passen- ger travel other than on commuter lines barely exists.)
1. First-class trains, representing passengers and mail, clearly are of top priority. People must be transported to their destinations, not only on time, but in style. Mail can't be delivered late. And some critical or special-order merchandise must be delivered on-time, all the time. Delays simply aren't acceptable in this part of the business.
2. Freight needs to get to its destination in a timely manner. It is not on a fixed schedule per se, but certainly must be somewhat predictable. This means that there is a maximum acceptable delivery time. Freight needs to get where it's going within, say, 24 hours, and any arrival time within that time period is completely acceptable.
3. The railroad would run to maximize profit (rather obvious). While freight revenue was probably much of the profit even in the fifties (look at the large number of freights represented in the game), it seems that the freight revenue was in some sense "protected" by the reputation garnered by running crack first- class services.
All of this points to the need to place quite a priority on passenger operations without sacrificing efficient delivery of freight.
So How Close Did Charles Roberts Come?
The rules of the original game ensure that most demerits will occur from not moving freights out, although moving too many of them out will cause increasingly large demerit accumulations from delayed passenger trains. But there are some real anomalies.
1. Moving out a freight train and simply parking it somewhere avoids all demerits. (This is not as simple to do as I make it sound. But the principle applies.)
2. Keeping a passenger train moving, even if it's moving the wrong way, avoids all but a token single set of demerit points. Having a train arrive at its destination, much less arrive on time, is only an indirect goal (because demerits accumulate only when a train can't move its full allowance).
3. Piling up trains at the east-west junction at Glen Yard is virtually encouraged. This hardly leads to optimized global operations, although it leads to competitive advantage in a two- player game.
It appears, then, that the game strays rather far from an ideal global state. The player can, and will, do some rather odd things which are driven by the scoring system rather than by a desire to run a good railway operation.
Making Improvements
Let's look at some different ways to conduct business. We'll base them on a few principles derived from our analysis of the "ideal state" discussed above.
1. Passenger trains and specials need to arrive on-time. There is really no other measure of success. They need to take a prearranged path, as "waypoints" could well represent mail drops and the like. (The game doesn't seem to deal with traffic on anything but a point-to-point basis. While trains have certain timetable listings for various towers, they don't have intermediate station stops, although they sometimes have required routings. We won't try to deal with that here; we'll just take it as is.)
2. Freight trains should reach their destinations within the 24 hour period represented by a complete game. (We also won't deal with the so-called "shorter" games suggested in the rulebook, which simply don't work.)
Developing New Rules
Let's look at some ways to implement the principles described above. Let's state our first two principles:
1. First-class and second-class trains that arrive on-time at their destination (and required intermediate routing points) garner no demerits.
2. Freight trains which arrive at their destinations, via any required intermediate routing points (i.e. an AV-EL train) garner no demerits.
And now let's see how we earn demerits:
1. First-class and second class trains that arrive late to their destination or any required intermediate routing points earn demerits as follows:
D = L*T*>P
where
D = demerits
L = number of 15-minute increments late
T = train factor (Tf = 4 for first-class, and Ts = 3 for second- class)
P = parameter, different for first and second class, discussed later, and called Pfl for late first class trains and Psl for late second class trains
Intermediate routing point demerits and final destination demer- its are not cumulative; the larger of the two is used. (This actually makes sense in considering rule 2 below.)
2. For first and second class trains that don't arrive at all by the end of the game, the following slightly complex rules are used (these do not apply to a train that crashed; see below).
D = G*T*P + U
where
D = number of demerits
G = number of sections of track left to travel to a possible intermediate routing point and to final destination (not cumula- tive, but defines the path for counting the sections)
T = train factor, as in rule 1
P = parameter, different for first and second-class trains, call Pfg and Psg, and discussed later.
U = late calculation from rule 1 as follows: assume that the train had just arrived at the final destination at the very end of the game, and calculate the late penalty accordingly
3. For third class trains that don't arrive at all by the end of the game, apply rule 2 above with
T = train factor = Tt = 2
P = Ptg
which will be discussed shortly. Notice that this increases the need for planning. You can't simply shove any old third-class train onto a track, as you could in the original rules. Instead, those trains that need to travel farther have to leave earlier if they are to avoid non-finishing penalties.
Selection of the various parameters is all-important. Non- arrival of first and second-class trains is somewhat serious (especially given the timetables, which make this unlikely except in a case of really poor management). So non-arrival penalties factors should exceed late penalty factors and parameters must be chosen accordingly.
The distance factor, G, does not take into account the fact that a non-finishing train might not (and likely can't) proceed to the destination (after the end of the game) in a direct and undelayed manner. This calls for an increase in the various P-g parame- ters.
In the odd case where a train has reached a required intermediate point but then didn't finish, we apply the late train penalty for the intermediate point and the non-finishing penalty for the final destination.
The choice of the five parameters, Pfl, Psl, Pfg, Psg, and Ptg, also determine the economic balance between the various classes of trains (although T, the train factor, builds some of this in already; in fact T is not really necessary but it makes things more explicit and visible and can also be used to control the relative scoring structure).
Setting Parameters
The parameter choices need much research and game-play to fine tune, and I invite suggestion and comment. I'll simply suggest one set of parameters that seems logical and let you explore further on your own if you wish.
Pfl and Psl can be set to 1 for simplicity, and behavioral rules incorporated into the other parameters.
Pfg and Psg need to be set such that late arrivals cost less than non-arrivals. As an example, suppose a first-class train fails to finish at a distance of ten track sections from the final terminus, and it was due in two hours ago. According to the late formula, formula 1, the train earns 32 demerits, which becomes factor "U" in formula two. This represents how late the train already is. The ten track sections represent how much later it would be if it were to proceed direct to its destination (after the end of the game. How much higher should the penalty now be for not finishing, taking into account that further delays are possible, and that not finishing is a bad thing? If Pfg were equal to 1, we would have 40 demerits, surely too low. If Pfg were equal to 2, we would have 80 demerits, for a total of 112. This may seem high but it accounts for perhaps 50% additional delay and a 50% non-finishing penalty. For second-class trains a similar argument might apply and setting Psg to 2 may also be appropriate.
Setting Ptg, for third class trains, is critical in import. Let's say a freight finishes the game the same 10 sections from its destination as in the example above. Now, how far away it is doesn't matter except in how it influences how late we can expect that train to be. Freights are slower, so that means if there were no further delays, it would be five hours late. To reflect this in comparison with first-class trains, Ptg has to be at least 2 (twice as slow). Ptg of two would yield 40 demerits, or a little less than a third of the first-class train total above.
Now, there is no late arrival factor for third-class trains, which implies that Ptg ought to be a bit higher, but not too much since these are non-timetable trains. But, it should be higher still considering possible additional delays. Ptg of three yields 60 demerits, or about half of the first-class train example. This makes first-class trains about twice as important as freights. Is this appropriate?
In the original rules, there is no way any freight could get more than 24 demerits absent a collision. On the other hand, a first- class train could get up to 96 demerits in an absolute extreme case. Typical numbers are hard to justify, but you might say 8 demerits for a first-class train and 12 for a freight might actually occur on the average (I have nothing to back this up with). Now, in my proposed system, we're working with much higher numbers, but we're interested in the ratio as an expres- sion of relative importance.
So if we have first-class trains roughly twice as important as freights, what behaviors do we drive? Essentially, if we have trouble with three freights that would allow us to sacrifice a first class train to obtain a probable better score. To me that seems low. Five or six freights seems better.
So I'd leave Ptg at 3, move Pfg up to 4, and Pfl up to 2. This would, in our example, give us 264 demerits for the first-class train and 60 demerits for the third-class train. We're getting closer now. This is just over a four to one ratio and looks about right.
In summary:
Pfl = 2 Tf = 4
Psl = 2 Ts = 3
Pfg = 4 Tt = 2
Psg = 4
Ptg = 3
What About Collisions?
A collision is really serious, so serious that it should never happen. The original rules awarded 50 demerits for a collision (presumably per train, or 100 in all). With the higher numbers generated by my proposed scoring system, collisions need to be evaluated differently, and need to be related to the parameter factors shown above.
My formula for a collision, applied per-train, is:
D = C*P*100
where
D = demerits
C = 15 for a first-class train, 3 for a second-class, 2 for a third class train
P = Pfg, Psg, or Ptg depending on the train class
The collision factor "C" is much higher for a passenger train because of the injuries that would result. Thus a passenger train collision would incur 6000 demerits, a game-wrecking amount, and I think rightly so. If you disagree it is simple enough to adjust the "C" factor. A third-class train would incur 600 demerits, a more modest number but still quite significant.
Conclusion
No matter what the scoring system, Dispatcher is, absent the random event cards, a closed-form game which will have an optimal solution (even though it's hard to find). Computer analysis would be really interesting (any takers?) as a means of finding the best solution. "Best" of course will vary with the scoring system. Having such a computer program, and varying the rules and parameters would be a fascinating study, but one that is not likely to be done soon.
I would certainly love to hear from other Dispatcher enthusiasts. What rules do you play by? Do you stick to the originals or have you too developed your own system? What do you think of my system? Are the parameters right? There is a lot here to dis- cuss.
I can be reached for the indeterminate future at:
and I welcome your e-mail.
Bob Newell, Santa Fe, New Mexico, October 14, 2001; Minor revisions February 6, 2006
In true retro fashion, this document was composed in its entirety with WordStar 5.5 on an Epson Equity II (286) computer.