Professional Gaming ? Modeling with Wargames
E160A -- An open panel discussion with audience participation, examining the state-of-the-art usage of wargames for modeling and decision support. Part of the Strategicorps event track. All ages / 50 seats. Run by Brant Guillory.
This sounded like something that might really be interesting, and since I paid for that Strategicorps ribbon so I really ought to use it. This might have been the best decision I made all week.
I arrived just as things were about to get started. With more panel members than audience, Brant decided to have us all pull our chairs into a circle to facilitate discussion. Good call Brant. Introductions ensued:
The Panel:
Brant Guillory (Moderator, former US Army Captain, PhD candidate in Communications, Bayonet Games) 1,2,3
Col. Matthew Caffery (Chief, Wargaming Plans & Programs Directorate, Air Force Research Laboratory) See 1,2
John Tiller, PhD (independent game developer with Matrix Games and HPS Simulations) See 1,2,3
James Sterrett, PhD (Simulations Instructor, U.S. Army Command and General Staff College (Northrup Grumman), Military Historian, Ad Adstra Games {Attack Vector : Tactical}) 1,2,3
Joseph Miranda (Editor Strategy & Tactics Magazine, MCS Group) 1,2,3
Maj. Michael Martin (CGSC graduate, soon-to-be PhD candidate in Modeling and Simulations at ODU in Norfolk.)
The audience (as best I recall):
Col. Ken Guillory (MLRS commanders in Desert Storm I: a man of vast military experience and father of Brant Guillory) 1
Robert Crandall (Programmer and Games Developer, Core Talent Games, Matrix Games)
Jim Snyder (Matrix Games, Lock and Load Games) 1,
Myself (DE)
Several others joined in as well, including a number of avid gamers, but I didn't get any more names.
---------------------------------------
The following is my (pitiful) attempt to summarize some of the discussion, more-or-less stream-of-consciousness style. I will attempt to attribute comments where I can. All quotations are approximate.
BG: This discussion is intended to be about the use of games and simulation to model warefare, rather than as a training tool.
JT: The modeling process is "legitimate but not realistic.
MM: Modeling guys get good data on the physical aspects, but not data on psychological aspects. A favorite tactic: "suppression fire and sneak around" works because it distracts the enemy.
KG: (on adjudication of information) personalities matter. Not all leaders are equally about to pass on important information.
MM: Try a search on "Correlation of Forces and Means".
MC: Air Force Command structure is nearly Theater to Pilot. This is necessary because functional groups (bombers, ECCM, refueling tankers) are not based together.
[DE: the discussion turned to data representing units - like maximum road speed - and what that means]
JT: A tank platoon moves at the speed of its most confused Sergeant.
BG: Average speed means more than maximum speed (says the man who rode an abramsM1A1 tank down the road at 60+ mph!).
JT: Mission is more important than the vehicle. Recon units will be faster because that is their mission, not (entirely) because their vehicles are faster.
KG: Not the maxumum speed OR the average, but what you need when you need it.
RC: Maximums are useful for specific instances (what you need when you need it again).
MC: [drew a chart dipicting the relationship between granulatity, completeness, and interpretability in simulations, which I did not capture adequately. The essence was that you cannot have all three at the same time.]
JM: CRTs (combat results tables) based on data resresenting actual outcomes.
??: Logistics controls movement
??: "Generals study logistics"
??: different levels and aspects of simulation
??: Outcome of simulation may be affected by the agenda of those creating it.
DE: [summarizing] There was additional discussion of whether there could be a game simply about logistics [there are many examples] or if there needs to be direct conflict [something to go BOOM at the end].
JM?: [a good story on why a simulation cannot contain all possible outcomes] "Under what circumstances can a Calvary unit capture a ship at sea?" [and it actually happened!]
??: Does "resting" units get rewarded in games?
MC: [modestly described what has been called the "Caffrey Loop"] What can we learn from history that can make better games?
--------------------------------------------
The official discussion concluded, but 7-8 of us adjorned to the bar across the street. The evening concluded with several very pleasant rounds of beer, and (among many other things) a discussion of the merits of weaponized Silly-Putty (I get the strangest ideas sometimes).
I would like to thank the entire group, and especially Brant for moderating a great session.
[7/6/09 -- Thanks to the active responses of many of the panel members, I have been able to update and correct this post.]
[7/8/09 -- I now recall there was discussion of how people with particular agendas might influence the simulation results, and how this might be overcome. I commented that we were now talking about modeling the simulation process itself. At this point, Brant threatened "that if I got any more analytical, I would be require to buy the first round of beer". For the record, I bought the first round. :-) ]
[This post has been back-dated to approximately the actual time it occurred.]
23 comments:
Your summary makes me wish I was there to hear all of that, and I hadn't thought I would be that interested in any talk that dealt with miniatures. I was wrong!
Although I'm not necessarily good at tactical games, I find studying tactics and battles really interesting.
This was a mix of military experience, computer simulations professionals, historians, and game developers (computer and board). A fascinating group, and they certainly broadened my horizons about gaming - and I'd like to think my hozizons are already pretty wide.
It's GOOD to meet new people and discover common interests, but I just got a lot more than I expected from this session.
I was there (Robert Crandall) wearing my software wargame developer hat and it *was* a great discussion. My notes are pitiful too - too many good ideas flying around at once to write them all down - but I captured a few more things -
- the emphasis all too often is on the depth of modelling at the cost of getting results quickly. We need plausible wargames that can be done fast, preferably entirely under computer control (but following various scripts) and ideally hundreds or thousands at a time so we get a nice distribution of possible results.
- "All models are wrong but some are still useful"
- James Sterret - look for selective fidelity - a modelling of what is important to you for the purpose of the game/sim and abstracting out the rest. It doesn't have to do everything to be a worthwhile exercise.
- further to the above, 'excessive perfection' is a waste of resources. Good enough now trumps perfection later. Besides, as soon as you add a human to the equation you lose all precision anyway.
- Excessive modelling can kill the design if it just creates extra hoops to jump through. Don't put something in unless you can see it, control it in a meaningful way, counter it if necessary or use it for something else. For example, don't put a detailed logistics model in unless you really want to teach logistics. If that is not the objective then abstract it out. Don't remove logistics entirely, just don't model it to the nth degree.
Hey Guys -
Glad to see an attempt at a wrap-up of one of the panels/seminars. I think next year, we may try to record them for later posting online, if we can get permission to do so.
Just a few tiny corrections:
I'm no longer a captain in the Army - honorable discharge received shortly before Origins :) (proud of my service, but striving for accuracy!)
I also no longer work with Matrix Games. I'd freelanced with them for a time in '06-'07, but not since. They're still great friends, but I haven't worked with them in a while.
Finally, the bit about the Navy loading ships by type/quantity was dad, not me. He's the one who suffered through *unloading* them in Saudi Arabia in 1990 while I was in college.
And in case any of the readers are wondering, yes, I really did hit 60+ mph in an M1A1 back in 1995 at Fort Hunter Liggett. It was a lot of fun :)
Oh, and Mike Martin is a soon-to-be PhD candidate in Modeling and Simulations at ODU in Norfolk. He jsut recently graduated from CGSC and is headed east shortly.
So Brant - what is the difference again between a "game" and a "sim"? I'm still not sure what your take on that is.
Laughing, Rob Crandall
Rob Crandall wrote:
- "All models are wrong but some are still useful"
I love that one, it's a quote from statistician George E. P. Box
@Brant: How about your dad's rank, did I get that right?
Thanks for checking in and contributing! I just updated/added/corrected my post.
James Sterrett emailed me a picture to add, and I think he will be stopping by to comment when he gets a chance. I also found where I has hidden Joe Miranda's business card, which provides a much better link.
DC: dad's rank is correct.
Rob: we'll stay away from that argument, lest we be cast out and not invited back... ever...a
Here's another relevant quote:
"Far better an approximate answer to the right question, which is often vague, than an exact answer to the wrong question, which can always be made precise."
- J. W. Tukey (1962, page 13), "The future of data analysis". Annals of Mathematical Statistics 33(1), pp. 1–67.
and the difference between a game and a sim? I might take you up on that. ;-)
OK, fair enough. :)
The only difference between a game and a simulation is the presence or absence of an external purpose. Games are self-contained; simulations are used for a purpose.
This doesn't mean that any given utilization is well-chosen. I can decide to use Mechwarrior to simulate the Battle of the Atlantic at a strategic level... but it would be a silly idea. I could decide to sit down and play JDLM (Joint Deployment Logistics Model) as a game; but, sadly, making JDLM a good game would be a decided challenge.
However, the fact of bad ideas doesn't invalidate the central point. The same model could, in principle and in practice, be used effectively as both a training (or analytic) tool, or as a game. TacOps, or Flashpoint Germany, or Battles from the Bulge, to name a few none-too-random examples. :)
I do readily admit the best use of this argument is 1) to deal with people whining that their exercise should be using a (wonderful) Simulation instead of a (silly) Game; and 2) to try to focus people on *what they need done* before they try to choose the tool they will use to approach the task.
I was recently reading that there is not even a single definition of "game" that everyone can agree on (Brathwaithe and Schreiber's book, I think), therefore defining how a simulation is different may be problematic. Here's a couple:
"A game is a series of interesting decisions" (Sid Meier's definition, which excludes games like Candyland and Monopoly).
A game is "a play activity with rules that involves conflict" (Ian Schreiber's definition, which may work well depending on how you define "conflict").
James' definition of simulation would also seem work for mathematical/statistical simulation, which are commonly used in my field to evaluate the properties of new statistical methods.
For my part, I think a game must also have some element of interaction for the player(s). Candyland is interactive, even if it does not include any decisions or conflict. A simulation does not require interaction, though I expect most military simulations would be interactive.
JS wrote: ... to deal with people whining that their exercise should be using a (wonderful) Simulation instead of a (silly) Game ...
Since I began to seriously look into the mathematics of games a few years ago, one thing that really strikes me is the similarity between games and reality. The mathematics of attrition applies equally well in reality and to cardboard chits on a map. Games may be an abstraction from reality, but the lessons they teach are real. I would suggest that if someone cannot learn from the "silly" game, the "wonderful" simulation is not going to do them much good.
PS: I have not been able to dig up an email for Matt Caffrey to make him aware of this discussion. Perhaps one of your might know where to contact him? --- Dan
I've emailed you Matt's email.
"I would suggest that if someone cannot learn from the "silly" game, the "wonderful" simulation is not going to do them much good."
Absolutely! :)
http://www.costik.com/nowords.html
My definition is that games and sims can, in fact, overlap. I'll detail it more later, but basically this:
Games are competitive. They may be less (tic tac toe) or more (Flashpoint Germany) realistic when compared to each other, but the essential element is one of competition/victory.
Simulations are attempts to model reality, and may be less (JDLM) or more (JANUS) competitive, but are marked by their increasing level of detail attempting to replicate as much of 'reality' as they can.
Where do they overlap? It is possible to have a highly detailed, competitive product that is both a sim *and* a game.
I'll post details from an academic paper when I get a chance.
Brant wrote> It is possible to have a highly detailed, competitive product that is both a sim *and* a game.
Certainly!
I think my definition of simulation would be a little broader: a model of a process. If it models a "real" process, then it has a purpose (as JS noted above).
I seem to have left open the possibility that simulations might not model reality, and this goes beyond what we usually think of as a simulation. Still, if we have model that includes a hypothetical process (perhaps combat vehicles armed with laser-based weapons) is it not still a simulation. If there is any possibility of laser weapons in reality, then doesn't this simulation serve a purpose?
We could "simulate" a process that is complete fantasy, but then it would clearly be a game (or a toy!). Some aspect of reality, or at least possibility, seems to be required for a simulation.
Trial Balloon #1: A simulation is a model of a process that has a real purpose.
Trial Balloon #2: A game is a decision process.
(and games such as Candyland and Monopoly get re-classified as "toys".)
I would *not* argue that a simulation must simulate a real process! :)
If we get more technical.... then, (per memory of https://www.dmso.mil/public/resources/glossary/ which won't load today)
A model is a mathematical description
A simulation is a model over time
Battletech is a simulation of a fantasy/SF world. It may or may not be a *good* simulation of that world (not the point here!), but it contains a mathematical description of how parts of the world work, and instructions on how to run those models over time.
"Trial Balloon #2: A game is a decision process.
(and games such as Candyland and Monopoly get re-classified as "toys".)"
NoNoNoNoNoNoNoNoNoNoNoNoNoNoNoNo
Candyland has limited choices unless you play the 'draw 2, pick 1' rule for older kids. Otherwise it's pure recognition. Monopoly absolutely has choices.
Read the Costikyan article I posted to. "Toys" are things you play with that lack inherent rules. A ball is a toy until you put boundaries and goals around it, and it becomes soccer. Flight sims are toys until you start racing, or competing for tricks or whatever other 'rules' you graft onto it.
Sims have choices in them. Games make those choices competitive by keeping score.
A sim need not necessarily have a choice, unless you treat the initial conditions as a choice?
I can run simulations in which I have no run-time input at all -- from a bouncing ball simulation to a simulated rocket launch.
Different note -- Candyland is arguably more of a simulation than a game. A simulation of *what*, exactly, I do not know... but given a set of rules an initial conditions it has a process for determining an outcome. :) (Perhaps it's a bad simulation of a game? ;-) )
"A sim need not necessarily have a choice"
If you start with that stipulation, then Candyland can be a great simulation of an automatic conveyor belt with a fault clutch and motor mechanism :D
JS: I would *not* argue that a simulation must simulate a real process! :)
I was trying to make a connection to your requirement that a Sim should have a purpose, and some connection to reality, however tenuous, seems to be needed. You could simulate a complete fantasy, but then the only purpose might be Fun, or maybe Art.
JS: "A simulation is a model over time"
I'm trying to decide how the "time" aspect matters. Statisticians use simulation models to study the properties of new methods, and these generally don't involve time. We generate a random sample from a specified population, apply the method (a particular analysis), and record the result. Rinse and Repeat. Then we look at the distribution of results. So we have 1) initial conditions, 2) "Do Something", 3) Result. I think (2) might count as a sort of abstracted time. Sort of ... on the other hand this is a simulation of a mathematical decision making process, and time is irrelevant.
I can see that for training purposes, run-time input is the whole point.
BG: "NoNoNoNoNoNoNoNoNoNoNoNoNoNoNoNo"
Um ... Uncle: Monopoly is not a toy. The basic version of Candyland has no choices, but it does require participation of (and interaction with) the player. I still think it could be a toy. (I'm familiar with that Costikyan article too, but I'm not in 100% agreement with him). Read on ...
JS: "I can run simulations in which I have no run-time input at all ...."
Yes. My entire experience with simulation is with "no run-time input". I have done some simulation of games that looked at modeling certain types of player strategy during a game, but the player "choices" were determined randomly in a pre-specified manner.
JS (again): "I can run simulations ... from a bouncing ball simulation to a simulated rocket launch."
If we can simulation a ball, then we can play ball with a simulation. Is a simulation of a ball then a "Toy"? Are ALL simulations toys?? I often tinker or "play" with a simulation just to see how it behaves. I'm not sure I can think of a simulation that cannot also be a toy.
BG: "... Candyland can be a great simulation of an automatic conveyor belt ..."
Candyland (the basic version) is essentially a random distribution, and is very nearly a Markov process. A conveyor belt is a fair descriptions too.
I just remembered one of Brant's final comments from the original discussion, which I will add.
yes, I'm commenting on a post from years ago, but the relevant discussions have started to be collected here:
http://grognews.blogspot.com/search/label/Theory
Thanks Brant! Comments are always welcome, but late comments are often SPAM, thus the moderation after 60 days.
Post a Comment