February 17, 2014– Several years ago, for my book Normandy Crucible I chose to make use of a boardgame as a kind of historical laboratory. I took an existing game on the Normandy campaign, SPI’s Cobra, done by Brad Hessel many years back, to generate virtual data on the actual campaign. (I’ll say something in a moment about how that went.) The other day, while researching a topic from World War II in the Pacific, I was delighted to encounter another instance of someone who’d done the same thing. This was an honors thesis in history done at Ohio State nearly a decade ago. There’s not enough to call this a trend, but clearly the notion is out there, so I thought I would say something about best practices.
A little more detail on the Normandy Crucible experiment: I set out to explore the contours of the German defense of the hedgerow country following the D-Day invasion in June 1944. I wanted to see if a different combination of German strategic approach and/or defensive tactics might have worked better and, in particular, how these factors impacted on the size of the force the Germans would succeed in withdrawing from Normandy once they began their retreat. To accomplish this I first took the Cobra game and brought it up to date. When it was published in the late 70s we knew very little of the Allied intelligence advantages with ULTRA, the design itself had certain awkward elements, and now we have much better historical data on German replacements and reinforcements. All these things were factored into the game.
To effectuate the laboratory schema, I chose a set of parameters–some strategies, some tactics, some were terrain-based defense emphases. Each of these became a scenario. Then the game was played numerous times with extensive data recorded, ending with the number of German troops who escaped the debacle. The game was played solitaire and with various opponents. Every scenario was played multiple times. I ended up with a very interesting collection of insights.
In our other example, in 2005 Mark Gribbell at Ohio State used the 1992 editions of the Avalon Hill games Midway and Guadalcanal and modified them to replay the battles of the Philippine Sea (June 1944) and Leyte Gulf (October 1944). Gribbell asked the question of whether the Japanese, had they had better pilots at that time in the Pacific war, could have done better at these decisive naval battles. We don’t know enough about his assumptions or data to judge his actual scenarios, but his conclusions were that the outcomes would still have been Japanese defeats although the Americans would have suffered somewhat higher losses. To judge from the honors thesis, the battles were replayed just a few times.
So, best practices? In my view the first key element is to ensure that the simulation platform actually has the capacity to answer the questions posed. In both of these examples an original boardgame (or games) were modified to accommodate the experiment. There is not sufficient information to describe the actual modifications made in the Pacific war laboratory, but in the Normandy case I am confident the game update provided exactly the platform required.
The second key element is to ensure the questions are the right ones. For Normandy, the ability of German forces to withdraw was directly measurable by the game. Secondary questions, such as the effect of commencing the retreat at different moments in the history, was also observable. In the Pacific war case, losses were a surrogate measure for the impact of pilot quality, but there the connection is less clear. For example, what about the effect of better Japanese aircraft, along with or apart from, pilots? Gribbell himself suggests something of this with a related conclusion he draws–that the American use of the F-6F “Hellcat” aircraft was a key factor in Japanese losses.
Finally, it is very important to ensure that there are enough iterations of the scenario to collect a broad range of data. This takes better account of the effect of different player strategies and tactics, as well as for the element of chance.
In my case one other lesson of this exercise was that some readers and historians are not yet ready to accept the simulation as a valid tool for historical analysis. In the text of Normandy Crucible I made some use of the simulation laboratory results, but I had originally done more. Comments from readers and editors encouraged me to extract most of this material and move it to an appendix, and some items I had to drop altogether. The simulation laboratory has yet to come into its own.