As any designer or publisher will tell you, game design is a lot of "playtest, playtest, playtest". But when designing wargames especially, it's also a lot of "research, research, research". If designing a wargame is about modeling the historical factors that influenced the course of a battle, campaign, or conflict, then determining those factors is a crucial part of the process. Not to mention, of course, finding the right map and coming up with a reasonably accurate order of battle.
When I first started designing wargames, I had a paralyzing fear of getting something wrong. Someone who really knew their stuff would come along and say, "No, that division shouldn't be in hex 2208, it should be in hex 2209, the entire game must be bonkers and obviously you don't understand anything about the history, what is this completely erroneous set-up based on, a Wikipedia article?" This particular form of what the cool kids call "impostor syndrome" caused me to doubt every design decision I made. I compensated by doing copious amounts of research, checking and double-checking and triple-checking every source. It took me the better part of a year to design my Blood on the Alma, and about as long for Blood in the Fog. Those are of course relatively obscure subjects, so if I did make a mistake, maybe no one would notice (or so I reasoned). Doing my ACW game Blood Before Richmond took even longer, as I was certain someone would know that this one brigade only had left shoes, so why wasn't their movement factor reduced accordingly?
"Fording the Alma, September 1854", Louis A. Johns, 1901. The 2nd Battalion Rifle Brigade leads the Light Division across the River Alma during the Crimean War. Most of the English troops were equipped with the Minié rifles. They were not aware of the incredible range of the rifle, so they used it as though it had the same range as their usual rifle.
There were three main factors that enabled me to get over my phobia of making a mistake. First, I actually had some games come out, and the more games that got released, the more I felt like I kinda sorta knew what I was doing. Secondly, I actually did have a couple of people call me on "mistakes" I had made. One of those involved artillery ranges in the Alma game. I had artillery firing at a range of two hexes, and the Minié rifles at a range of three.
"This is completely unrealistic!" said the chap on the internet, who I am paraphrasing. "The infantry can stand outside of the artillery range and fire at them with impunity! Do you expect us to believe this?"
"Well, yes," said your humble narrator. "Because that's precisely what happened."
And this was definitely something worth replicating in the game. Now, here's the thing, though: the guy wasn't really wrong, exactly. The way I modeled it in the game was, in retrospect, not really appropriate for the game's scale. The maximum range of the rifles was about 300 feet greater than that of the artillery. The artillery range worked out to slightly less than two hexes, and the rifles worked out to slightly more. In both cases, I rounded up, but the result was that, in the game, the rifles were out-ranging the artillery by 800 feet, not 300. If I wanted to model the range differential, I should have done it at a different hex scale. There were also some chrome rules in the game that, in retrospect, weren't really appropriate for the game's chosen brigade-level scale, and made the game too finicky.
Which brings me to the third factor that freed me from the need to compensate by over-researching - the realization that different levels of research are required for different scales. If I'm doing a game on the American Revolutionary War that's focused, say, on the role of supplies in the Northern Theater between 1775 and 1777, then I don't really need to read ten books on the Battle of Saratoga and I don't need to know every regiment by heart. If on the other hand I'm doing a game on Saratoga, then I need to know that information.
Even that's a generalization, however. I'm working on a game called Table Battles, which is something like a cross between W1815 and Yahtzee. Folks who took advantage of last year's holiday sale got a sneak peek of this system with Christmas at White Mountain. It's a system that functions at a very high level of abstraction. It's not a detailed simulation; it's a light dice game with some of the flavor of the historical engagements. If I was doing a Saratoga scenario for this system, then I don't really need to read ten books on the battle. I need to read enough to capture the broad strokes, because it's a broad strokes kind of game.
That said, even with a broad strokes approach, I still need to have a general grounding in the period: armaments, training, military doctrine, political philosophy, etc. This is achieved via what I would call "passive research". That is, I read a lot, about a lot of topics, usually without having a game immediately in mind. In fact, it's easier to do so without a game in mind, because then I'm not fighting the urge to shoehorn the new information into any preconceived notions or a game mechanism I've become enamored with.
The great thing about researching a topic the first time around is that it's that much easier to research it the second time. The research I did on Blood Before Richmond, for example, gave me a great foundation in the ACW in general, and for my game on Seven Pines more specifically. Designing Seven Pines; or, Fair Oaks was like spending time with some old friends. Though I'm not sure if any of my old acquaintances were ever as frustrating as George McClellan.
There are some topics I stay far away from as a designer, not because of any lack of interest on my part, but because the amount of research seems too daunting. I'm a sucker for Napoleonic games (and movies and novels), but I don't think I could ever do enough research, passive or otherwise, to be able to do it justice. There's also the fact that, despite several attempts to learn a few languages, I can't speak or read anything other than English (and a few words of Esperanto). Using only English-language sources on the Napoleonic Wars is bound to result in an Anglo-centric bias, and to be full of inaccuracies. I learned that lesson from my Franco-Prussian War game, Von Moltke's Triumph, where my lack of continental sources resulted in some rather embarrassing order of battle errors.
That said, would I be able to do, say, a Table Battles scenario on a Napoleonic land battle? Sure. The amount of research that's needed all depends on the scale of the game, its level of abstraction, and what you're trying to achieve with it.