fight for Western Europe. This projected conflict had been analyzed continually for decades and both sides’ intelligence had been so well developed that, by the mid‐1980s, nearly the entire world understood how the battle on the north German plain would unfold: attack corridors, force compositions, and equipment, even opposing commanders were all known. Tom Clancy’s novel “Red Storm Rising” (Putnam, 1986) provided a realistic description of what that encounter would have looked like and demonstrated the amount of information commonly available about that potential conflict.18 Quantifying the kinetic combat capabilities of forces became the focus of the analysis that underpinned the US defense acquisition decisions, and this played perfectly into the strength of closed‐loop combat simulations.19 However, US Army analytic organizations realized that closed‐loop combat simulations could not be relied upon as the single tool needed to do analysis. While the automated decision rules allowed for the development of stochastic models that could be run numerous times to ensure there was a representative set of battle outcomes, the automation of the human decision‐making process was recognized to be too simplistic to rely on for a complete assessment of combat operations. Both the Army’s Center for Army Analysis (CAA) and the Training and Doctrine Command (TRADOC) Analysis Center (TRAC) developed analysis protocols that first used wargames to thoroughly examine different courses of action (COAs) or concepts of operations (CONOPS) before deciding on a single scheme of maneuver that was then instantiated in their closed‐loop combat simulations.20
Wargaming has taken on a more prevalent role in DoD since the beginning of the US irregular warfare (IW) campaigns in Afghanistan (2001) and Iraq (2003), although there have also been major efforts to develop computer simulations to model IW operations such as counterinsurgency and stability operations. Initially, some modelers attempted to add IW complexity to existing kinetic‐focused combat simulations. They added a third side to their simulation to represent the population. Others added civilians on the battlefield, so kinetic engagements between two uniformed, armed forces had the potential to cause “collateral damage” among the populous. They were missing the point. The whole dynamic of warfare changes when the center of gravity shifts from defeating an adversary’s military to influencing a civilian population. In an IW campaign, kinetic engagements between armed combatants are infrequent, small, sporadic, and often counterproductive. The Counterinsurgency Field Manual FM 3‐24 said it best: “You can’t kill your way out of a counterinsurgency.” Whole of government approaches and modeling civilian populations’ attitudes and behaviors are required, and modeling these attributes presented significant challenges to defense modelers.21 DoD stood up the Human Social Culture Behavior Program in 2009 to address some of these challenges, specifically “to develop, implement, and demonstrate forecasting and predictive models of human behavior for both analytic application and warfighter training in support of non‐traditional warfare operations.”22 The United Kingdom’s Defence Science and Technology Laboratory developed the Peace Support Operations Model (PSOM).23 This is not a closed‐loop simulation, but a simulation that requires periodic decision‐making by human subject matter experts – a computer‐hosted wargame. PSOM was used in 2010 and 2011 for training UN peacekeepers in Central Asia24 and to assess campaign options in Afghanistan.25 A survey of simulation tools used to model conflict in Iraq and Afghanistan was completed by RAND in 2014.26
As the conflicts in Iraq and Afghanistan approached their first decade, the US DoD began to realize that both wargames and closed‐loop combat simulations have important and distinctly different roles in the analytic process. The United States’ involvement in Iraq and Afghanistan has highlighted that counterinsurgency and stability operations cannot be modeled well in existing closed‐loop combat simulations. While agent‐based simulations show promise for modeling human behavior in regions of conflict,27 there are no closed‐loop IW simulations that parallel the quantitative analytic capability of those used by DoD to assess major kinetic combat operations. In historical terms, modern‐day wargames are much like the Prussians’ Free Kriegspiel, while today’s closed‐loop combat simulations are more similar to Rigid Kriegspiel. Each tool has its purposes, and in most cases those purposes are not overlapping. Wargames should not be used for quantitative assessments and closed‐loop combat simulations cannot replicate human commander’s decision‐making processes.
Wargames Today
As analytic wargames began to regain some traction in the 2010‐time frame, they were attacked by combat simulation advocates. Analysts who teethed on closed‐loop combat simulations derided wargames as “a simulation of one replication” or a “sample size of one,” noting that you could not run a particular wargame multiple times, varying random variable values to generate quantitative output for statistical analysis. What they failed to understand was that a wargame’s focus is on qualitative data, decisions produced by human players, while the computer‐based closed‐loop combat simulations are focused on quantifying the attributes of a force engaged in high‐end kinetic combat.
Running a series of wargames to generate multiple replications or running wargames to compare and contrast different concepts or technologies is problematic. Running any simulation for multiple replications is typically done by holding most variables fixed and introducing randomness for certain, identified random variables (such as system‐on‐system probability of hit) so the statistics of the multiple replications can be calculated to examine the range of expected outcomes, given the introduced randomness. It is difficult, if not impossible, to produce multiple replications of a wargame because of the learning effect that the players experience, so the difference in replications is confounded by the players learning more about the operating environment and the opponent’s method of prosecuting combat with every subsequent replication. If the first replication of the wargame produced clear winners and losers, would the loser use the exact same strategy for the second replication? Would the winner expect the loser to not learn and try the same course of action? You could attempt to eliminate the learning bias by having different players play each replication, but that would assume you could find a large supply of players with identical experiences. In practice, some organizations that utilize wargames do play them multiple times with the same players, but these cannot be considered multiple replications, at least not in the traditional sense. When the TRADOC Analysis Center ran the H‐I‐T‐L simulation JANUS to develop a concept of operations (CONOPS), for later instantiation in the closed‐loop simulation CASTFOREM, the blue side had around 30 “pucksters” plus a command and staff, and the red side had around 10 pucksters and their own command element. Pucksters typically maneuvered and fought elements of the force, such as a tank company in a brigade operation. In order to obtain a “record run” that could be instantiated in the simulation, they would need four to five runs to familiarize all the pucksters with the operational plans. One or both sides would usually petition to move the starting locations of their forces between runs, so these cannot be considered “replications” in the pure sense. Only the results of the final record run were used to inform the CONOPS instantiation in the simulation.28
At the Office Secretary of Defense (OSD) Office of Net Assessment, players play a wargame multiple times where the players’ learning is one of the subjects of study. Understanding how a command and staff’s thinking evolves when combating an adversary’s new technology or concept in a wargame allows the evolution of doctrine without putting forces at risk.