The several techniques of war gaming continue to assume positions of increasing importance in the decision-making process in government and in industry. After 300 years of development and refinement, war gaming is being used extensively to evaluate equipment, to test tactics, strategy and plans, and to train executives and combat staffs.
Within various agencies of the federal government, games are being conducted to study political, economic, and strategic aspects of policy, actions, and reactions. The Joint Chiefs of Staff operate the Joint War Games Agency and use the services of several contractors. The Joint, Unified, and Specified commands such as SHAPE, PACCOM, and SAC conduct extensive gaming in broad strategy. The individual services operate their several independent gaming establishments; at the Washington level, the Deputy Chief of Naval Operations, Plans and Policy, operates gaming under the Assistant for War Gaming Matters; the USAF Deputy Chief of Staff, Plans and Operations, conducts gaming through the Directorate of Studies and Analysis, and the Chief of Staff, U. S. Army, operates the Strategy and Tactics Analysis Group. The Marine Corps conducts war games at its War Games Division, which is under the Marine Corps Landing Force Development Center, Marine Corps Schools, Quantico, Virginia. The organizations, responsible to the service headquarters, conduct games designed to evaluate proposed weapons systems, and to evaluate organizations, strategy, tactics, logistics, and intelligence. In 1963, a RAND Corporation report estimated that approximately 200 organizations were engaged in such analyses in support of military decisions. A list of some of the better known organizations which conduct this gaming is contained in Table I, compiled by the U. S. Army’s Strategy and Tactics Analysis Group. Some of these activities use the war game to train combat staff officers and to evaluate operation plans. But perhaps their most important function is to develop tactics applicable to new equipments and weapons and to evaluate prospective weapons systems. The civilian corporations conduct studies, analyses, and games on contract or as part of both their product development and their sales programs.
Gaming in American industry developed as an outgrowth of military war games as a program within the American Management Association in 1956; it now operates to accomplish three basic functions: provide training in decision-making, test corporate strategy and policy, and plan operations. Training games are now conducted for executives by the American Management Association, the IBM Executive Course, University of California at Los Angeles, Indiana University, and Oklahoma University, as well as many other educational institutions and large corporations in petroleum, transportation, insurance, real estate, and banking. Carnegie Institute of Technology developed one of the first realistic games for industry. To meet selected production schedules, players were required to co-ordinate maintenance, finance, hiring, overtime, purchasing, and other variables effecting output. Corporate strategy and policy is tested by “living out” a period of corporate life, by game, several times under several sets of strategy and then determining which produced the most advantageous results. An excellent example of operations planning by gaming is the General Electric simulation which provides such data as shop layout, capacity, scheduling, and inventory requirements. Gaming before production has become accepted as an essential tool to management planning in much of industry.
It has become increasingly necessary for those involved in the theory of conflict to be familiar with the advantages, the disadvantages, and the pitfalls of the several gaming systems. In order to use the results of analysis by gaming, to detect the faults, and to evaluate results, one must consider the theory of the various gaming systems.
In operational and administrative applications of gaming, the complexity of the computation must never be permitted to obscure the fact that the derived results cannot be more accurate than the input data upon which the play is based. Further, there is an ever present danger that the final decision, which the play of the game is intended to provide, can be unintentionally delegated to the technician who programs the play. There are a multitude of decisions which must be made in the design of a game—if the programmer does not refer the basic decisions to the proper level, he has usurped responsibility and thrown his own bias into the play. Many war games are only elegant treatments of peripheral factors to which the big model simulators happen to be adaptable. Dependence on elaborate models and simulators tends to overemphasize the variables that lend themselves to representations which the computer can handle and to de-emphasize the judgment factors.
The limitations on gaming are basically derived from subjective judgments of programmers, umpires, and players; time requirements, imponderable morale, and esprit de corps factors; complexity of enemy reactions, and scarcity of factual input data for significant aspects of combat actions.
Gaming should be considered to be an expeditious and orderly system of storing data, performing routine calculations, and organizing one’s thinking. The computation of requirements and evaluation of systems can best be accomplished in an operation which lends itself to programming into a computer as a plan and an estimate of the situation from an operational commander and which leaves few options available to the enemy. .
There are three basic principles which have been used for the a priori analysis of conflict. The three systems are: game theory, war gaming, and conflict simulation.
In 1916, F. W. Lanchester published a treatise on air power which contained a mathematical analysis of conflict. The analysis consisted of a system of differential equations which determined the effects of two military forces acting on each other. It assumed that the rate of attrition in each force depended upon the strength of the opposing force. A review of this method indicates that, because of the complexity of the equation (and the real world situation), many simplifying assumptions must be made in order to arrive at a significant solution. These simplifications appear to nullify the effectiveness of the system for the purpose of computing the effectiveness of forces. The only real-life situation which could be so computed would be the very simplest, such as that which could be reduced to a rate-of-fire and a single-shot kill probability. A technique has been used to minimize the effect of these simplifications. Two forces are compared by fighting, each individually and separately, against the same third force.
Game Theory. Game theory provides a mathematical system for the computation of the results of contests where interaction probabilities are known. Game theory might seem an appropriate method of computing the “pay-off” in a conflict such as that which is anticipated by air forces that are to be employed against a complex of land targets. This could be considered a multi-move game between two players where each seeks the pay-off in the form of target kill. Consider the game as consisting of a series of strikes, or moves, each of which consists of continuous air missions where at the start the two players, Red and Blue, have certain forces. On every strike, or play, each player may dispatch planes on each of the three missions: air defense, counter-air, and close-air support. The effectiveness of the air defense interceptors, the bombers, and the anti-aircraft installations of both the Red and Blue forces must be assumed to be a variable numerical proportion of forces committed in order that the complex equation for the value of the game can be solved. Further, game theory does not adequately treat other complexities such as the effect of saturating a defensive system, the effect of numerous penetration tactics against the early warning net, and the effect of partial destruction of defensive facilities by aircraft with large bombing errors (targets which are hit accidentally).
The number of plays required to determine the solutions resulting from the large number of options available to the players becomes phenomenal. It can be shown that if t aircraft inventories exist for Red and Blue, if m allocations of aircraft are to be made to the strategic mission, and if n strikes are to be considered, the number of pure strategies for each side will be equal to m,tn. Thus, if m = 3, t = 9, and n = 5, mtn = 2.9X1021 and 8.41X1042 solutions exist. Clearly, such a number of computations must be performed on an electronic computer. Although game theory does not prove adequate as a method of computing force requirements and testing tactics, the volume of computation which must be performed is here indicated. Apparently, the only advantage of this system of analysis is that it could provide, under certain conditions, for the ready computation of the optimum proportion of forces to be allocated to each mission type. The optimum fraction of attack aircraft to be committed to counter-air attack, anti-air facilities, and to strategic targets must, however, be computed prior to each strike. While such might be computed by using the game theory, it will be seen that ratios can be determined by testing the limited number of logical ratios (in a simulation) for optimum results.
War Gaming. The derivation of the war game is such that it is often considered to be a tool of military planning, strategy, and tactics. The war game has been defined as: “A simulation, in accordance with predetermined rules, data, and procedures, of selected aspects of a conflict situation.” Other sources differentiate between simulations and war games. It can be argued that a control group conventional war game does not share the “general validity and high confidence awarded simulation.” A war game is best defined as play wherein personnel are required to render judgment decisions during the play of the game. Teams are formed to act as contestants, make decisions, and conduct specific operations against each other. The argument against analysis by war gaming is chiefly directed against the overriding effect of the actions of the players. The results reflect, to a high degree, the competence of the players and, in the course of repeated plays, the effect of the learning process influences the data. The players being a part of the model represent a dynamic and indeterminate variable. In open games, where all information is available to all players, competence becomes a less important factor; however, in open games the effect of strategy and tactics becomes confused. The referee or umpire, if assigned the responsibility of evaluating and making judgment decisions, affects the results in a like manner. The time required to play the game usually precludes the repetition necessary to cover the available number of optional strategies. Clearly, the construction of the game and the rules established may bias the results of the play.
In spite of the shortcomings of such a control group war game, the complexities of real life will, in some instances, force the use of that method to determine force requirements, strategy, and tactics. The German Army used such game techniques prior to both World Wars to test the adequacy of plans and forces. The Japanese used large scale war gaming to assist in the preparation of their plans for World War II. Real world results have not consistently followed the game solutions. This type of war game, conducted with men making decisions under simulated combat conditions, certainly may not produce pure scientific results. But, when the interaction of factors is of such complexity that the scientific simulation is not possible, the conventional war game is the next resort.
In any type of game, the input data which describes equipment and forces become part of the game. The compilation of these data is one of the most important parts of the game designed for any purpose other than training and is an area susceptible to fatal errors.
The results of the war game may provide data which are only estimates but, properly executed, the game will provide the best analysis of broad scope, complex, multioption conflict. It is this control group game that is most useful for the training of industrial executives and combat command staff's.
Simulation or Simulator War Games. Simulations are “operating representations of conflict events and processes.” Simulation is the dynamic representation of armed conflict programmed into a computer in accordance with a model. The model is designed in accord with scientific method as a logically connected statement of the conflict that can be programmed into the computer. Simulator models were originally designed as analog computers, but such computers were inflexible in the degree to which they could be adapted to different situations without the assistance of judgment decisions during the play. The high speed digital computers have proven more flexible and are capable of the storage of vast quantities of reference data. Thus, they are adaptable to an unlimited number of models.
The simulation, ideally constructed, has no human intervention during play and is designed for replication to provide statistical data which are valid for comparison. The computer performs three basic functions: it carries a train of events chronologically in accordance with the program logic; it performs probability computations; and it performs supporting computations, basically logistics.
The probability of success which is computed is the “pay-off.” In its very simplest form, probability of success (P8) is equal to the product of the several individual probabilities of success upon which the final pay-off is dependent. As an example, the play would consider the pay-off of a bomber mission, that is, the mission effectiveness (Ps), as follows:
Ps =PeXPrX(1-Pk)XPgeXPnXPwXPd
Ps = Pay-off, probability of success (effectiveness of a bomber mission).
Pe = Probability that aircraft will be in commission for launch.
Pr = Probability that the reliability of the aircraft will be such as not to cause the mission to abort before reaching target.
Pk = Probability that the bomber will be killed by the defensive system
Pge = Probability that the bombing error will not be gross.
Pn = Probability that the bomber will navigate to the planned target.
Pw = Probability that the weapon will operate as designed.
Pd = Probability that the weapon will do specified degree of damage to the target.
Although true simulation is not adaptable to all military operations at present, an acceptable model can be designed for operations in which the combination of options available to the enemy are such that each combination of options can be played out and the several results analyzed statistically. Simulation is the most accurate where there are few options available to the enemy. Simulation is, therefore, most applicable to operations in which complexity comes from a large number of independent parts but where the interaction between the parts is controlled or can be predicted on a probability basis. An example of best accuracy in simulation would be that of the interaction of attacking bombers opposed by defensive anti-aircraft fire. In this case, the offense has the initiative because the track of the bombers can be selected without restraint but there is only one reasonable option available to the defensive element. The only serious consequence of the anti-aircraft activity would result from its firing as rapidly as possible at every bomber that flies within range. The quantity of information and data which must be processed in the complete air defense problem is voluminous and its interaction is complex, but the options available to the enemy are virtually singular. The problem of resolving the question of enemy reaction can, in some instances, be solved by reference to enemy tactical doctrine, past performance and logic, but always the safest assumption is to assume that any enemy will react to the utmost within his capability.
Logic. The most important single advantage of simulation by high speed computer is the capability to store, select, and process vast quantities of information with incomparable rapidity. Obviously, neither the model nor the computer can perform acts of insight or imagination; such must be performed by men in the design of the model and thus the simulation must contain decisions. If conjectural assumptions are fed into the computer in the form of equations and rules of play, the results will also be conjectural to the degree that the unknowns bear on the problem; thus, all assumptions should be a command prerogative. To promote confidence in the solution, conjecture must be avoided in favor of conclusions based on logic, valid data, and on hard facts. The scientific validity of simulation stems from the ability of the model to arrange numerous expressions of logic which operate within the computer as equations in Boolean algebra.
Certain simplifications can be made in the design of the model for the purpose of reducing the complexity of stored data and the complexity of computation; however, the simplifications must be limited by the required degree of accuracy of the result.
Program. Supporting play and computation are required during the play of the war game or the simulation for the position of ships and aircraft must be computed, fuel and ammunition availability must be determined, the effects of weapons must be calculated, and the play of subordinate components of the game must be co-ordinated. To accomplish this, a computer may be programmed to interrupt the primary problem so as to provide for subroutines (computations) in the proper sequence. The importance of sequence stems from the effects of saturation of effort which must be considered. The effects of reaction time and of communication efficiency are also determined by proper simulator sequence. Even though computer time is only a small fraction of real time, the relative position of events in time remains fixed.
Probability and Performance. Probability is extensively used in simulation as it is in war gaming. The reliability of equipment, the effectiveness of weapons, and the frictions of war obey the laws of probability. All such areas must be analyzed by statistical methods to determine the correct probabilities which obtain. Statistics relative to many factors in the area of frictions of war are available in published data as are the conditions of visibility, cloud cover, wind, rainfall, snow, temperature, tides, and moon phase, sea conditions, etc.
Thus, the probability of any given condition obtaining in any location over an extended period may be computed. Performance data relative to weapons and warning systems are determined from tests results, sensor data, and from basic intelligence, and all are usually expressed as a probability. That data which is estimated thus degrades results accordingly.
The high speed computers are capable of applying probability to performance data in several ways. A random number, within certain limits, is selected and the logic system directs the proper conclusion depending upon whether or not that number is larger than a reference number specified as stored performance data. The results of a play based on probabilities will usually be obtained as the most probable pay-off. The most probable or expected value is obtained from the play if input data represent the arithmetic average of all possible results. If such is not considered to be an adequate pay-off, additional forces should be committed. With good fortune, better results may be obtained and, if less fortunate, the pay-off may be less. Minimum probable results may be determined by supplying to the computer, data representing performance which is the most disadvantageous that can be expected with a high degree of confidence. Maximum probable results can be determined by providing data which represent performance which is the most advantageous that can be expected with any reasonable degree of confidence. The stochastic system for determining the range and distribution of results involves the random process of selection of performance data. The operating speed of modern equipment makes possible multiple runs in a relatively short time. When this system is used, the play should be run a number of times and thus the statistical range and distribution of results is obtained.
Clearly, large operations are a combination of smaller interacting operations. Because of limitations of machine capacity, the large operations must be subdivided into smaller components and played out individually on the computers. Simulation of combat by the adaption of a detailed model of the conflict to a large high speed electronic computer appears to provide the most effective method for the measurement of military effectiveness.
The computation of the number of Navy heavy attack aircraft required to perform a specific operational mission serves as an excellent example of the simulation process and demonstrates the considerations necessary in such a computation. The model must be designed to simulate the offensive and defensive activities that occur during nuclear air strikes against land targets for the purpose of analyzing attrition of strike aircraft penetrating enemy defenses. This model can be readily adapted to the computation of force requirements and to the computation of probability of success if valid force capability inputs can be provided.
The Strike Model provides for the aircraft performance capability in range, speed, altitude, and delivery maneuver; further provision is made for defensive weapons and doctrine. The most complete data available on enemy defenses must be compiled for the operating plan.
The model is programmed for use on a high speed electronic digital data processing system, a digital computer. Such a system would be capable of processing over 30 simultaneous strikes. The computer requires magnetic core storage with a capacity of some 33,000 words of 36 data bits each. Other storage capacity must be available as disc and tape storage through attached units. The system is flexible in that it is an assembly of individual components. The random storage capacity of the system can be expanded to provide storage for any reasonable quantity of data.
The model is planned as an “event type” simulator game in that each event is placed in store in chronological sequence. Certain significant events are the following:
Bomber reaches early warning radar line
Raid detected by radar
Defender commits interceptors
Interceptor take-off
Intercept
Strike maneuver
Interceptor returns to base
Raid detection in surface-to-air missile area
Possible missile launch
Missile intercept
Bomber weapon release
Weapon detonation
Bomber kill
Bomber reaches target
Bomber effects satisfactory bomb drop
Weapon detonates
Raid data are placed in machine memory in detail; profiles, flight path, delivery maneuvers, targets, ECM plan, and weapon plan must be introduced. All requisite data are placed in storage; location of defensive elements, target characteristics data, missile data, anti-aircraft data, fighter performance data, reliability factors, weather data, navigation reliability data and bombing accuracy data.
Success or failure in each event is determined by comparing the probability, for the event in question, with a random number. As an example, each 1/25 second the radar horizon for a raid may be computed. A random number is then generated and compared to a curve (the penetration depth versus detection probability) to determine the point at which the raid is considered to be detected, a check is made for probable kill by defensive anti-aircraft fire. If the bomber is not killed, the raid passes on to subsequent events. In a similar manner, all events are processed in chronological sequence until the conflict reaches a conclusion in computer time or until all raids are killed.
The results of the game are expressed as an indication of the damage inflicted on targets and as the loss inflicted on own forces. The game can be run any number of times to test the probability distribution of results. If results are not considered to be adequate, increased effort is required by the offensive; thus, force requirements are determined.
Conclusion. The genius of the simulation cannot be fully appreciated from this short description but it is well suited to conflict analysis if the requisite input data is available and if the complexities of the real world problem can be reduced to logic. It must be remembered, however, that accuracy of results are dependent upon the accuracy of the input data and the program.
War gaming and simulation can both provide a basis for evaluation of forces and for the determination of equipment requirements. Electronic computer simulation appears to be the more exact gaming technique if the requisite conditions for simulation do obtain. Control group war gaming, however, may be more effective where the problem involves a large number of command options and judgments. While the hazards involved in analysis by gaming are such that it can not be construed that the results will provide irrefutable conclusions, the gaming process will certainly provide better insight into the problem which the game is designed to solve.