What is chess good for?

On the face of it, the bloodless, low-tech game of chess might seem to bear little resemblance to modern warfare. "But it resembles real war in many respects," maintains Jan Kuylenstierna, one of the Swedish researchers. "Chess involves a struggle of will, and it contains what has been termed the essentials of fighting - to strike, to move and to protect." By studying chess and other adversarial abstract games such as checkers (draughts), researchers can strip away some of the confusion of the battlefield and identify the factors that are most important for winning, says Jason Scholz, who leads the Australian work. "The strength of this approach is our level of abstraction," Scholz says.

But neither group is studying standard games. By modifying key variables, such as the number of moves al lowed each turn, or whether one player can see all of the other's pieces, they are investigating the relative importance of a host of factors that translate to the battlefield, such as numerical superiority, a quick advance and the use of stealth.

"There's all sorts of anecdotal evidence that there are certain factors in warfare that are important, and people talk about having a strong operational tempo, and that kind of thing," says Greg Calbert, a mathematician on Scholz's team. "But even today there's debate over what really counts. How important is stealth over tempo, or tempo over numerical strength? That's what we wanted to find out." As well as informing fundamental military theory, this kind of information could have a big impact on how army procurement officers choose to spend their budget. There might be urgent calls for more tanks or better surveillance devices - when, in fact, to win the next war the money might be better spent on faster communications systems, for instance.

One major difference between chess and war is that chess does not contain what the military terms "information uncertainty". Unlike a battle commander, who may have incomplete intelligence about his opponent's level of weaponry or location of munitions depots, one chess player can always see the other's pieces, and note their every move. So Kuylenstierna and his colleagues asked players to compete with a board each and an opaque screen between them. A game leader transferred each player's moves to the other's board - but not always instantaneously. For instance, one game modification resulted in a player being prevented from seeing their opponent's latest two moves.

These games, and other variations on regular play, led the team to a clear conclusion: being stronger and having more "battlespace information" than your opponent are both less valuable when there is little information available overall to both sides - but the advantage of a fast pace remains. "The value of information superiority is strongly tempered by uncertainty, whereas the value of superior tempo is much less affected," says Kuylenstierna.

Uncertainty is often a problem in war. So in practical terms, launching a rapid attack might provide a better chance of winning than trying to gain more information about the battlefield situation, or ensuring that you have numerical strength over your opponent. "To what extent these findings have had any influence on decisions made by the Swedish military I dare not say - but they continue to sponsor our work," Kuylenstierna adds.

The Australian team had to write new software to allow virtual agents to play the tens of thousands of games needed for a powerful statistical analysis of the results. "We had to rewrite extensively the code for chess - and we worked really hard, believe me," Calbert says. As well as tempo, planning and strength, they looked at stealth (one agent had pieces invisible to its opponent), and the level of "networking" between an agent's pieces (involving the exchange of information on the "value" of the particular move, if any, that each piece could make).

And they also found that a fast tempo can be important, particularly in combination with "deep planning". Deep planning involved, at every move, each agent considering all their previous moves and their opponent's responses, and their responses to those responses, and using this to develop a "tree" of possible strategic paths they could follow to win. "A deeper planner is one who can search deeper into time, and has more possible end points," says Calbert. In general, deep planning plus a fast tempo was devastating - even if the opponent was numerically superior.

What's more, these findings held true whether the game was chess or checkers. And while the games might appear similar, the aims and strategies most likely to lead to success are quite different, says Scholz. He thinks that achieving the same insights from analyses of the two games suggests that his team has uncovered some general rules - rules that are likely to be applicable in other similar adversarial situations, such as war.

But what do the experts think? Retired Australian Air Vice Marshal Peter Nicholson agrees that fast tempo is a key to military success. "It's something that many military commanders have been doing instinctively for a long time," he says. "Napoleon was one of the first proponents of it in nation-state warfare. And the Mongols were another. Their rapidly moving small forces of armed horsemen completely threw conventional forces off balance."

Sun Tzu, author of the Art of War, had his own take on this: "An attack may lack ingenuity, but it must be delivered with supernatural speed," runs one translation.

There are a number of potential routes to raising battlespace tempo. A nation's military might invest in faster ships or in computer systems that can rapidly fuse new data and information from the field - leaving humans only with the job of making the decisions. "What the results encourage us to do is to look at some of the command and control decision processes and find the bottlenecks," Scholz says.

The Swedish and Australian approaches to game analysis each have their own strengths, says Kuylenstierna. While the Australians could look at many thousands of games, the Swedes used real people. But in writing the agent-playing software, Scholz's team did integrate techniques from a new area of maths called neurodynamic programming. This allowed the agents to learn as they played and, as a result, more accurately mimic real human game-playing behaviour.

Using the same new mathematical techniques, and building on the chess and checkers work, Scholz and his colleagues are now creating improved computer-based war games for use in military training. Good artificial intelligence has been lacking from most war games until now and they hope their work will provide more realistic characters and situations, and therefore not only better training but also an improved method for considering new strategies for real warfare. One important advance from the chess simulations is to allow multiple moves at the same time, as would happen in a battle.

Nicholson says he welcomes the new work, and considers the chess research one of the "tools in the armoury" for developing, testing and evaluating operational concepts and strategies. "It's not the panacea," he says. "It's one of several methods, which are all valuable and each have their place."

The Australian team are careful not to suggest that the results of their chess and checkers work should re-write the military theory books. "We've been working on this project for two years, whereas military commanders devote their lives to what they call the operational art," Calbert says. "That art certainly involves manoeuvres not just in the physical battlespace but in a social and political space as well - and those spaces are really intangible to scientists at the moment."

2,000 years of role playing games

China
The board game Go, known in China as weiqi, is a game of territory and encirclement, and has long been linked with warfare. Some of the earliest military references appear during the Dong Han dynasty, from AD25 to AD220. They describe weiqi as a game of war, and some modern scholars infer that the Chinese might at that time have been using it to model military strategies. Mao Zedong reportedly insisted his generals study weiqi - and there are rumours that today senior members of the Chinese military must be proficient at the game to progress through the highest ranks, says Jason Scholz of Australia's defence science and technology organisation.

Persia
The Persian game of Shatranj is believed to be adapted from the Indian chess-precursor Chaturanga (although there are some scholars who argue that Shatranj came first). Like the Indian version, the Persian game includes elephant pieces and horses, and Persian nobles were taught Shatranj as part of training in military strategy. It has even been suggested that pawns' ability to move two squares in their opening move in modern chess is a Persian modification, to better model a strategy in which foot soldiers with spears rushed ahead of the rest of the attacking army - but the true origins of military influences on chess, and the game itself, remain murky.

War in Iraq
The build-up to the war in Iraq coincided with the first results from the chess simulations run by Jason Scholz and his team. "We watched with great interest the dialogue between General [Tommy} Franks, who wanted to use more materiel, and Donald Rumsfeld who wanted a fast tempo and lighter units," Scholz says. Based on the chess results, which favoured a fast, decisive attack strategy, Scholz says his advice would have been to go along with the US defence secretary's ideas. "In the end, there was a compromise," he says. "But a relatively fast tempo did really gain a very decisive, rapid advantage in Iraq." However, trying to win a battle as quickly as possible might not always be the best strategy, he adds: "You can win a battle quickly but hearts and minds are not so easily won - and of course we do have continuing trouble in Iraq.""