With the cheat-free PBEM/tournament idea in mind, John Harrison organized the Experiment game. The game itself ran through almost 10 turns, between mid-February and early April 1996. For those who weren't involved in the game, the basic premise was that to make the game fair, everyone should be allowed to revert and do anything else explicitly allowed by the Warlords program, since otherwise a cheater could do those things secretly to gain an unfair advantage.

While editing of the saved-game file was not allowed, it was explicitly agreed that one could exploit any feature of the game software (particularly the revert) in the quest for victory. If you had gold, you could revert the start of the turn until you got a hero with three allies in a strategically critical city. If you visited a temple, you could revert until you got an easy quest. In a battle, you could revert until you won. Moreover, you could pretend to end your turn, load up the other player's turns, and peek at their armies and stacks and builds, before returning to your own turn and using that intelligence to your benefit. Anything that you could do within the Warlords application was acceptable.

The game was begun, using a random Mac scenario and the default Mac army set. On his first turn John found some cash in a ruin. This let him get a hero and three wizards on the second turn, and with wizards reverting their way to victory all over the map, he very quickly overran a neighboring player. But John, as game organizer, had chosen his side last, and as a result he started in a relatively weak and isolated position. So fortunately the rest of us had time to catch up.

It soon became clear that the changes in rules had a big effect on the game: the most effective tactic was to revert a few hundred times until you received a hero and three wizards in your forwardmost city, and then to use those wizards to overrun everything in their way, reverting battles as often as necessary to win. In places where water and mountains slowed down the wizards, dragons proved invaluable. Meanwhile, quests were fulfilled at an amazing rate. By turn 10, half the players had been eliminated, and the survivors still had 3-5 heroes apiece, with dozens of archons and a variety of wizards and dragons running around.

Not only did wizards dominate the game, but there was no point in building any support troops aside from bats and scouts, which were easy to build in large quantities, were extremely mobile, and which could be dispersed into forests and hills and over mountains and water to provide some measure of defense against the wide-ranging wizards.

It became clear by Turn 10 that with heroes, quests, and the default army set, the game was essentially unplayable. In order to stay competitive, I found myself reverting at least 300 times per turn, and turns were taking several hours to complete. However, the fast pace of the game meant that battles were generally fought between single units. Besides, it didn't pay to build up large stacks because a single attacker and a few hundred reverts could kill almost anything. 8 scouts could take out 4 dragons with only an hour's work!

So individual battles did not require many reverts. The bulk of the reverting was done to get heroes and allies in the right places, and the next largest source was in getting good quests and good quest rewards. (After all, why settle for a quest that will take two turns to complete, when you can get one that you can complete this turn? And why settle for three archons when you can get five?)

With battles occurring primarily between single units, it was also clear that the default army set was hopelessly inadequate, since only the fastest-producing and most-mobile units had any value whatsoever. The numerical strength of units was largely irrelevant, since even a bat could bring down a dragon if one had enough patience and a fast computer. Also, being assured of victory made expansion much more rapid - I used scouts and bats to capture griffin-cities and pegasi-cities, for instance - and as a result the game was thrown off of its usual pace, suggesting that in future games the army speed and build times should be made a bit slower.

However, if one eliminates all the reverts needed to get heroes and allies and to complete quests, and if one were to balance the army set so that in one-on-one combat all units had an even chance of winning, almost all the reverting would be eliminated, and turns would take a reasonable amount of time. So I think the basic premise of allowing reverts is sound, provided one eliminates the need for the bulk of the reverts. This knowledge, gained as a result of the Experiment game, forms the basis of the Tournament Philosophy and Motivation, and leads directly to the Tournament Rules.