Firaxis has in the past revealed and exposed variables that impact the behavior and performance of the AI players so that the variables can be adjusted in the editor to effect the enjoyment levels of game play.
I have identified a key variable (actually a set of variables) that needs to be made available in the editor in order for us to approach some level of functionality at higher levels of difficulty.
Currently there is only one variable called “AI Cost Factor” that is adjusted to set the advantage that all the AI players have relative to the human player. This one factor impacts every AI player the same and impacts all aspects of each AI player.
As an example, at Deity difficulty level the AI Cost Factor is set to 6 while the cost factor for the human player remains fixed at 10 and this yields the following cost advantages for the AI.
Nothing in this post will complain or whine about the advantages given to the AI players at standard difficulty levels.
The code should be revised to include a simple change that lets the editor set different cost multipliers for five different activity areas instead of just one cost advantage factor that distorts all teh game play areas equally. These factors should be:
Population Growth Rate
Unit Production Rate
Technology Research Cost
Improvement and Small Wonder Production Rate
Great Wonder Production Rate
All of these cost factors should be set to the current default difficulty levels in the CIV3 product so that the standard product will represent no change to game player or the end user.
I would implement each of these cost factors by including a cost factor for each type of activity and a percentage factor that would be applied to determine the probability that each AI player (after the first AI player) would have the advantage. As a default this percentage factor would be set to 100% so that all the AI players would be certain to have the advantage. If the percentage factor were adjusted to 50% then the first AI player generated in the game would have the full AI cost advantage and all subsequent AI players would have a 50% chance of having the AI cost advantage and a 50% chance of being the same as the human player in this particular cost factor.
A third enhancement that could and should be included would be the civ specific advantage multiplier and selection dialog box. Each civ could have an individual cost advantage multiplier that would be multiplied times the standard cost advantages to help differentiate the different civs from one another. The selection dialog box would be located at the general difficulty level setting page and would determine if the civ specific multiplier would apply to any one of the five cost factors or the additional choices of “none” and “all”.
I have only been able to test the play impact of one of differentiating one of these cost variables because the current system has hard coded all the AI civs to be the same and at the same time hard coded all these civs to have the same advantage relative to the human player. The only variable that can currently be play tested on a limited basis is the civ specific unit cost.
The results of this first test were so significantly positive as to warrant immediate sharing of the results.
To test the results of providing more thoughtful application of the cost factors, I created copies of all the key early units and then made them available on a civ specific basis only. As an example, on Deity level I created a unit called the “eqSettler” or equalized settler using all the standard graphics. I then set the cost factor for the eqSettler to be 20 shields and 1 population point (that would cost 20 grain bundles) while the AI civs still used the standard settler cost which only cost then 18 shields and 2 population points (at total equivalent of 24 grain bundles.) I interpreted the eqSettler to be about as close as possible to being equal to the AI version of the regular settler. To test the impact of equalizing the settler production capabilities, I made the eqSettler available only to the human player.
I also tested the impact of equalizing the costs of other units by using a similar eqUnit approach to other units like spearmen, archers, swordsmen, horsemen, and pikemen. This approach basically made the human player equal to the AI players in the area of unit cost and unit production but left the AI players with a supreme advantage in all other cost areas including population growth, research, and infrastructure.
The equalization approach could also be tested in reverse, by restricting the eqUnits to only be available to the AI civs while reducing the AI cost Factor for the difficulty level back to being equal to the human player (10) this would make all cost and growth factors equal but would still allow the AI players to exploit a significant advantage in unit production rate.
Just by differentiating the AI unit cost advantages from all the other advantages, the enjoyment factor of playing at pseudo deity level was vastly increased. When the unit costs were equalized, the growth and expansion rates of the civs were more balanced even though the AI’s continued to exploit a significant set of advantages due to their one settler head start and enhanced growth rate combined with 40% reduction in the costs of technology, wonders, and improvements. When the AI civs were only given a unit cost advantage, they could expand quickly and support that expansion with increased military units at an accelerated rate but at least the human player had some chance of building more than perhaps one great wonder in the entire game.
A key point in this proposal is to provide some differentiation between the AI civs at the higher game play levels. The current implementation has the AI civs set to a higher level of game progress in every way and results in substantially less difference between the different AIs on the Emperor and Deity levels. All of the AIs almost approach a psychotic level of builder frenzy and settler diarrhea leading to the singular appearance of ganging up on the human player. Depending on the “luck of the draw” from the random number generator, the AI players either get into early conflicts and create opportunities for the human player to win or the AI players end up balanced while the human player is terminally squeezed without opportunity.
