Proposal for revision of DC's artificial intelligence module, by Reu:




Current sub-system 1: Random toggling of play-styles.

Description and plus side:
AI will roullete through 5-6 different styles of play-style. This is to prevent the player from catching the pattern in the AI's attack strategies and exploit holes. This will prevent the player from feeling like he is fighting against an AI by emulating personality and several routes of thought.

Down side:
Due to random toggling, some play-styles would seem very unintelligent to be executed in certain scenarios, player might feel that the AI is stupid.

Solution 1: Detection precautions for such scenarios to check against before executing such play-styles.

Solution 2: Active detection for player to be in vulnerable position or cancel current play-style mode if deemed ineffective at that point in time.

Play-style 1: Rush down - extremely offensive

Play-style 2: Counter attack - wait for opponent to attack and counter

Play-style 3: Suprise attack - Keep away from opponent and sneak in attack from far

Play-style 4: Run away - Do nothing but run away. Used to break monotony of only attacking.

Play-stle 5: Defensive - Do nothing but guard. Used to break monoty of only attacking.

Play-style 6: Random - Do anything illogical to throw player off-track from reading the AI.



Current sub-system 2: Fight memory

Description and plus side: Introduce a set of pre-determined precautionary functions to prevent exploits in the character's gameplay by the player should he find a loop hole that works for most, if not all of the AI's play-styles.

Down side: The precautionary function may open up a new loop hole instead. Since it is going to be a fixed reaction every time the player attempts to exploit the existing loop hole.

Solution 1: At least 3 levels of contingency precautinoary functions. This will make it much harder, almost impossible for the player to find the loop hole. This may not always be possible.

Solution 2: Revert back to original play-style. If AI detects the player exploiting the second stage of loop hole, reverting back to the original play style will close the loop hole, while although opening the first, the player will have to set up himself to open the first loop hole, while the AI will have a chance to throw his chain of thought out or defeat him.









Problem with current system 1: AI will completely obliterate a new player.

If the AI is too easy, experienced players will not feel challenged, if it is too hard, new players will get frustrated.

Solution 1: Detect difference in health variable. If AI is found to have a huge advantage, go into wuss mode and allow player to hit him.

Solution result: After testing with actual players, the "wuss mode" seems to be an insult to some players. By going easy on them, they feel that they have lost their chance to get even. Perhaps this is flattery to the AI system, since it can invoke such feelings.

Solution 2: Detect time spent in hit state. Longer time spent in hit state, would allow AI to perform more damaging combos, since a new player would not be able to produce very long combos. Objective result would be to produce proportionate damage for both sides.

Down Side: This solution requires a dependency on the player's actions. Hence, if the player increases his attack time at a fast rate, the AI may not have sufficient life variable to even out.


/End proposal





My quads are damn suan. At least I know that I can still sprint as fast as the chest-mastah.

Went climbing today. Maybe wasn't such a good idea to go after gym? Since like a girl could go further than me for a route? Anyway, main reason for going was to brush up on technique, and train fingers.

Perhaps I got more than I bargained for... Got lots of good advice from the experienced climbers there. But I think I sprained my fingers - again.

Who the hell sprains his fingers every time he goes bouldering? Good job, Reu.

Eyes are sleepy, but need to go run later.