Color Blind
Colorblind is a concept game that would allow a player to control an assassin, in a town of color blind citizens. Who each wear etheir a Red,Green,Blue sweater with a Red,Green,Blue Scarf. After you assassinate someone, the citizens who witness it yell out (Speech bubble) the colors the murderer was wearing, and run to the nearest guard. Although their is one problem, as you could guess.
Citizens make their way to the guard to report the crime with a possibility of messing up certain colors. They also spread this to other citizens, as common beliefs spread. Guards then have to figure out the color of the player, and will eliminate that color from the city.
However the player can change either the shirt or scarf after killing a guard, and the AI needs to account for that. The bonus for them is guards have less probability of being color blind.
Citizens make their way to the guard to report the crime with a possibility of messing up certain colors. They also spread this to other citizens, as common beliefs spread. Guards then have to figure out the color of the player, and will eliminate that color from the city.
However the player can change either the shirt or scarf after killing a guard, and the AI needs to account for that. The bonus for them is guards have less probability of being color blind.
First Guards used a simple probability calculation to accommodate for the color blindness of the citizens. Above is a statistics sheet of the eye witness reports. I imagine it like a bulletin board at a police station. Since the highest number of Sweaters of one color is Red(4+2+4) is 10. Red(4+1+6) Scarf wins as well. The reason for the +4 and the +6 is because we trust guards twice as much (or exact ratio if Ai knew the amount) since we trust guards over citizens, other guards will be doing these calculations as well, so info shared amongst guard could be quite accurate.
So Red/Red is the most likely candidate. If a guard hasn't been reported killed in a long time. (So player has not changed color for a while). The guard is even more certain of the color. After a specific amount of certainty, the guards will massively execute that color from the city, assuming the death toll wouldn't bring the player too close to victory.
A second calculation is run, which calculates the change in trend in color reports. For example, if
RR 18 RB 20
BR 2 BR 3
So the scarf is defiantly not blue or green. But if a trend quickly spikes past the ratio of B:R (if 18:2, so roughly 9:1) and becomes something like (3:1). It is safe to assume the suspect has changed shirts. That must also mean a guard has died. Meaning the player has figured out guard routines. They use Nav-Mesh(With Unity) way-point routes like the civilians do. It is hard to kill a guard so a trap is necessary. Means it's time to change it up. Possibly send more guards to that quadrant of town, or the one's surrounding that area.
Third calculation which could be beneficial could be to track number of times a player is spotted (by dead guard or spotted by guard) which direction (other quadrant, or even routes) he most commonly takes. And the preemptively set traps.
So Red/Red is the most likely candidate. If a guard hasn't been reported killed in a long time. (So player has not changed color for a while). The guard is even more certain of the color. After a specific amount of certainty, the guards will massively execute that color from the city, assuming the death toll wouldn't bring the player too close to victory.
A second calculation is run, which calculates the change in trend in color reports. For example, if
RR 18 RB 20
BR 2 BR 3
So the scarf is defiantly not blue or green. But if a trend quickly spikes past the ratio of B:R (if 18:2, so roughly 9:1) and becomes something like (3:1). It is safe to assume the suspect has changed shirts. That must also mean a guard has died. Meaning the player has figured out guard routines. They use Nav-Mesh(With Unity) way-point routes like the civilians do. It is hard to kill a guard so a trap is necessary. Means it's time to change it up. Possibly send more guards to that quadrant of town, or the one's surrounding that area.
Third calculation which could be beneficial could be to track number of times a player is spotted (by dead guard or spotted by guard) which direction (other quadrant, or even routes) he most commonly takes. And the preemptively set traps.
How to mess with the Ai
Purposely let live the ai who misreport in the same scarf color. Ex) All Ai who go to report a green scarf, when your wearing a red one. The ai will have a harder time, but it'll depending on your skill of letting the right ones flee the scene. You can throw knifes, however you need to retrieve those, and guards will confiscate it when they find the body. The citizens can bleed out and run for some time depending on the accuracy of your knife.