sman1975
Emperor
Currently working on the AI Covert Ops function. Spent most of yesterday testing AI missions, cleaning up the code so it better accommodates AI on AI attacks, notifications, etc.
Today, the focus is on the 1st component of the AI function: when & how the AI decides to attack. I'm looking at 9 different aspects of a "leader", and weighting those values to come up with a "Covert Operations Propensity" (COP) coefficient. The variables in the formula are: Victory Competitiveness, Boldness, Warmonger Hate, Meanness, Approach to War, Approach to Fear, Espionage Flavor, Offense Flavor, and Defense Flavor.
The formula weights each of these variables based on how important they are to the leader when deciding whether to launch an operation or not. The formula uses these values to determine just how aggressive the leader is. Most of the aspects "add" to the total value, but the Approach to Fear and Defense Flavor actually subtract from that total value - i.e. the more "afraid" a leader is, the less likely they are to do something so monstrous as unconventional warfare...
Below is a snapshot of the spreadsheet I've used to derive the COP value. The numbers in yellow are the weighting values, demonstrating relative importance to the overall calculation. I'm leaning towards using the 33% adjusted numbers at the bottom. This would mean that if on a given game turn and the leader has the option to launch an attack, this number represents the percentage change that they will.
I'm still working through what constitutes what gives the leader the option to attack. At this moment, it depends on how many "Scheduled but not executed missions" the leader has open (currently 3 is the max) and how many turns since the last mission occurred (currently 10 turns). So, a leader can only have 3 open missions, and the last mission had to have taken place 10 or more turns ago. If those conditions are true, they have the "option" to launch this turn - based on the adjusted numbers in the spreadsheet.
For example,
- Elizabeth has 2 open missions, and the last mission executed 8 turns ago. She does not have the option to launch.
- Alexander has 2 open missions, and the last mission executed 12 turns ago. He has a 33% chance to schedule a mission this turn.
- Askia has 3 open missions. He does not have the option to launch.
These numbers are subject to change. In fact, they're moving all the time as I run more and more 500+ turn Live Tuner games and collect data on how many missions were launched, by whom, against whom, and the total changes made (how extensive the game impact). Obviously, this will also change after players start telling me "it's too much" or "it's not enough" during testing. I still wanted to at least attempt to get the frequency of attacks looking somewhat a player would expect.
Any thoughts?
Today, the focus is on the 1st component of the AI function: when & how the AI decides to attack. I'm looking at 9 different aspects of a "leader", and weighting those values to come up with a "Covert Operations Propensity" (COP) coefficient. The variables in the formula are: Victory Competitiveness, Boldness, Warmonger Hate, Meanness, Approach to War, Approach to Fear, Espionage Flavor, Offense Flavor, and Defense Flavor.
The formula weights each of these variables based on how important they are to the leader when deciding whether to launch an operation or not. The formula uses these values to determine just how aggressive the leader is. Most of the aspects "add" to the total value, but the Approach to Fear and Defense Flavor actually subtract from that total value - i.e. the more "afraid" a leader is, the less likely they are to do something so monstrous as unconventional warfare...
Below is a snapshot of the spreadsheet I've used to derive the COP value. The numbers in yellow are the weighting values, demonstrating relative importance to the overall calculation. I'm leaning towards using the 33% adjusted numbers at the bottom. This would mean that if on a given game turn and the leader has the option to launch an attack, this number represents the percentage change that they will.
I'm still working through what constitutes what gives the leader the option to attack. At this moment, it depends on how many "Scheduled but not executed missions" the leader has open (currently 3 is the max) and how many turns since the last mission occurred (currently 10 turns). So, a leader can only have 3 open missions, and the last mission had to have taken place 10 or more turns ago. If those conditions are true, they have the "option" to launch this turn - based on the adjusted numbers in the spreadsheet.
For example,
- Elizabeth has 2 open missions, and the last mission executed 8 turns ago. She does not have the option to launch.
- Alexander has 2 open missions, and the last mission executed 12 turns ago. He has a 33% chance to schedule a mission this turn.
- Askia has 3 open missions. He does not have the option to launch.
These numbers are subject to change. In fact, they're moving all the time as I run more and more 500+ turn Live Tuner games and collect data on how many missions were launched, by whom, against whom, and the total changes made (how extensive the game impact). Obviously, this will also change after players start telling me "it's too much" or "it's not enough" during testing. I still wanted to at least attempt to get the frequency of attacks looking somewhat a player would expect.
Any thoughts?