Posted July 01, 2020
Hi all!
It’s time for our next developer diary. Today we want to give you some more insight into one of the key systems in The Guild 3: The AI. Before we go into more detail on that topic, we want to say a few words on why AI is the topic of today’s dev diary. We received lots of feedback from you, both in the forums and also as comments in our surveys. The bottom line is that the AI dynasties are not as big of a challenge as they are supposed to be. What’s also mentioned often is that they act too arbitrarily. We absolutely see it the same way and that’s why we started to focus more on the behavior of the AI, both in aspects of functionality and balancing. Before we can go into more detail on how we want to improve the AI, we need to take a step back and give you more insight on how the AI is currently working.
Every action a family executes has an impact on their dynasty values like honest, cunning, generous, etc. You can find those values in the dynasty UI. Those values both represent how a dynasty is perceived and also how likely they act in certain situations. That’s an important point because those values have a big impact on how likely the AI is going to choose certain actions. Additionally, those values also have influence on the objective a family is trying to achieve. Before an AI family member decides which action to pick, the family gets an objective that they want to accomplish. This objective contains a pool of suitable actions.
For example “Grow family”. The objective of the family is to increase the amount of family members. As you know, there would be several ways of achieving this. They could adopt a child, have a baby or even try to marry somebody from another family. Based on the values of the dynasty and the objective they are following, an internal “weight” is calculated for all the suitable actions linked with that objective. Then the AI selects one of the actions and tries to execute it.
Usually, actions require a target. Let’s take the “Attack Someone” action as an example for now. So the AI is supposed to attack a character of a competing dynasty. How is that character chosen? Currently, and this is one of the problems we are facing, the selection algorithm is very simple. Right now, the target selected would be a dynasty that is one rank above the attacker’s dynasty. However, the attacker’s dynasty could actually have a good relationship with the dynasty of the higher rank, or the dynasty of the attacked could be so much stronger that it would be an obvious suicidal attempt to even try. Just with this small example you can see where the problems are. Additionally, the variety of objectives already supported by the game is not yet as big as we think it needs to be to create a good gaming experience.
As you can imagine, there are several different algorithms for different actions but with the examples above it’s clear why we want to make significant changes to improve that system.
So we’ll review all actions and objectives to fix and improve them. As mentioned above, the target selection process is also under review right now to improve and ensure that the targets are reliably found and make sense in the context of the respective family’s progression.
Another big topic we are working on is to create “personalities” for the families. Currently, they are just surviving, however, to make them into worthy opponents, that is not enough. We want the AI families to be able to thrive in their own way. Their personality will help to guide them throughout the game. In order to achieve that, they will be given a bigger pool of objectives. Some of them will be designed to compliment the personalities and are used only if they match the personality of the family. We want to expand the regular pool of objectives to include more game-relevant behaviors.
So, as you can see, there are several issues with the AI system that’s in place right now, but as you can imagine, we’ll not keep it at that at all. We started focusing on the AI and with our upcoming patches (after EA 0.9.9), you will be already able to experience these changes.
So what are we actually changing? First of all, we’ll add many more objectives in order to create more variety for the AI in the way of choosing an action and also to narrow down actions only to objectives where they actually really make sense. We’ll investigate every target selection algorithm and extend those in order to make the target selection more meaningful and logical. Please note, that of course there will always be an element of “randomness” or “irrationality” involved, so the AI doesn’t always appear to make the perfect decision and to avoid being too predictable for the player.
We’ll give you more detailed updates on the AI with every release where changes to the AI are made. We hope you will like the plan we are putting in place for the AI and are looking forward to your feedback.
It’s time for our next developer diary. Today we want to give you some more insight into one of the key systems in The Guild 3: The AI. Before we go into more detail on that topic, we want to say a few words on why AI is the topic of today’s dev diary. We received lots of feedback from you, both in the forums and also as comments in our surveys. The bottom line is that the AI dynasties are not as big of a challenge as they are supposed to be. What’s also mentioned often is that they act too arbitrarily. We absolutely see it the same way and that’s why we started to focus more on the behavior of the AI, both in aspects of functionality and balancing. Before we can go into more detail on how we want to improve the AI, we need to take a step back and give you more insight on how the AI is currently working.
Every action a family executes has an impact on their dynasty values like honest, cunning, generous, etc. You can find those values in the dynasty UI. Those values both represent how a dynasty is perceived and also how likely they act in certain situations. That’s an important point because those values have a big impact on how likely the AI is going to choose certain actions. Additionally, those values also have influence on the objective a family is trying to achieve. Before an AI family member decides which action to pick, the family gets an objective that they want to accomplish. This objective contains a pool of suitable actions.
For example “Grow family”. The objective of the family is to increase the amount of family members. As you know, there would be several ways of achieving this. They could adopt a child, have a baby or even try to marry somebody from another family. Based on the values of the dynasty and the objective they are following, an internal “weight” is calculated for all the suitable actions linked with that objective. Then the AI selects one of the actions and tries to execute it.
Usually, actions require a target. Let’s take the “Attack Someone” action as an example for now. So the AI is supposed to attack a character of a competing dynasty. How is that character chosen? Currently, and this is one of the problems we are facing, the selection algorithm is very simple. Right now, the target selected would be a dynasty that is one rank above the attacker’s dynasty. However, the attacker’s dynasty could actually have a good relationship with the dynasty of the higher rank, or the dynasty of the attacked could be so much stronger that it would be an obvious suicidal attempt to even try. Just with this small example you can see where the problems are. Additionally, the variety of objectives already supported by the game is not yet as big as we think it needs to be to create a good gaming experience.
As you can imagine, there are several different algorithms for different actions but with the examples above it’s clear why we want to make significant changes to improve that system.
So we’ll review all actions and objectives to fix and improve them. As mentioned above, the target selection process is also under review right now to improve and ensure that the targets are reliably found and make sense in the context of the respective family’s progression.
Another big topic we are working on is to create “personalities” for the families. Currently, they are just surviving, however, to make them into worthy opponents, that is not enough. We want the AI families to be able to thrive in their own way. Their personality will help to guide them throughout the game. In order to achieve that, they will be given a bigger pool of objectives. Some of them will be designed to compliment the personalities and are used only if they match the personality of the family. We want to expand the regular pool of objectives to include more game-relevant behaviors.
So, as you can see, there are several issues with the AI system that’s in place right now, but as you can imagine, we’ll not keep it at that at all. We started focusing on the AI and with our upcoming patches (after EA 0.9.9), you will be already able to experience these changes.
So what are we actually changing? First of all, we’ll add many more objectives in order to create more variety for the AI in the way of choosing an action and also to narrow down actions only to objectives where they actually really make sense. We’ll investigate every target selection algorithm and extend those in order to make the target selection more meaningful and logical. Please note, that of course there will always be an element of “randomness” or “irrationality” involved, so the AI doesn’t always appear to make the perfect decision and to avoid being too predictable for the player.
We’ll give you more detailed updates on the AI with every release where changes to the AI are made. We hope you will like the plan we are putting in place for the AI and are looking forward to your feedback.