What is your favorite aspect of the proposed game? Why?
I love the simplicity of the gameplay because it makes it easy even for unexperienced players to have fun in the beginning and get better!
What is your least favorite aspect? Why?
Only moving one figure a time might make the gameflow very slow, as the player has to wait after each movement for the ai to move. I guess this makes it easier for AI, but I can see it slowing down the gameplay a lot.
Which single change or addition would you suggest to most improve the game?
The game idea itself seems solid and if the AI training works out, it will surely be fun. However, I am missing details about the setting, story, what levels you are going to create, will it be stages, how can you win the whole game, do your soldiers die and never appear again once they die and so on. It would be nice to have a clearer view about such things too!
What is your favorite aspect of the proposed game? Why?
I really like that you are brave enough to try the Unity ML-Agents in a game. If it works, it will definetly be a very new experience to play against an AI without a predefined ruleset.
What is your least favorite aspect? Why?
The simple mechanics of the game. Your game has a very limited set of strategic possibilities with the current concept of only three units and limited obstacles.
Which single change or addition would you suggest to most improve the game?
I would probably try to train the AI with Curriculum Learning and adding more complex scenarios with each itteration. This way you can make sure that your AI has mastered the basics of the game, before giving it a more difficult task. This way you could add more features to the game, depending on how good the AI is at learning them. Also, this will give you a better chance to fine tune your algorithms reward functions, observations, actions and hyperparameters for each small step and will probably give you a higher success rate at training a good ML-Agent.
What is your favorite aspect of the proposed game? Why?
I really enjoy your take on turn based strategy like Advance Wars ect., which features a grid and rock-paper-scissors units. This genre allows for very interesting and smart strategy design without intense workload. From the three units you have now, you are ablte to scale complexity by adding new ones.
What is your least favorite aspect? Why?
Personaly, I think choosing the AI as key feature of your game is a disadvantage. Of course it is a high risk high reward game you playing. With a working and training AI you add an amazing replay value to it, but I think the game will lack features, since a lot of effort will be put in the AI alone. If the AI project fails, it will be tough to appeal to a player with the current Game Design.
Which single change or addition would you suggest to most improve the game?
I like to see more actions that the player can do, for instance how about items or cards that can be used one-time and affect one round. It will give the player more strategical options, especially if there is a reactive AI that is facing him.
What is your favorite aspect of the proposed game? Why?
I like the technical aspect of having a "real" AI as an oppnent in the game.
What is your least favorite aspect? Why?
The simplicity, since I think especially this type of game can profit from more complexity. However, I understand that this is due to the limited time we have, and also training the AI.
Which single change or addition would you suggest to most improve the game?
Consider adding more strategic possibilites by allowing the player to move multiple or all of his units per round.
What is your favorite aspect of the proposed game? Why?
The fact that you are using Unity-ML to create your own AI. This should lead to fascinating battles of wits between human and AI.
What is your least favorite aspect? Why?
The simplistic ruleset could make it too easy to let the AI or the player to find the optimal strategy, which could lead to a deadlock.
Which single change or addition would you suggest to most improve the game?
I know due to the time constrained it would be hard to train the AIs but I would suggest adding more complexity in form of more units, terrain or fog of war. If this is not possible add a time limit for the turn of the player.
What is your favourite aspect of the proposed game? Why?
The simplicity and chess-like feel of the game. I really like it when games make a lot out of very little.
What is your least favourite aspect?
I haven't ever done anything with ML-Agents or anything like it, so don't quote me on this, but I feel like the game-rules are currently so simple, that a well-trained AI will play close to perfectly making the game frustrating rather than fun.
Which single change or addition would you suggest to most improve the game?
I think the game would benefit from some more variety. Maybe add different types of obstacles that affect the actors in different ways and are randomly placed in each round. This would add both depth to the gameplay as well as make each round unique.
What is your favorite aspect of the proposed game? Why?
The (planned) level design that includes line of sight, slower movement areas, blocking objects etc. makes the game very interesting regarding strategy.
What is your least favorite aspect? Why?
8x8 maps are too small in my opinion. Of course, it makes the AI easier but it also takes some great possibilites. If there are many obstacles, special terrains and units on the field, there won't be much space left for actual strategy, like when two areas are only connected through one bridge, the only option to encounter the enemy is this bridge which definitely takes some strategic options and I think that's what the game makes fun.
Which single change or addition would you suggest to most improve the game?
Your proposal does not include how you're planning to create the levels but I think a good random creation would be great to add variety.
What is your favorite aspect of the proposed game? Why?
I think your ambitious concept of playing against a real AI has a lot of potential for interesting strategies and scenarios.
What is your least favorite aspect? Why?
I believe there is a danger in the trade-off of simplifying the rules in order to make the game playable for the AI. It could happen that the game will be simple enough to train the AI but far too simple and with too little mechanical depth for the player to be enjoyable.
Which single change or addition would you suggest to most improve the game?
I was missing some form of progression through the matches. Are there multiple stages? Does the AI get smarter/harder to beat? I would add some more maps to create a small campaign maybe.
What is your favorite aspect of the proposed game? Why?
I really like the idea of using the ML-Agents and really hope to see some unpredictable (in a good way) AI behavior.
What is your least favorite aspect? Why?
Personally, I'm not the biggest fan of round-based games, especially if they're too simplistic, which is sort of a "requirement" for your AI. It's unclear if the game will be enjoyable, if it is small and simple.
Which single change or addition would you suggest to most improve the game?
If the game happens to be too simple or not challenging enough, you could consider making it "semi-round-based". So you, and your opponent, get to make one action every 15 seconds or something like that.
What is your favourite aspect of the proposed game? Why?
The art style looks very appealing
What is your least favourite aspect? Why?
Rock-paper-scissors interactions might be boring. It's understandable that you want to have the rules as simple as possible but I think the experience of the players might suffer because of it.
Which single change or addition would you suggest to most improve the game?
Make sure it doesn't turn into tic-tac-toe experience when both the player and the AI employ the best strategy.
12 Kommentare
Unbekannter Benutzer (ga59kuh) sagt:
What is your favorite aspect of the proposed game? Why?
Line of sight adds a lot of depth to a grid-based strategy game, even on this small scale.
What is your least favorite aspect? Why?
Using the ml-agents for your AI sounds like a good idea, however I don't think there's enough time to properly train and tune it.
Which single change or addition would you suggest to most improve the game?
At the moment the possibilities seem somewhat limited, try to add additional possibilities like e.g. height differences.
Unbekannter Benutzer (ga78rum) sagt:
What is your favorite aspect of the proposed game? Why?
I love the simplicity of the gameplay because it makes it easy even for unexperienced players to have fun in the beginning and get better!
What is your least favorite aspect? Why?
Only moving one figure a time might make the gameflow very slow, as the player has to wait after each movement for the ai to move. I guess this makes it easier for AI, but I can see it slowing down the gameplay a lot.
Which single change or addition would you suggest to most improve the game?
The game idea itself seems solid and if the AI training works out, it will surely be fun. However, I am missing details about the setting, story, what levels you are going to create, will it be stages, how can you win the whole game, do your soldiers die and never appear again once they die and so on. It would be nice to have a clearer view about such things too!
Unbekannter Benutzer (ga97zar) sagt:
What is your favorite aspect of the proposed game? Why?
I really like that you are brave enough to try the Unity ML-Agents in a game. If it works, it will definetly be a very new experience to play against an AI without a predefined ruleset.
What is your least favorite aspect? Why?
The simple mechanics of the game. Your game has a very limited set of strategic possibilities with the current concept of only three units and limited obstacles.
Which single change or addition would you suggest to most improve the game?
I would probably try to train the AI with Curriculum Learning and adding more complex scenarios with each itteration. This way you can make sure that your AI has mastered the basics of the game, before giving it a more difficult task. This way you could add more features to the game, depending on how good the AI is at learning them. Also, this will give you a better chance to fine tune your algorithms reward functions, observations, actions and hyperparameters for each small step and will probably give you a higher success rate at training a good ML-Agent.
Unbekannter Benutzer (ga78vup) sagt:
What is your favorite aspect of the proposed game? Why?
I really enjoy your take on turn based strategy like Advance Wars ect., which features a grid and rock-paper-scissors units. This genre allows for very interesting and smart strategy design without intense workload. From the three units you have now, you are ablte to scale complexity by adding new ones.
What is your least favorite aspect? Why?
Personaly, I think choosing the AI as key feature of your game is a disadvantage. Of course it is a high risk high reward game you playing. With a working and training AI you add an amazing replay value to it, but I think the game will lack features, since a lot of effort will be put in the AI alone. If the AI project fails, it will be tough to appeal to a player with the current Game Design.
Which single change or addition would you suggest to most improve the game?
I like to see more actions that the player can do, for instance how about items or cards that can be used one-time and affect one round. It will give the player more strategical options, especially if there is a reactive AI that is facing him.
Unbekannter Benutzer (ga54reh) sagt:
What is your favorite aspect of the proposed game? Why?
I like the technical aspect of having a "real" AI as an oppnent in the game.
What is your least favorite aspect? Why?
The simplicity, since I think especially this type of game can profit from more complexity. However, I understand that this is due to the limited time we have, and also training the AI.
Which single change or addition would you suggest to most improve the game?
Consider adding more strategic possibilites by allowing the player to move multiple or all of his units per round.
Unbekannter Benutzer (ga58kav) sagt:
What is your favorite aspect of the proposed game? Why?
The fact that you are using Unity-ML to create your own AI. This should lead to fascinating battles of wits between human and AI.
What is your least favorite aspect? Why?
The simplistic ruleset could make it too easy to let the AI or the player to find the optimal strategy, which could lead to a deadlock.
Which single change or addition would you suggest to most improve the game?
I know due to the time constrained it would be hard to train the AIs but I would suggest adding more complexity in form of more units, terrain or fog of war. If this is not possible add a time limit for the turn of the player.
Jan Wagener sagt:
What is your favourite aspect of the proposed game? Why?
The simplicity and chess-like feel of the game. I really like it when games make a lot out of very little.
What is your least favourite aspect?
I haven't ever done anything with ML-Agents or anything like it, so don't quote me on this, but I feel like the game-rules are currently so simple, that a well-trained AI will play close to perfectly making the game frustrating rather than fun.
Which single change or addition would you suggest to most improve the game?
I think the game would benefit from some more variety. Maybe add different types of obstacles that affect the actors in different ways and are randomly placed in each round. This would add both depth to the gameplay as well as make each round unique.
Unbekannter Benutzer (ga86lop) sagt:
What is your favorite aspect of the proposed game? Why?
The (planned) level design that includes line of sight, slower movement areas, blocking objects etc. makes the game very interesting regarding strategy.
What is your least favorite aspect? Why?
8x8 maps are too small in my opinion. Of course, it makes the AI easier but it also takes some great possibilites. If there are many obstacles, special terrains and units on the field, there won't be much space left for actual strategy, like when two areas are only connected through one bridge, the only option to encounter the enemy is this bridge which definitely takes some strategic options and I think that's what the game makes fun.
Which single change or addition would you suggest to most improve the game?
Your proposal does not include how you're planning to create the levels but I think a good random creation would be great to add variety.
Unbekannter Benutzer (ga87huy) sagt:
What is your favorite aspect of the proposed game? Why?
I think your ambitious concept of playing against a real AI has a lot of potential for interesting strategies and scenarios.
What is your least favorite aspect? Why?
I believe there is a danger in the trade-off of simplifying the rules in order to make the game playable for the AI. It could happen that the game will be simple enough to train the AI but far too simple and with too little mechanical depth for the player to be enjoyable.
Which single change or addition would you suggest to most improve the game?
I was missing some form of progression through the matches. Are there multiple stages? Does the AI get smarter/harder to beat? I would add some more maps to create a small campaign maybe.
Felix Neumeyer sagt:
What is your favorite aspect of the proposed game? Why?
I really like the idea of using the ML-Agents and really hope to see some unpredictable (in a good way) AI behavior.
What is your least favorite aspect? Why?
Personally, I'm not the biggest fan of round-based games, especially if they're too simplistic, which is sort of a "requirement" for your AI. It's unclear if the game will be enjoyable, if it is small and simple.
Which single change or addition would you suggest to most improve the game?
If the game happens to be too simple or not challenging enough, you could consider making it "semi-round-based". So you, and your opponent, get to make one action every 15 seconds or something like that.
Unbekannter Benutzer (ga97fek) sagt:
What is your favorite aspect of the proposed game? Why?
Playing against a neural network ai and getting a feel for the current AI development, but...
What is your least favorite aspect? Why?
...it might just feel like playing against a random and kinda weird ai.
Which single change or addition would you suggest to most improve the game?
Something to make the AI feel more like a "real being". Like adding a "feeling" state for the AI to spark apophenia.
Unbekannter Benutzer (ga27vax) sagt:
What is your favourite aspect of the proposed game? Why?
The art style looks very appealing
What is your least favourite aspect? Why?
Rock-paper-scissors interactions might be boring. It's understandable that you want to have the rules as simple as possible but I think the experience of the players might suffer because of it.
Which single change or addition would you suggest to most improve the game?
Make sure it doesn't turn into tic-tac-toe experience when both the player and the AI employ the best strategy.