Hero AI Contest

[Elprede]

HERO AI CONTEST #1



Contestants must program an AI that controls a single hero to do battle. The winner is determined in a tournament, in which the AIs test their mettle in a series of duels or team battles.

---

• Submissions must follow the contest theme.
• Submissions may not be started / made before the official launch of the contest.
• Teamwork is not allowed (physically working on someone else's entry).
  • Giving feedback isn't considered teamwork and is always welcome. However, if you feel someone is abusing it, you may appeal to the appointed moderator.
• Post 2 Works-In-Progress (pictures or videos) in the contest thread during the contest process. It must be clear that you are working on the submission.
• The submission must be posted in the respective resource section, before the contest ends.
  • The resource description must make it clear it's a submission from the respective contest.
• Post your final entry in the contest thread before contest is over. The post should be a little presentation, having your submission attached.
• Any coding language is allowed. That includes the in-built AI editor.
• Coding efficiency is nonessential as long as it does not hinder the enjoyment of watching the AI in action.
• No judges or polls needed.
• The patch version does not need to be considered as maps made in older patches can be adapted to newer ones and vice versa to a certain extend. Its time to bridge the gaps!

Breaking any of these rules may lead to disqualification from the contest, depending on the weight of the issue. In case of doubts towards a certain rule you are very encouraged to discuss it in the contest thread so others may help with sorting it out.

• 1st place: 750 experience points
• 2nd place: 600 experience points
• 3rd place: 450 experience points
• Entry: 150 experience points

The three winners will receive an award icon representing the winning entry.​

​

The contest shall begin on 17 May 2021 and conclude on 17 June 2021 GMT (day is included).

---

Assigned Staff: Open

Results

 

[Achille]

This is an interesting concept though I am curious, is the AI supposed to be made modular for any POSSIBLE vanilla hero or does the contestant get to choose? A concern that may arise is that if the winner of the contest is determined by the winning hero it may not completely reflect the capacity of the AI.

So another way to consider what I am asking is whether or not the contestants get to choose what hero they are designing this AI for.

But another question I have relating to the fight itself is whether or not the Heroes will be max level when they fight, or if they will instead level as the tournament progresses?

Overall I’m pretty excited to see this and maybe even participate, but I would love more information on the actual fight itself and what it would entail.

 

[Riki]

I honestly don't think this kind of Contest can be feasible.

@Archian @Kyrbi0, what do you guys think?

 

[Kyrbi0]

I mean, I don't mind it staying open for a while to see if it garners any interest. Currently it hasn't even been a week.

But yeah, I'm not sure how much interest there will be.

 

[Cheshire]

I would love to try this out, even though I'm not sure if I would be able to do it in the allotted time. and @Achille has some good questions...

 

[Leods]

I'm curious about this, but I've the same concerns as @Achille
I guess once they are answered we will know if there is or isn't enough interest for it

 

[Achille]

So, I keep thinking about this contest suggestion, though my questions remain unanswered. I have been logging in to hive semi-frequently to see if there are any points of clarification or updates but all I see and feel are echoes of my eager curiosity. In light of this I drew up some suggestions in terms of how this may be made more feasible simply because I am bored and I hope that they may be useful.

First concerning the selection of heroes. For this suggestion I am going to be unoriginal and steal from the high IQ plotting of @FeelsGoodMan in the recent Arena Design contest. His idea, from what I understand, was to have contestants create custom arenas. After this he would run a later contest in which boss fights would be designed for the arenas. I think that such a two phase design could work quite swell in this suggested contest. The first contest could be a hero design contest, where the top entries are selected for the subsequent contest where an AI will be designed for the purpose of combatting in a tournament. Now there is the problem of garnering interest enough to design at least three competent entries. This suggestion remains of course a suggestion, and should it be opted too complicated then there is another substitute I would like to suggest:

Instead of creating an AI for a single hero, contestants may instead draft up a team of three default Warcraft III heroes of varying level brackets (4/5/6 is my suggestion). Then contestants must design an AI that can competently use their slated trio against an opposing team. This way there is a bit more fairness/stratagem to the planning other than designing a really basic Mountain King/Blademaster 1v1 AI. This of course leads in to my last suggestion/thought:

I do not believe the winner of the contest should be decided by this tournament alone. I love the tournament aspect of this contest suggestion and it plays a huge part of why I am so damn interested in the concept, but it just does not make sense to me as it is suggested. A much more sensible implementation would be in using the tournament solely as a substitute for the poll scoring which is usually around 30% of a project’s score. The remaining 70% would be made up by judges. My reasoning for this is that coding efficiency/originality absolutely should have at least some bearing on the scoring of the AI. The Arena forums are a factory of wonderful and engaging assets and maps, and I would hate to see an AI that leaks a location every time its given a simple move order.

These are my suggestions; I hope people will agree/disagree with me here because it would show interest in this contest that I still firmly believe holds real potential. My drive for rationalizing this contest suggestion is because I would love to see a streamed Hive event like the product of this contest, I can only see it being amazing for this awesome Warcraft III community of ours.

 

[Archian]

I like the idea though Although, it reminds me of the AI in bossfight contests.

 

[Riki]

@Achille, you have some good questions and suggestions. This one in particular caught my attention and could change my opinion:

Instead of creating an AI for a single hero, contestants may instead draft up a team of three default Warcraft III heroes of varying level brackets (4/5/6 is my suggestion). Then contestants must design an AI that can competently use their slated trio against an opposing team. This way there is a bit more fairness/stratagem to the planning other than designing a really basic Mountain King/Blademaster 1v1 AI.

It would be cool to implement a team battle that is managed by the AI. I also agree with your suggestion that the tournament count for 30% of what the poll would be and leave the remaining 70% to the judges.

Now however, we have the problem of attracting enough interest for this type of contest, and that is precisely what worries me about carrying it out. If we can get at least 4-5 entries that would be great.

 

[Elprede]

This is an interesting concept though I am curious, is the AI supposed to be made modular for any POSSIBLE vanilla hero or does the contestant get to choose? A concern that may arise is that if the winner of the contest is determined by the winning hero it may not completely reflect the capacity of the AI.

So another way to consider what I am asking is whether or not the contestants get to choose what hero they are designing this AI for.

But another question I have relating to the fight itself is whether or not the Heroes will be max level when they fight, or if they will instead level as the tournament progresses?

Overall I’m pretty excited to see this and maybe even participate, but I would love more information on the actual fight itself and what it would entail.

The contestant gets to choose the hero they design the AI for. You raise a valid concern. Some melee heroes are better in 1 vs 1 matches, but how a contestant codes their AI to control their hero can make a difference. Terrain and positioning comes into play here.

Heroes being max level at the start could make the tournament exciting at the start by skipping the slow buildup. Every hero will be able to use their full arsenal. This option allows the contestant to focus on coding on how their AI use their chosen hero and analyze the terrain. Leveling up as the tournament progresses can be fun as this option gives the feeling of escalation. But it can make coding the AI excessively complex.

So, I keep thinking about this contest suggestion, though my questions remain unanswered. I have been logging in to hive semi-frequently to see if there are any points of clarification or updates but all I see and feel are echoes of my eager curiosity. In light of this I drew up some suggestions in terms of how this may be made more feasible simply because I am bored and I hope that they may be useful.

First concerning the selection of heroes. For this suggestion I am going to be unoriginal and steal from the high IQ plotting of @FeelsGoodMan in the recent Arena Design contest. His idea, from what I understand, was to have contestants create custom arenas. After this he would run a later contest in which boss fights would be designed for the arenas. I think that such a two phase design could work quite swell in this suggested contest. The first contest could be a hero design contest, where the top entries are selected for the subsequent contest where an AI will be designed for the purpose of combatting in a tournament. Now there is the problem of garnering interest enough to design at least three competent entries. This suggestion remains of course a suggestion, and should it be opted too complicated then there is another substitute I would like to suggest:

Instead of creating an AI for a single hero, contestants may instead draft up a team of three default Warcraft III heroes of varying level brackets (4/5/6 is my suggestion). Then contestants must design an AI that can competently use their slated trio against an opposing team. This way there is a bit more fairness/stratagem to the planning other than designing a really basic Mountain King/Blademaster 1v1 AI. This of course leads in to my last suggestion/thought:

I do not believe the winner of the contest should be decided by this tournament alone. I love the tournament aspect of this contest suggestion and it plays a huge part of why I am so damn interested in the concept, but it just does not make sense to me as it is suggested. A much more sensible implementation would be in using the tournament solely as a substitute for the poll scoring which is usually around 30% of a project’s score. The remaining 70% would be made up by judges. My reasoning for this is that coding efficiency/originality absolutely should have at least some bearing on the scoring of the AI. The Arena forums are a factory of wonderful and engaging assets and maps, and I would hate to see an AI that leaks a location every time its given a simple move order.

These are my suggestions; I hope people will agree/disagree with me here because it would show interest in this contest that I still firmly believe holds real potential. My drive for rationalizing this contest suggestion is because I would love to see a streamed Hive event like the product of this contest, I can only see it being amazing for this awesome Warcraft III community of ours.

Coding AI for one single hero is already difficult enough. The idea of an AI controlling a team could be saved for the next contest. If we were to take in factors like how the AI uses the hero's spells, how the hero will fight (hit and run is one strategy), and how hero will use the terrain to its advantage, designing a 1v1 AI becomes more interesting. A contestant could also design their AI in a way that takes into account how the hero fights each specific hero. A lot already goes into coding a single hero that makes it beyond basic.

The tournament could still have team battles, but the contestant is only responsible for coding AI for a single hero. The contestant may factor in how their single hero fight together with other heroes. The coding for the AI must be done before the tournament starts.

I have to disagree on the 70/30 split with the tournament on the lower end. AI is all about how it functions. Performance is important but not as important as function. A 60/40 split would be better with the tournament taking the bigger cake. You can have most original and efficient coding for the AI, but if the AI performs poorly in game, it is not fun nor useful. However, coding an AI takes quite the effort, so the coding should be evaluated as well.

I like the idea though Although, it reminds me of the AI in bossfight contests.

I would say the AI concept is more like the one in Castle Fight or Direct Strike.

 

[MyPad]

What is the primary version of Warcraft 3 that we'll be making our entries in? The list of heroes we can choose from is limited to melee heroes, right?

If so, I think I'll go with the Death Knight as an option.

 

[Panda]

I feel like it would be cool to program 2 heros instead of just 1 into that of a "boss battle" scene and whoever creates the better scene wins. Im not sure how making 1 hero as a AI, go into a "battle" is considered a contest.

Like, what type of battle is this 1 hero going into? Another hero? units? creeps?

There is also no real "theme" involved or given as well.

 

[Elprede]

What is the primary version of Warcraft 3 that we'll be making our entries in? The list of heroes we can choose from is limited to melee heroes, right?

If so, I think I'll go with the Death Knight as an option.

You can work on the entry in any version you prefer as long as it can be compiled later with the other entries. Let’s stick to melee heroes for the time being to simplify things a bit.

I feel like it would be cool to program 2 heros instead of just 1 into that of a "boss battle" scene and whoever creates the better scene wins. Im not sure how making 1 hero as a AI, go into a "battle" is considered a contest.

Like, what type of battle is this 1 hero going into? Another hero? units? creeps?

There is also no real "theme" involved or given as well.

Boss fight contests already fill the “boss battle” category. What I am suggesting is a tournament of rounds, in which it is AI vs AI and hero vs hero or heroes vs heroes. However, you are only responsible for one single hero.

This contest idea is a throwback to a contest hosted in wc3c.net.

The contest could have a theme, but I don’t see it as strictly necessary for this contest. It might even be counterproductive to the contest idea. What would you suggest?

 

[Footman16]

I think if you were able to draft up a map with the tests/tournament and how the AIs would be tested that would be a good step to showing what is expected for any potential entrants

 

[Panda]

You can work on the entry in any version you prefer as long as it can be compiled later with the other entries. Let’s stick to melee heroes for the time being to simplify things a bit.

Boss fight contests already fill the “boss battle” category. What I am suggesting is a tournament of rounds, in which it is AI vs AI and hero vs hero or heroes vs heroes. However, you are only responsible for one single hero.

This contest idea is a throwback to a contest hosted in wc3c.net.

The contest could have a theme, but I don’t see it as strictly necessary for this contest. It might even be counterproductive to the contest idea. What would you suggest?

What i would suggest is doing the AI for both heros in the "battle" having them cast spells and everything with like a cool looking backdrop. Instead of just doing AI for 1 hero because what would the other hero be doing? Would that hero have AI as well or just be a computer?

And, for a theme you could use it for specific tilesets that can be used.. like, There can only be "winter" tilesets used or "Summertime" tilesets used in this "Boss fight" Or could just be everything goes as well.. not sure.

 

[Elprede]

I think if you were able to draft up a map with the tests/tournament and how the AIs would be tested that would be a good step to showing what is expected for any potential entrants

I will see if I could cook something up.

What i would suggest is doing the AI for both heros in the "battle" having them cast spells and everything with like a cool looking backdrop. Instead of just doing AI for 1 hero because what would the other hero be doing? Would that hero have AI as well or just be a computer?

And, for a theme you could use it for specific tilesets that can be used.. like, There can only be "winter" tilesets used or "Summertime" tilesets used in this "Boss fight" Or could just be everything goes as well.. not sure.

Each contestant is coding the AI for one single hero. Then their AIs will face each other in the tournament. That is the gist of the contest. It is not like other contests, in which entries are isolated and judged individually.

As for your suggestion on theme, that could work as it would affect the map chosen for the AIs to fight in.