A Macromedia Flash 8 game to try out mechanics similar to chess and maybe eventually advance wars/fire emblem.
The controls and win states are given in game and in the screenshot, replicated here so its in text form.
The objective of the game is to destroy all opponent units, or take control of all the mines (squares that are around a diamond). De-Miner units convert a neutral (grey) or opponent mine to your color, at which point you can move your other units safely accross. Flying units, which are the Bomber and Fighter, will not be destroyed by mines.
Select a piece with with the mouse and drag to place. All units move 1 space of the same size. Land units are a single small tile, Air units are a 2x2 large tile.
| Name | Description |
|---|---|
| Bomber | Moves one large square at a time and destroys aground units under them |
| Fighter | Moves one large square at a time and destroys air units |
| Soldier | Moves one square at a time and attacks any ground units |
| De-Miner | Moves one square at a time and cannot attack. They convert the mine they are standing on at the start of the turn. |
| Bazooka | Moves one square at a time and attacks all unit types |
- Deminers actually head toward opponent mines
- Code cleanup and restructuring.
- Changed sides to be numeric instead of strings to allow indexing into the global unit array, which has also been added. Large structural chanegs won't impact gameplay, but allow for simpler code and implementation of future code.
- Added ability to change whether a side is controller by a human or the ai.
- Moved AI execution into a setinterval when using ai vs ai because where it was caused too much recursion
- Can now begin to analyze patterns in the ai and notice any stupid moves it makes.
AI could use some work. It needs to be tested with some one on one games. I'm thinking of adding multiple difficulties by structuring the AI and expanding it. I'm thinking of layering it so that an individual unit knows what to go after, what it's strengths are, then make it decide which unit of the randomly chosen type is in the best position. Then finally have it choose the best type of unit to move.
The code could use some better structure. I should use some more classes for packaging unit types and a sides units. This may also make the above ai improvements simpler.
Another update to the graphics would be in order, though this simple version is easy for testing, and I want to keep them easily identifiable.
A point was made about unit movement range. That would complicate the AI a bit but I can see where some units could benefit. I'm thinking soldiers could use an extra square to give them some actual tactical bite over the granadiers. Possibly the fighers but they don't kill anything but bombers anyway so I'm unsure.
Add automatic movement to both sides and a delay between them to alllow simming. If I see some games play out I might be able to reevaluate the design.
Thinking about it now, It would almost be worth it for the chosen unit to have a weight of which unit is best to attack. So each unit type would have an affinity for attacking certain units, and then we would multiply or do some other scaling by distance to choose the best target to move toward. Then we take that further by evaluating each units best score and moving that one. Though that may result in the cpu chasing down the user one unit at a time. Using the weights on at least a per unit basis could be useful to make the unit flee a dangerous situation, or stay in place.
Need to allow for human vs human games and eventually peer to peer or some form of online. I need to play against a human opponent so i can see strategies develop and see if the game itself is fun.The size of the grid may need to shrink.
- Set up each unit type with a prioritized list of attackable unit types. Likely be better for bazookas to go after bombers instead of ground units. makes sense to be adjustable per unit.
- Create logic to prevent a unit from moving into a space that is sure to kill it.
- Pull over what mark has done to the graphics
- Add Sound effects
- Add explosion animations
- Give it a real name
Starting to feel as though the next large ai update will take awhile and cause the game to suffer large slowdowns. Actionscript can do alot but I don't know how it will do with the stuff I want to add next.
Also began thinking of names. Field Control, was one I liked. Really the only thing I liked. I'll need to get some thoughts from other people, cause my creativity is at a low lately.
Speaking of the AI I'm thinking I can't improve more without adding that weighting system. It'll actually take a bit of effort. Here's what needs to be added:
- Functionality to determine if a unit in a specific square can be killed by surrounding enemies.
- A Path finding function, so we can get accurate distances. This scares me most, and may wait since we can use the current distance formula, but tuning will be difficult. Also units with 1 square of movement that are still trapped will move until that gets added. A decidedly stupid thing unless everyone else will get killed by moving.
- A function to determine if the move results in an enemy destruction, and also score it based on the units attack priorities.
- A scheme of rating the above and comparing them so that the unit chosen makes the best move.
Eventually I'm hoping that if that works in actionscript, we can then extend it to scoring all units and chosing the one with the best move, at that point I think the ai will be complete as can be.
Once the AI is set up and tunable, I'm not sure what changes to make. Adding sound and animation would be interesting, and so would giving the user access to the ai settings. Resetting is another important feature, possibly online play but thats a bit far off at this point.
Not liking the way this feels. It's no longer agressive enough. It is more intelligent about what a certain unit should do, but not about which unit should work.
This may be fixed by the later changes. Hope so.
It feels like the AI is now far too defensive, too cautious, too unwilling to lose units. Technically though it's too stupid to know when units are in danger, besides the random one it picks. The next change is the big one to finally give some global senses. Lets see what happens.
Well after alot of additions i did not expect to make it seems the ai is good enough to beat me in a quick game. Not sure if i'd say it's challenging but this is a first.
I have noticed that without introducing some randomness, ai vs ai games always result in weird infinite loop situations, with the same moves being made back and forth forever. Introducing even a mild amount of randomness screws up the ai, to the point of almost being random entirely.
The resizing does not work correctly as there is a flip between i and j and what they represent. A size of 12x12 seems to make the game quicker and the distance problems less evident. May go with this from now on, not sure.
Really don't know how to fix the ai vs ai problem at this point. The ai should probably vary in some way so that the sides behave differently. At the same time I'm clueless as to how to write any more. It's painful to admit it, but I have no idea how to improve this, and it needs improvement.