Spellshore Tactics

[Firekraker51]

I think I agree with that, yeah. If you could do interesting things with the terrain, could completely make the game much more interesting to try to win. A lot less repetitive, perhaps.

[Greg]

Man, everything's too repetitive for you guys. You've been spoiled with all those fancy new games with like a gazillion programmers. In my day* we shot down the same damn aliens with the same 7-pixel tank and we liked it**, dammit.

Incidentally though any ideas for bonuses? Obviously with the addition of the cost system it would be easy to make ones that increased or decreased cost, but anything else?





*not actually my day
**I didn't actually like it that much.

[Firekraker51]

Hmm, you could make a bonus that only shows up on unowned territory, and when you claim the tile that has that bonus as your own territory, you also get +1 claim on every adjacent tile as well. There could also be, perhaps, something you pick up, such as a bomb, that you could use on enemy territory to make the tile you use it on, and all adjacent tiles, lose some of their claim.

Oh, and I just had an interesting idea. This would only work, though, if you wanted to make more than one type of terrain for your map, each distinct. Perhaps at the start of the game, you could make everyone choose a "type", which correlates to which terrain you could more easily conquer, and which you are not as adept at claiming. Such as, if you pick a type that is adept at, say, grasslands, then for each grassland tile you put claim on, it would be worth 2, and for each on the opposite tile, say, mountain, (or would it be sea? Hm.) any claim you put on it would only be worth half.

Then again, that sounds way too complicated to code, so if you don't wanna do it, then nevermind.

[Greg]

It's not too complicated to code, all it would involve is adding an extra variable to each tile and comparing this with a corresponding variable in the player data during the cost calculation. What I am trying to limit though is making the game too complicated to understand. It's already a lot more complex than I intended and I don't want to clutter people's minds with variables. It's a neat idea, though, I'll think about it.

Also, I'm now working on improving the AI since what's currently there is mostly just a placeholder for target practice, all it does really is make a list of tiles it could afford in its influence and chooses one randomly. It completely ignores bonuses and can't place buildings.

This sort of thing is where an education in computing would come in really handy so I could use your help, Nathan, but initially I was thinking just assign value points for certain features, like bonuses, buildings and cost, and then pick the tile with the highest value. Of course, if I wanted it to have proper strategy I'd also have to somehow assign points for strategic significance but that's just way beyond me.
Something that would be really cool and not too hard is to evolve the AI by having variables controlling the value of features, and then incorporating winning values into a database that game instances can download from. It would be hack city though so there would have to also be a default button.

Placing buildings will be a bit more of an issue, though. There would have to be a quick way of locating the right shape on the board, and planning ahead to make shapes for the buildings to fit into. And that's not even considering tactical importance like putting walls on the edge and cities in the middle.

[Nathan]

Assigning values is called a heuristic. Typically you don't assign strategy yourself you use a heuristic search. SO calculate several moves ahead and see where the best board is based on the heuristics. For a game like this you need to take into account your opponent so yoiu have a max-min search, where you assume your opponent plays the best that you can.
These are algorithms that are easy to find.
But they might be overkill.
It might well be easier to just program a few techniques and just look ahead one move.

[Greg]

Yeah, sounds like overkill, maybe save it for a sequel. Man, we say that so much it should have an acronym, I hereby announce SFS as forum slang (not to be confused with Sexual Frustration Syndrome or Somalian Forehead Syndrome).

But yeah, I'm not sure how to go about programming a strategy. I suppose the basis of it is to get the best value for your claim, i.e. make your moves go 'further' than your opponents'. The main barriers I can see are:
1) forming chains to bonuses, this involves looking ahead several moves to calculate the minimum cost of taking a bonus vs its benefit, and then making the required moves to take it.
2) placing buildings, this requires first 'knowing how' to use buildings, each one will probably need its own instructions, and it'll also need a mechanism for reserving and taking over the tiles needed. I could just have it place buildings randomly into any slot they fit in but that seems a bit too crude.

[Nathan]

Well, for buildings that need protecting , search for places near the home base. For buildings that have specific benifets search in order from the place they have benefit.

[Greg]

Well, search how? All I can think of is using a tree to check tiles within a a couple of links of the home, then checking all the tiles around each one in the base pattern to see if they're suitable for building. If no suitable bases are found, go back to the one that had the most complete base shape and increase the value of the remaining tiles.

Sounds like a pain to program and it'll probably lag it up, though. Especially since it has to do that every move it has a building.

[Greg]

OK, I'm officially getting vexed. I had this neat idea of using a recursive loop which moves out from a starting point, exploring all available routes, then passes values back up the tree to the root assigning values as it goes, so in other words, each tile is assigned the value of all possible paths passing through it. It's a simple enough idea, but for some reason I just can't get it to work. I'm not sure if it's a mathematical error on my part or a feature of AS. At the moment I'm trying to pass values using return, but that doesn't seem to be working out.

[Silon]

Doesn't sound like my kind of game but my opinion doesn't matter anyway.

[Nathan]

Doesn't sound like my kind of game but my opinion doesn't matter anyway.

thats good to know.