Tin Soldiers: Napoleonics

[Rowly]

Any chance of a Napoleonic period game guys.
Romans is great!
Cheers

[Deride]

I had responded privately, but I wanted to respond publically as well.

I can confirm tha Koios Works is working on a 3rd title in cooperation with Matrix Games. I can not confirm the exact era or type of game. (The Matrix guys would need to pipe in with any details they are comfortable with releasing at this point.)

However, I think that the Matrix community is really going to love this next effort. We have a tremendous amount of details, an incredible enhancement to our 3D engine (including some of the techniques you see in the much bigger budget games) and further refinement of our base AI capabilities. The game is going to be much more moddable than our previous efforts.

Additionally, our current team has expanded. We have 2 full-time artists in Dallas, 1 full-time artist in Germany and 1 full-time artist in Greece. We have 2 full-time developers, 1 part-time developer and 1 full-time level designer. All music/sound is being outsourced, but the game will have a brand-new music track and sounds.

In other words, I am very excited about our most recent effort, and I think the communitiy is really going to love it...
Deride

[ravinhood]

You should hire about five full time AI perfectionists really. Get rid of a couple of artists.

[e_barkmann]

why? The AI in TSJS is excellent and the gameplay is strong. Happy to have more candy

[ravinhood]

The AI in TSJS is excellent

Excellent isn't good enough for me I need "perfection". And I don't need more eye candy.

[Deride]

So, what does a "perfect" AI mean?

Is it allowed to cheat?
Is it allowed to think as long as it wants?
Must it adapt its play stype to the player?
Must it never lose?
Must it follow the same rules as the player?

I think the problem that you find in a game-based AI is that it really should be following this set of characteristics: 1) it should cheat as little as possible -- or at least not be obvious to the player that it is cheating, 2) it should be fast enough that it does not frustrate the player, 3) it should put up a good fight, but eventually lose to a skilled player, 4) it should be enjoyable to play against by players of various skill levels and 5) it should help "teach" the player proper techniques on how to win the game.

So, I think you will find that the "ravinhood perfect AI" is not the "Deride perfect AI" or the "Bob perfect AI" or the "John perfect AI", especially given the constraints that we have in current computing capabilities vs. a human brain -- with sometimes more than 30 years of knowledge and experience to play from.

Deride

[ravinhood]

I would settle for an AI that merely "learns from it's own mistakes". If it's charged 100 times the same way from the same direction into the same ambush, don't you think it could learn from that and try something else?

The problem with AI's is they are too "predictable", without adjustments to how they deploy, maneuver, hell even feint once in awhile, they become predictable and boring. I know when I play someone and a particular strategy or tactics doesn't work, I try something else, come from a different direction, try some feints, soakoff even.

The frustration even builds moreso when the AI has the best army, the best faction, the best resources and comes at you with 2 to 1 or better odds and still gets torn to shreds. COG ai is a prime example of this. One thing I rarely see out of any AI's is flanking abilities, with movement toward the center as a strong feint. The RTW AI is notorious for this, just a garbled rush, leaving it's flanks open for attack time and time again. No reserves, why doesn't the AI ever have any reserves?

Combat Mission has a horrible attacking AI, but, on defense, when entrenched does a pretty good job of setting up a good cross fire defense and protects it's flanks with some nice AT/AI guns. One of the most impressive "defensive" AI's I have faced.

As far as cheating, hell people cheat the AI all the time, find the exploits, look up cheat codes, anything to overcome the AI in many cases. I have no problem with a cheating AI, in fact it's what I hope for more in AI's these days since there's really few or any with any real challenge to them without cheats.

And why oh why can't the most difficult setting be difficult and/or pratically impossible? I have no problem with "normal" being a normal stupid lackluster AI, I pretty much expect it. But, on the most extreme difficulty levels damn I'd like somewhat of a challenge. There's not many that even offer this on the highest difficulties. SPARTAN is one of the few games I can give a good recommendation for the most difficult level living up to it's title. Cheats like hell, I know, I built it with the cheats I wanted it to have, more resources, an exceptional build and research tree structure, where it works to get it's best units out quickly and defends in a half decent manner in it's rear.

I'm just saying I don't think developers put enough time into structuring an AI. They take one template and put it into several games, with very little change or adjustments to make it better. A million dollars into AI production alone I think would bring about a much better AI than 10000 dollars or just a copied template from another game (HPS).

Today wayyy too much "artwork" in computer games and not enough gameplay and/or challenge.