Your starting difficulty level

[Degtyarev14.5]

Red Harvest,

Yeah, I know what you mean... I appreciate reality, I know that a strategically strong AI is just a dream... I guess I'm just dreaming at this point in time. Do you blame me? So many people here at the Org have harboured such high hopes for RT:W over the past year or more, and having seen some of what is in store for us, I'm desperately hoping that an advanced AI will restore some of our faith in the game.

Let's be brutally honest here: as I have previously made very clear, I don't care what units are included in the game, because regardless of how fantastic (in its true sense) some of them may be, ultimately, we're going to mod R:TW beyond recognition anyway. To me, it's the R:TW engine that matters.

Please refer to a couple of my previous posts on this topic to see where I'm coming from:

https://forums.totalwar.org/vb/showthread.php?t=35446 , about two-thirds down the page.

https://forums.totalwar.org/vb/showt...6&page=1&pp=30 , about halfway down the page.

Although we may have access to certain parameters (such as how in M:TW one could specify a unit's general role, eg. Kataphractoi: "ATTACKER,AMBUSH,ANTI_MISSILE,CAVALRY") I expect that the core of R:TW's AI shall be hardcoded. As much as I hate to say it, a broken AI shall stay broken, no matter how much we try to compensate with tweaks.

Edit: I am particularly concerned about how the tactical AI shall handle skirmishers. Will hastati be able to throw their pila on the run, or will they hesitate, glance around, lift their- too late, trampled by cavalry. We could tweak this somewhat in M:TW by increasing the javelin range from 2000 to 2500, which made it at least manageable, but I think you will agree that this was hardly a satisfactory solution.

I played a computer differently than a human--same basic idea as in TW. Tactically, a computer could kill me so I avoided tactical openings. But I could see far deeper strategically. The computer was good enough tactically that I could trick it into a poor strategic position by sacrificing pawns or pieces--converting its strength into a liability.

Exactly!! This is what I mean about being able to "read" an AI. One whould note that the "strength" you refer to, as you have noted, has been developed over decades of concerted effort. Most AIs lack this, and as such don't have any strength whatsoever. This is precisely why they need to resort to cheating to be competetive.

To build a truly strong AI takes time and experience with the final product. Unfortunately, this is not a luxury the programmers are ever allowed (to my knowledge.) To really do it right means the game rules must be frozen and the unit abilities must be set in concrete well before release.

I believe this is called planning.

Now if they could take a "dream team" of strong players with *differing* styles to work with the programmers, they could build a very strong AI.

Agreed. I have acknowledged this in stating that "an AI can only be as intelligent as it's programmer" earlier in this thread.

A new problem arises when you have a very strong AI: it can be tough to "dumb it down" so that the masses will play it. Most folks I knew with chess programs and chess computers complained that they could never beat the machine on any level--but the were not "serious" players, just casual looking for a 5 or 10 minute game. So it can actually backfire. Your customers might be offended by having to play with "sissy mode" options and then still struggling to win.

I believe JeromeGrasdyke has addressed this to an extent, earlier in this thread:

You might want to keep in mind that the difficulty levels are pitched slightly differently in Rome than in Medieval. "Easy" is really very romper-stomper easy - almost to the point of deserving the moniker 'kiddie mode' - while "Normal" should be a breeze for any seasoned strategy gamer, as it's pitched at the mainstream. "Hard" is where most genre fans will find a decent challenge, and "Very Hard" should give a good game to the experts. Or so we'd like to think

So we hope, Jerome, so we hope... I expect that "Very Hard" shall give experts a good game many times over and not just the first time, too.

OT: Red Harvest, I had no idea you were such a venerable chess player! Please don't read any sarcasm into that, because I really mean it. I don't care if you can't compete with Kasparov, just as I don't care that my pianistic abilities can not compare to those of Vladimir Ashkenazy. You have my respect and admiration.

Also OT: I shall be changing my nick sometime in the next few days. I guess I've finally outgrown ArseClown. vBulletin code permitting, I'll be venturing down the Dark Path as ... *cue: Shostakovich, Symphony No.5, I, development section* ... Degtyarev14.5.

A.

[Oaty]

Heres my oppinion

You know even for MTW the A.I. was'nt too bad on the battlefield. When the A.I. was at its worst was when they were greatly inferior to you. So it would constantly juggle its troops around figuring how it should be setup to defend from you. When the quality of troops is equal I do'nt think it was all to bad. IMO the A.I. was really horrible was in troop selection. It's biggest aim was to make a huge army of missile troops and peasant class units. Where this problem really stems from is its teching up. It made teching farms,mines and trade secondary. So the A.I. tries to buid an anti rush defense army. Well because it just spent all its money on crap and has very little left it can no longer tech up for income. So it all comes down to the A.I spending there money poorly from the bginning.

Also everyone plays differently some stretch there line and have no reserves, where as I go for more of a box formation with half the army being reserves and some missle units.