12 replies to this topic

[LarkinVB]

so I'm not alone with my opinion

Note #1 and #7 of the first post and think about our 15 - 20 turrets (not counting those to be build at big generators soon).


I like this :

lets put it this way;
as soon as i see a turret, i know i have won....this has been right 100% of the time so far.

Edited by LarkinVB, 17 February 2005 - 04:27 PM.

[thudo]

Yea yea but the Skirmish AI micros better than any human so while a human has to scroll the screen to build turrets that time is used efficiently by the AI + as Giskard has states once we gave the AI the ability to research the tech tree and thus the req/power upgrades, he now is NOT retarded no more.

Remember: turrets ARE NOT CREATED immediately.. its a slow process and while their been placed throughout the duration of the game, the AI is massing. Hey..we can change the attack_rating to -1500 if we wanted and he'll always attack and more times than not an human player will likely not survive due to continuous bombardment. Does that sound logic: throw AI units at an enemy continuously? No it doesn't. Seems like any other dumb Skirmish AI from most of the RTS games I've played since '92 ! <damn I'm old.. hehehh>

Turrets made by humans is scroll-time wasted. Turrets made by an AI is an afterthought due to efficient microing and persistent massing regardless. Drop the attack_rating further and then tell me making turrets would cripple the AI. Attack_rating for SM, CSM, and Eldar is 150 now (was 200) and 250 for Ork (was 300).. Made a HUGE difference but imagine if we drop it more? Ouch!!!

[LarkinVB]

Seems all good anti turret arguments are going into void ....

Release as you wish but don't count on me if you get flak by good players telling you that they are wasted.

Get this : A good human player will disable all of them without ANY losses at all !!!!

I'm not against them completely. A few at start and some more at LP from tier2 on with high req/power demands. Not less than 300/150.

Lets be democratic : Thudo many turrets vote = 1. Larkin + Corsix less turrets vote = 2. You told us we are a team. So please accept our vote. If you make the AI defend turret/LP points I may change my mind.

Edited by LarkinVB, 17 February 2005 - 07:14 PM.

[thudo]

I'll make the requirement for turret building 300/150 for the Advanced Skil (Hard skill cost is already set high)l..

I'm happy with that..

I don't want to fragment the team over something silly.. The AI does need to defend and react better though. We'll need to determine that in the next release. I can't imagine the size of the AI's army. I mean its bloody big now - it'll be super gigantic now! Essshhh!

[LarkinVB]

Sorryto be unclear, for me hard=advanced.

[thudo]

Yeaa that'll get a few people why ADVANCED equates to HARDER/INSANE.

[LarkinVB]

You already fragmented the team. But its not necessarily you, I'm a dickhead too . I made a complete game on my own including balancing so I must admitt I can't stand your direction. Sorry and please don't take it personal.

[thudo]

Well I do take it a little personally as I formed these threads not to eventually see something hideous like this occur. Please read my response to your "abrupt departure".

[giskard]

You guys are kind of missing the entire point by focusing on turrets.

Any good player would choose a tactic and build his base according to that tactic. He wouldnt build his base one way then choose a tactic at random like the AI currently does.

So by rights, your AI should looking for 3 things at the start of the game that will help it win using its chosen strategy.

1) Choose a Tactic
2) Choose suitable armylist for that tactic.
3) Choose a suitable Build queue that helps the AI achieve its goal.

This is what I want to achieve in my long term goals.
My default AI is a fortifier type thats very agressive when it attacks, but by the time im done even I want know what tactic its using until I attack it or it attacks me for the first time.

My first step was giving it a nice selection of army lists at all difficulty levels. Which works fantastic btw. Next is to replace the random code used to select the army list with one that generates tactics, build queue orders and an army list.

All achieveable i believe and all stuff i'll post about as i go.

Giskard

[thudo]

Turrets are just a small part of things. Now we've made it more paletteable requiring 300req/150power so this should allow the AI to slowly "meld them in". I have always realized turrets don't win games.. turtlers never do. Maybe my fault is I felt it was such a technological achievement to have "turrets on LPs" in our first release hence the reason I was pushing it. We still will have turrets just not a turtling AI at the start and one that builds em slowly without impacting the true reason to be challenged by an AI opponent - constant, smart, reactionary, and always aggressive.

The golden egg of AI is tactics - how to make it think like an irrational human and be unpredictable. Can we achieve that in DoW where all other RTS games have never done? Don't know quite yet. Maybe we can get close but I know with the help of all of you we'll certainly get their quicker than as mavericks.

Edited by thudo, 18 February 2005 - 03:01 AM.

[LarkinVB]

The problem is that turrets won't win games but make the AI loose games sometimes. Its common sense that turrets are completely useless against good players. Is this AI mod only for bad players ?
But I guess I made my point and won't repeat myself like a broken record any more.

[thudo]

The Project is meant to appeal to everyone as well as being, admittedly, a technical achievement. If Turrets totally retarded the AI beyond belief then I would remove them in a Manhatten moment but if we "span them" over a game then their impact will be lessened + their cost now is higher so the AI will concentrate more on unit development rather than static defenses. In fact, if the AI is "raking in the dough" mid-to-late game with the req/power upgrades then we're even more golden!

So turrets are now a non-issue. We're got a good medium-ground and can move forward with other things. Whats good is if there are issues "down the road" and turrets are the culprit then we'll raise the turret cost and move forward again.

[giskard]

To be honest Larkin, your view on turrets is perhaps heavily bias towards one style of play.

In this forum alone you have ...

Larkin = no turrets
Thudo = some turrets
giskard = Fortified bases

So id say between us we roughly cover most styles of play.

Some players like my self wait to attack and when we do we walk right over the AI the first time our forces engage each other. Which means 30min to an hour just building the base and exploring the map, perhaps taking a few bottle knecks to make things easier then we go for the kill and its over in 5 minutes.

Boring......

The game ends so fast it make you feel like uninstalling Dow because it is no challenge there and the rusher game is a no brainer, always the lowest common deminator for every game since the dawn of time eg Dune 2 the first RTSG (Which i brought the day it came out and played to death btw), same tactic over and over kind of thing. When you own or have owned as many RTSGs as I have you want a little more from your games.

Besides, adding turrets to my mod has not had the effect your describing. On the contery the AI is extremely aggresssive on the offensive and sometimes, depending on its start location. Very good on the defensive too.

So theres a lot of battles for minor map points now which is what makes the game fun for players like me who want a game to last a hour or two and be a challenge the whole time rather than 10 minutes and its all over.

Lets face it, the AIs default build queue is a rusher build queue, its designed to build units as fast as possible and not much else, a few small tweaks is all it needs to do what you want to do, tweaks that are already part of my mod. We have seen the limits of this approach and its inspired the Skirmish team to want to do better.

Whether they have the ballance right yet is unknown but it is a ballancing act most of the time with lots of tweaking needed before its right.

Once they get it right, they will have a good base for variaty of different AIs.

Giskard