AI

posted on January 1st, 2011, 10:14 pm

However, what we do have here is the option to not click on any thread you know to have been started by a certain user. Just wait until it's about 2 pages in, by then it should be certain to have been derailed.

Which is why I shall be re-posting this thread once I get a little more progress made.

that feature isnt implemented on this forum.

tokra may have views that are unconvential, but he has just as much right as the majority to voice them. in this case though i do believe he went too far with the multiple instances of the same complaint.

i leave you with this quote from JS Mill

posted on January 2nd, 2011, 3:01 am

As I said previously:

The FO AI is in fact very well optimized - most everything builds exactly when it needs to, no stops in construction or pauses in research, and no hang ups with units. But of course, until a whole new system of AI is put in, it is only a placeholder. A long run placeholder.

And I shall reiterate: the AI in Fleet Ops is a placeholder. A long term solution is planned, but it will take years of effort to create the suitable replacement, as the A2 AI was severely constrained when development and release of the game was rushed. For now, if you wish a better/more fair/smarter opponent, you will have to play online.

As for altering the current AI to be more reasonable you will probably have little luck other than using the Normal AI. The tools used to create the current AI build list and combat responses are far more precise and calibrated than that which can be created using manual entries. Of course, no one will stop you from trying, but you will be hard pressed to find any way of optimizing it any further.

Reposting will get you nowhere other than hearing the same tired refrain - or being altogether ignored - I am afraid. You cannot rush a part without damaging the whole. Patience is a virtue, and development of any good system requires time.

posted on January 2nd, 2011, 6:01 am

Actually Im having pretty good luck testing it so far Dom.

The problem isnt the buildlists, for the most part, nor is it the way the AI is configured to behave, it is mostly the RTS_cfg settings which control certain attributes of the AI in a game, namely the precentage of the resource cost for a station/tech/ship specified in the .odf file. That results in the AI either progressing very slow, or WAY too fast. Or to quite from your own guide.

Lastly, the RTS_CFG.h file contains all sorts of global gameplay, AI and interface settings including modifying AI / human player handicaps and changing the maximum zoom and map height.

Simply put, the handicaps were set in such a way that the minimal SP testing/forethought shows clearly.

How poorly ?

Code: Select all

float HARD_AI_COST_HANDICAP = 0.1;
float HARD_AI_SPEED_HANDICAP = 0.6;

float MEDIUM_AI_COST_HANDICAP = 0.2;
float MEDIUM_AI_SPEED_HANDICAP = 0.7;

float EASY_AI_COST_HANDICAP = 0.75;
float EASY_AI_SPEED_HANDICAP = 0.8;

Lets use the easy AI as a base example.

The easy AI pays 75% of the price of an object, and builds it in 80% the time it should take.
The medium AI (this is the Hard AI ingame) pays just 20% of the objects cost, and builds it in 70% of the same time a player would.
The Hard AI (Merciless ingame) pays only 10% to build something in 60% the time you could.

This means that unless you play on the vanilla easy setting, you will be spammed, horribly, as the merciless AI out builds you ten to one, and the hard AI at five to one.


Now, Im sure that once again people will come in saying 'this is a MP mod'

Well if so, piss off.

There is a SP feature, I am simply attemtping to juggle the handicaps to make it more playable.

Testing thusfar I'm coming close to deciding to make the easy AIs cost/time handicaps both at .6

This allows the AI to build enough ships quickly enough to at least make you pay a little attention, without being too hard for a new player to flops (not that new players would be likely to check in here for mods right off the bat... but with all teh spam this WOULD be at the top of the forum, and with a lot of replies, so they might ..)

Past that, I want to make Hard actualy hard, a fun 'actualy pay close attention' type setting.

Merciless will likely get adjusted upwards and be less silly, Im thinking I may end up going as low as 30% for the merciless AIs buld cost, but I've not tested it at that yet so any concrete number would be premature.

posted on January 2nd, 2011, 12:35 pm

Now, Im sure that once again people will come in saying 'this is a MP mod'

Well if so, piss off.

this is a MP mod and no i wont piss off, make me

this isnt a wholly mp mod, its a mostly mp mod. their focus is mp.

posted on January 7th, 2011, 8:21 pm

Testings slowed down since works picking up some and I DO play other games too.

What Im doing is playing as various races against other races as I tinker with the two aforementioned settings.

It's (as Im sure Ive said before) going to take time becasue I've got to replay a lot, with differnt settings, to determine what numbers to use for easy/hard/Merciless.

Im getting closer to deciding on settings for easy/hard, Im thinking merciless I'll focus on beign designed for team or 1on1 matches, and test it that way (and of course in such a case it'll be a tad harder if you play 2on1 or more, but not as stupidly spammy at 10%  as it was before)

posted on January 19th, 2011, 11:56 am

Why do you think the numbers are wrong? What makes you think that there weren't hours of unpaid work gone into making the AI a decent oppenent for people of all skill levels.
If you think Merciless is too hard, play Hard. If you can't beat Hard, keep practising, use tactics, stragetgy, counters to win. You will get better over time and effort. You don't need to spam, and you don't need to turtle. An opinion is an opinion, but evidence is evidence.

Also, I have to say this; it is not a surprise you think the AI is way too hard if you do not play multiplayer.

What you are effectively doing here, is (if you can mod it perfectly) is tailoring the AI to your skill level right now. Which is why no one else will download it, and in fact over time you will cease to find it enjoyable to play, because you will get better. Then you will either have to mod it again, or just go back to stock Flops.

[Tok`ra]

posted on January 26th, 2011, 2:45 pm

Last edited by Tok`ra on January 26th, 2011, 2:48 pm, edited 1 time in total.

Why do you think the numbers are wrong? What makes you think that there weren't hours of unpaid work gone into making the AI a decent oppenent for people of all skill levels.
If you think Merciless is too hard, play Hard. If you can't beat Hard, keep practising, use tactics, stragetgy, counters to win. You will get better over time and effort. You don't need to spam, and you don't need to turtle. An opinion is an opinion, but evidence is evidence.

Oh I can beat Merciless if I play rock paper scisors and counter while sniping shipyards and miners.

But the rock paper scisors balance aspect is another flawed issue for another thread.

Also, I have to say this; it is not a surprise you think the AI is way too hard if you do not play multiplayer.

What can I say, I dont like rock paper scisors balance.

What you are effectively doing here, is (if you can mod it perfectly) is tailoring the AI to your skill level right now. Which is why no one else will download it, and in fact over time you will cease to find it enjoyable to play, because you will get better. Then you will either have to mod it again, or just go back to stock Flops.

Or do like I'm doing now, and posting some numbers here on the thread, and allowing those who wish to do so to test it.

float HARD_AI_COST_HANDICAP = 0.38;
float HARD_AI_SPEED_HANDICAP = 0.38;

float MEDIUM_AI_COST_HANDICAP = 0.45;
float MEDIUM_AI_SPEED_HANDICAP = 0.45;

float EASY_AI_COST_HANDICAP = 0.5;
float EASY_AI_SPEED_HANDICAP = 0.47;

Basicly, easy AI is harder, Hard/medium AI have a wider differnce between them. Admitedly, it's actualy a narrower gap than before, but each is made  less spammy, thus you focus more on having fun than simply countering ship with ship.

Anyway, nowere near done, I still need to experiment more with varying the speed/cost modifiers up more, but after some testing ingame it seems much more reasonable than it was before.

The easy AI is still pretty easy, but able to produce ships much easier than before.

Medium/Hard AIs will (I'm hoping) be more fun to play against, and borg AIs wont rush your base with multiple cubes within ten or so minutes of game start as they often would in vanilla flops (well, at least on larger maps. On smaller maps you could at least try to drag them into an expansion battle and suck up their resources and have their build list replacing losses) . Medium/Hard AIs I aim to make a challenge to play while at the same time as mentioned above not allowing them to spam so easily/much. Easy AI I want it to still easy enough for a new flops user to not get crunched endlesley (assuming new users even check this subfourm) while still at least giving experinced players something to do other than see how many ships they can get at max rank (I think my record in vanilla flops was going thru the .odf's name list for excel2's so that all ships built with the -B suffix, tho things were lagging at that point)  or other 'im bored, I'll dick arround against an AI that cant beat me even if I fall asleep against'



Anyway, for anyone that wishes to test this, go to your flops/data folder, and open "RTS_CFG.h' with notepad, and look for the line starting with "float HARD_AI_COST_HANDICAP =" and plug in the adjsuted values I posted above.

it's not perfect, not even nessacarily better than Vanilla (yet) but its getting worth posting the figures for testing.

For those that think merciless is easy........ do this.

float HARD_AI_COST_HANDICAP = 0.01;
float HARD_AI_SPEED_HANDICAP = 0.01;

posted on January 27th, 2011, 6:04 am

Well the AI in FO and stock both broke bad. This thing I find most interesting is that non of you guys have done any leg work to fix that.  Over the last few years alot of work has been done to improve the playability of the AI. It has been proven that the AI does not need to cheat to be tough. The info is out there if anyone cares to look.

posted on January 27th, 2011, 9:44 am

Armada 1 and 2 are quite different from Fleet Operations in terms of gameplay. They are both much easier to be put in a static script due to their simple gameplay.

If you are looking for real competition, try multiplayer

posted on January 28th, 2011, 4:40 am

Well the AI in FO and stock both broke bad. This thing I find most interesting is that non of you guys have done any leg work to fix that.  Over the last few years alot of work has been done to improve the playability of the AI. It has been proven that the AI does not need to cheat to be tough. The info is out there if anyone cares to look.

That info is based on assumptions unfortunately and doesn't stand up to rigorous testing and most importantly doesn't work with FO's setup. The AI for A1 and A2 will need to be replaced in order to meet any 'real goals'. That being said, as I stated earlier, if you watch the AI, you will quickly realize how nicely optimized it is in FO - in terms of pure build list.

posted on January 28th, 2011, 5:19 am

Yea if you play the AI especially brutal you will find it will adapt to your tactics, for instance I was playing an 8 person ffa with all merciless and late in the game when it was just me and a fed when i kept spamming e2s to kill their sovys, phalanx, and e2s so they switched over to mass defiants and akiras. they even left a fleet sitting in the middle of where my ships would be going to repair so they could pick off repairing ships. Pretty clever i think

posted on January 28th, 2011, 5:26 am

Actualy fa11out, the brutal AI isnt smart unless it cheats.

That 'brutal' AI only pays 10% of the cost of a vessel.

If you make the AI pay the same amount you have to also pay, then suddenly that AI becomes utterly worthless.

Optec, I dont care about multilayer.

Could y'all take your discussion elsewere, this thread is abotu fixing the SINGLEplayer game experince.

posted on January 28th, 2011, 5:42 am

But the single player is fine  provides a challenge

posted on January 28th, 2011, 5:43 am

Then what are you doing yakking here Tok'ra? - you should be out there studying up on AI so you can design one to replace the current version 

posted on January 28th, 2011, 5:58 am

Then what are you doing yakking here Tok'ra? - you should be out there studying up on AI so you can design one to replace the current version 

Well, if anyone would bother to read the thread, you'd see that I was attempting to get people to test what I was working on.