XAI 4.0 released!

[xeryx]

XAI will be generic, for everyone. The Bai is taking way more time to get everything fine tuned. So most likely due to times sake the XBAI will be released stand alone. While the in between 4.0 version of the XCAI will be also released, we'll call it 3.8.

[Tyre]

Awesome xeryx, I can't wait.

One more thing. I posted this probably a year ago but this same problem is persisting. If I march 2 armies together the ai general just hangs back and spams "march". Have you been able to narrow this down?

[xeryx]

Yeah Tyre, IIRC it has been toned down.

Hi everyone, i've been lurking on this forum for a while now and after playing Stainless Steel i wanted to experience a better AI
And this seems to be the best AI mod out there, so I was wondering what i would need to download to use this mod. I'm hesitating between Darth Ages & Kingdoms Grand Campaing Mod. Wich one should i get the to have the best experience with XBAi & XCAI ?
I am sorry to bother you with these questions because i know u are very busy, but I am very grateful for the work you guys have done!

thanks everyone

PM's are good for those questions too!

When we release the XBAI, you should be able to use it easily on any mod.

The CAI is more plug and play now, but there will be instructions for you to follow to get the best experience.

[Tudor]

I've just installed Soul Firez's mod and I'm really impressed with it. I'm posting my feedback here rather than on the mod thread, because it's more relevant for the AI than for anything else. I've chosen the mod for a chance to play with something closer to 4.0 rather than for its flavour.

I started a campaign with Sicily using the aggressive CAI version and I'm very pleased with it so far. The CAI seems to have finally been tweaked to the point where it no longer accepts treaties and deals without matching objectives. I have been unable to mindlessly swap map information and alliances the way I used to. I was also pleased to discover that relationships properly fluctuate according to my actions. Fighting alongside an ally improves my standing, while being allied to a neutral's enemy, gradually decreases it. This is particularly relevant as I have managed to get an increase in standing between me and a faction of a different religion (i.e. the Byzantines) just by holding an alliance and supporting them occasionally at sea. Naval invasions also seem to have been fixed as I have seen a number of overseas exploits by AI factions. There's still the odd chance of an irrational naval blockade, but if I have to choose between that and an AI unable to attack by sea, I'll pick the former.

As far as the BAI is concerned, I am also very happy with it. The AI is still no genius, and it will never be, since all you can do about it is fiddle around with thresholds. However, it has become very responsive, and that makes a huge difference. Now, not only do I have to plan a set up to trick the Computer, but I also have to assume that he will use sensible reserves, match troops in rock-paper-scissors fashion, bring in reinforcements and reform as necessary. The bottom line is that although I can still push him off balance, this is no longer enough. Now I have to ensure the availability of sufficient resources to check his reserves and to exploit a breakthrough, once it is achieved.

This was very well illustrated in the battle I have just fought, and refreshingly lost to the Egyptians. Once the Pope called a crusade, I immediately joined, and soon I was besieging Cairo with a 2/3 full stack. I had a decent mix of troops but lacked sufficient missile units. The city garrison was of an almost identical size, but had the typical oriental mix of missiles and mobility around anchors provided by spearmen and axemen. They sallied with a night attack after having brought another stack to back them up as reinforcement outside the city (so the AI has finally learned to coordinate its armies).

I tried to bait the AI by placing a unit of loosely spread peasant archers in my line, backed up by Norman knights, Italian spear militia and dismounted Norman knights ready to countercharge through the ranks as required. I had done so successfully before in this campaign, but I was in for a surprise. Apparently, the AI correctly assessed its superiority in missile troops, as well as the need to wait for the reinforcements to join up, so it just moved into skirmishing range and started to pepper my line to great effect. By now, they had control of the engagement and I was forced to do something. I advanced with a column of cavalry on my right flank. I kept a unit of crusader knights moving forward in order to screen my mounted sergeants which I intended to crash into the enemy archers as soon as I would have moved past them. The AI, however counter attacked with heavy cavalry, and a unit of Saracen spearmen. The key element here is not the reaction which, I believe, is built into the rock - paper - scissors logic of the hardcoded AI, but the fact that you tweaked the BAI to hold proper reserves, which it could deploy according to the said logic. Once the melee started, the Egyptian line advanced in support of the cavalry, and eventually took the bait in the middle, as the AI is still not, and probably will never be capable of understanding a false front. From then on the battle tipped my way. Countercharging through the loose archers started to erode the enemy, and I continuously pushed to exploit my temporary superiority in heavy cavalry on the flanks. However, the availability of AI reserves meant that it took a long while before anything decisive was accomplished. All this while, AI cavalry, although inferior in strength maneuvered, charged and countercharged protecting the desert archers and mamluks continuing to pepper my main battle line from the center rear of their formation. Once their right flank was defeated, I made quick work of their center, which had already been considerably weakened by my dismounted Norman knights and their supporting column. I was finally able to collapse their entire line and exploit the rout, but by the time this happened, I had already lost half my force, and their reinforcements were already in a position to engage.

As it turned out, these reinforcements contained an overwhelming force of heavy elite cavalry (ghulams, bodyguards and Khwarizmi cavalry). There was not enough strength left in my force to deal with this new onslaught. It's also worth mentioning that part of the reinforcements went directly into the city by a different gate in a sensible move to safeguard the main objective, while the heavy cavalry charged as fast as it could crashing into my flank without waiting to reform, or assemble a line, as the vanilla AI so often pointlessly does (if it can even be bothered to use the reinforcements and not freeze them at the edge of the map).

Bottom line for the BAI:
- AI's decision to skirmish turned an intended soft center trap into a conventional double flanking battle with a supporting column assault in the center
- proper use of reserves, reforming and charging delayed the flanking actions, preventing a decisive breakthrough and accentuating the effect of superiour supporting archery for the Egyptians
- attrition caused on the flanks limited my ability to exploit the breakthrough, once achieved
- rapid advance by reinforcing cavalry, to the point of losing contact with their infantry, allowed the AI to charge my army before being able to reform after routing the garrison. The beauty of it is that had I decided to reform earlier, I would have been totally unable to exploit the garrison's rout, and probably wouldn't have been able to defend successfully anyway.

All these, don't make the AI really smart. It still has no imagination, its moves are still shallow and still favour regiment level microing over meaningful deployment of whole sections of its army. However, they do make it challenging, and that in itself is a huge achievement from the team's part. I wouldn't ever contemplate campaigning without logistics against this AI. And with the CAI reluctance to ally, I'm actually a bit scared...

[xeryx]

Sorry guys I am limited to phone access. I will be on this weekend.

[Selerz]

Hi mate, you should contact the Third Age team, because it would be awesome to have a modified XBAI and XCAI in that mod!

[Argent Usher]

Hi mate, you should contact the Third Age team, because it would be awesome to have a modified XBAI and XCAI in that mod!

Well honestly it is a lot of work to redo the unit balancing etc. and on the other side, the WARHAMMER team was a bit faster.
KK is tough enough to implement a final XAI version 4.0 and we have long been in contact if he had questions or we have suggestions for improvements.

We do not say: "Never..." , but certainly not soon.


Cheers A.U.

[emperor77]

This might sound noobish, but is there any way to get XAI for unmodded Medieval 2?

[Hymne der Nacht]

Well honestly it is a lot of work to redo the unit balancing etc. and on the other side, the WARHAMMER team was a bit faster.
KK is tough enough to implement a final XAI version 4.0 and we have long been in contact if he had questions or we have suggestions for improvements.

We do not say: "Never..." , but certainly not soon.


Cheers A.U.

Sounds frightening- Orcs with the evil genius of XAI. Their Trolls will lick my gondorian ass.

[xeryx]

Stay on target!

UPDATE: The siege section is going slow guys, sorry with RL work interfering with my time in the evenings, I only have a little time on the weekends for the next couple of weeks. Once I get transfered back to where I should be working, I will have more time. It is difficult to say the least. We are however making progress. Just hang in there!

[Derfele]

Hey Xeryx and XAI team, it looks like sweetness is coming to town!

I would like to have been around much more, but my wife is very unwell & awaiting a major op.
Thus I am fully working & fully domestic (2 kids) which means all work & no play currently...

However, a big thanks to all of you who have been working hard to develop this most excellent mod

cheers & much thanks - Derf.

[diadok]

i'm not a specialist at all, but for this example :
- the game works, but maybe (quite sure) it delete any value for outflanking between heavy cavalry and pikemen. it does not made the game crashing, because engiine don't NEED it to work.
- when you go back to the .xml after testing: does your modif already there ? or the engiine has re-write "heavy-cavalry" ?

i have change myself .xml only one time: trying to improve skyrmishing, it was hard to find values which can makes best results... but changes were availables, no doubts about it.

XBAI get available results, no doubts that they have not waste their time

but i agree that a lot of mods says "better AI"... whereas it is really hard to see improvements from vanilla... that's why AI need a complete re-write of several files, where values get lot of interactions, needing hours of testing... so the work of a team completely dedicated to AI and only AI, is absolutly necessary, and that's why we NEED XAI :

King Xerix the Matrician, and the Following Rabits Guard

[xeryx]

OK I am home for the weekend, but being a DAD takes priority.

I have moved the previous post about the "xml kernal" to the general discussion thread.

Sorry guys we didn't get this version out by the end of this month, that was the goal. But because of me getting "volunteered" to work in Denver to help save the day. I was totally exhausted most of the time, and I still have one week to go.

I am still working on sieges, and hopefully I will make much more progress, than I did the last weekend I had a chance too. I am a perfectionist, and 4.0 will live up to it's name.

I wont be able to watch the boards that much, because of my workload..so please forgive that.

So currently there is NO release date, but the good news is that only the siege section remains. Please apply for beta testing if you desire.

I also wanted to post a clarification about "Scripting" our views are this, a script is something that is set. It can be an event that happens at a certain time. WE do not believe in set events or telling the "AI" that it MUST do a certain thing at a certain time. Our belief if to give the "AI" options or choices.

For Battle AI: We only want favorable conditions for it to do what it should. We do not believe in predefining scripted events to take place. This leads to predictability and and does not truly enhance the AI. For this reason, We do not script events or battles. Only the normal events are used. Here is why, once the scripted event has taken place, then what is the AI suppose to do? If we do not strive to improve the battle AI, then it will not be able to make sound tactical decisions, regardless of the "scripted" events that take place..it will still be just vanilla AI and you will still be able to easily beat it. That is not to say, there are not times to "script' events but that would be more for the Mod designer to do.

We consider, what we do here "coding" because we are creating conditions for the AI to make better tactical and strategic decisions. We obviously, are not changing the HARDCODE. We change the rules in which the AI makes better decisions. Remember you cannot bend the spoon!

There is alot of mis-information out there, especially when it comes to AI. All of you that play XAI KNOW, that what we are doing here is making a difference. We call upon you and your support, to make a difference. Cuz we love you guys!!

[Glabro]

Ah, finally the siege part is coming together. We all know how important it is for the AI to actually be able to capture your cities, after all!

No fuss from me because of the delay. I'm hooked on Mass Effect 2 until it's through in any case, so it's great to get that out of the way in order to immerse myself in Total War!

I also noticed something nice (which is old news to you guys). Empire will be playable without XAI thanks to the multiplayer campaign feature. One just needs to find opponents reliable enough to keep on playing...

[xeryx]

@ Diadock: Complete balancing of the battle arena is what is most important, so that both the units and BAI work best together. The BAI so far from what we understand cannot be, re-written. But there are parts that better commands and more commands can dramatically improve the choices the AI makes.

CAI: Has been completely re-written along with the Faction Standings files. No other mod can touch that.

I have a full workload, and am trying to catch up..thanks for the patience!

[diadok]

ok, ready, patience is my second name

[TheBlindKing]

ok, ready, patience is my second name

i am not patient as diadok
so pls master tel me when we have our new ai(so this time it would be truly AI a think)

thanks for great works so far...keep going

[xeryx]

The beta test is under way with some new testers. I am working on the XBAI this weekend.

Edit: I have reworked a major and critical section for the Sieges, so now onto some testing and we will see what else needs balanced. So 4.0L is released to the beta test team and team members!

[sellmax]

+ Improved the consequences for a players actions, especially trespassing.

HI,

i would like to see improved consequences FOR A AI actions, especially trespassing.
So the AI does not pass my territory without asking me, telling me or change of Relationship.
So i can be sure if it is passing or entering my territory it means war.

Also,...

I would like to have an AI, which can perform like that,...

1 - If the AI is getting short on money, it should consider to raise taxes and or disband not really needed units, to build financial buildings which will increase money income. And maybe not to start a war with a faction, only maybe take a rebel settlement.

2 - Avoid two front wars only if really strong

3 - Avoid war with Factions except rebels as long as all rebel settlements around that faction are taken.

4 - If faction A is already in war with another faction B, this faction A should not start a second war with another faction C, unless faction A is stronger then both together, or faction B or C are really weak so the faction can be destroyed in 5 or 10 turns.

5 - if in war with a faction AI should try to get alliances with neighboring countries of the enemy, and cut his relationships to any other country to weaken trade income.

6 - basic rule
AI should know which unit good versus another
example swordsmen good against spearmen but the clubmen are not good against spearmen nor swordsmen because they are generally weaker units

counter unit knowledge

resulting in

Ai sends spies to enemy cities to know which counter units it has to send
or
AI spies enemy armies on fields to get to know which counter units it will need

Finally the AI should

change their relationship constantly based on its needs

for example

AI is in an economic crisis it decides to change all relationships except rebels to be neutral and good

But if AI is strong good money income strong army then it wants to expand

But if it is weak say money income is good but no army, or if it is in development
then it considers not to make war and tries to get friends trade rights map information
and so on

There should be more different settings for weak AI and so on

That would be awesome if you can implant that

For any more questions or detail
because its just a rough idea haven't gone it to much detail

please ask

[xeryx]

Thanks sellmax, all of your suggestions are either in XAI, or we are unable to control that particular aspect of the AI. Great Ideas though!

Testing is going well, and we are tallying up any issues to get worked on. it is a slow process, but we want a minimal amount of issues when released. The great news though, is XAI has NEVER crashed a game!

Had tie to incorporate many new improvements to the sieges.

Lets just put it this way. I can now watch the AI defend and attack in a siege all by itself and do it smartly!. I use 2 attacking armies, one mine and an allies, then I just let them attack, and watch the AI repel the invaders. All without any scripting of the armies! It is pure AI behavior!

I will get you guys some screen shots this weekend!