AI Improvements

AI Improvements (Full Version)

All Forums >> [Current Games From Matrix.] >> [Ancients] >> Hannibal: Rome and Carthage in the Second Punic War

Message

Thales99 -> AI Improvements (10/9/2010 10:42:11 PM)
I've bought this game a few days ago and enjoy it a lot so far. The AI puts up a tough fight, but there is always room for improvement. Maybe a thread gathering AI bugs and ideas for improvements is a good idea. Here are the things I have noticed so far: - Often "Roman Siege" cards are used by the AI when storming small cities defended by only a single militia unit. These cards could be saved for more difficult targets (unless the AI has accumulated too many cards?). - I observed a case where Africanus accepted mountain attrition to move from Etruria to Cisalpine Gaul and then stormed Genua (defended by just one militia unit) from there. He should have stormed Genua from Etruria instead. - I played a card forcing the AI to remove two units from northern Italy. It removed two Latin legions and left two Roman cavalry in play. I think the AI should choose the less capable Roman cavalry (1/1 vs. 1+/2) in such cases. - I sometimes observe generals (with inferior troop numbers) exiting from a border city into a region held by Hannibal. This gives Hannibal two chances to easily destroy a force without having to use trick cards (interception during AI's turn and - if the general manages to retreat into camp - again during the player's turn) . This often happens after Hannibal has entered Cisalpine Gaul at the beginning of the game.

mercenarius -> RE: AI Improvements (10/9/2010 11:52:06 PM)
You make some good points. The storming of border cities is something that I am adding to the next patch. We hope to make this available sometime on Monday the 11th. This behavior will be in the AI at the Hard level and probably the Normal level as well. The use of the siege card in the case that you mentioned was wasteful. The AI probably did have a full hand and needed to play a card anyway. But perhaps this could be refined a little in the future. The choosing of losses in the case that you describe doesn't make sense in game terms, but I don't want the AI discriminating between Roman and Allied units. However, there is some room for improvement when deciding between infantry and cavalry. But, what if those cavalry units had been Latin Allied units? I hope that you will want to try the patch when it will be released. Technically, it will be a beta (at first) but the improvements in the AI should interest you. There will also be a few improvements for dealing with sieges of Rome and sally/relief battles in general.

smartspick -> RE: AI Improvements (10/10/2010 1:56:37 AM)
Mercenarious, whats the ETA on the beta patch? the 11th?

mercenarius -> RE: AI Improvements (10/10/2010 4:39:13 AM)
Yes. It will probably be late in the day (Eastern Daylight Time) but it should be the 11th. Of course, sometimes things get crazy on Mondays. But that is what I am aiming at.

Ron -> RE: AI Improvements (10/10/2010 2:06:22 PM)
I have seen the AI use the Roman Seige cards wisely, ie Major cities, as well as on Minor cities. As noted elsewhere, at times AI does not realize the danger to Rome and seems intent on recapturing Gaul or some such instead of concentrating all available forces to Rome. I have been playing Hard the last few games and I have noticed a trend with the AI's desire to continue building/reinforcing the Navy when the strategic reality in Italy has rendered that option wasteful and just plain wrong, ie the Carthagians have near parity in Italy and control Etruria and Umbria. Equally the Carthagian Senate is divorced from real events and its decisions tend to break immersion, ie after several turns of Reinforce Spain/Upgrade Defenses an Army of 12 stuck in Spain does not need further reinforcing nor does an Army of 7-8 in Carthage. Perhaps random I don't know, but during my Normal games I would occasionally get Punic Trick cards, yet in three Hard games thus far I have only gotten one! What are the bonuses for/against on Hard? Great game! Thank you. Edit: I did get another Punic Tricks card at the end of my current game, Turn 16 - perhaps the others ended too early?

Thales99 -> RE: AI Improvements (10/13/2010 8:41:13 PM)
I just managed to finish my first game using the new beta patch (absolutely flawless, no bugs observed). While the AI gave a good fight, I again noticed its tendency to deliberately move understrength and numerically inferior armies into territories held by the player. During the first turns of the game, the AI sent an army of 5 units (vs Hannibal's 12) to Cisalpine Gaul and landed with 5 units (vs 18!) in Spain. The AI successfully evaded combat and retreated into camp, but the disparity was so large that I attacked the camp and destroyed the enemy army with minimal losses in both cases. A handful of these battles (including a failed invasion of Africa) broke the back of the AI, even though it defended well afterwards and held out quite long. IMHO, the already competent AI could be further improved by disallowing this kind of behavior. Instead it should try to attack/recruit/reconquer wherever the player is weak or has left the region (which it already does quite well in many cases). When actively moving towards the player, it should concentrate its forces before to achieve clear numerical superiority. I also think that an invasion of Africa should only be undertaken if the AI is clearly winning or there is a good chance of gaining Numidian support or outright capturing Carthage. Otherwise the player can attrite the AI with his - normally inactive - Numidian allies for free.

mercenarius -> RE: AI Improvements (10/14/2010 5:59:38 AM)
You raise a number of worthy points. Let me just say that I don't want to rush into altering the play balance any further until I have feedback from a number of players. In the meantime, may I ask if you are playing at the Normal or the Hard level? Did you use the Standard First Move? I do find that it's a little more challenging to decline to use the Standard First Move, as long as it doesn't drive you nuts if Hannibal has a bad march to Italy. Forgive me if I sound like a broken record on that one point. What you suggest has merit. I am not dismissing it. And the 5-unit minimum that you often see the AI use is something in particular that could be changed.

NefariousKoel -> RE: AI Improvements (10/14/2010 9:01:39 AM)
quote: ORIGINAL: Thales I also think that an invasion of Africa should only be undertaken if the AI is clearly winning or there is a good chance of gaining Numidian support or outright capturing Carthage. Otherwise the player can attrite the AI with his - normally inactive - Numidian allies for free. I hadn't seen them invade Africa too terribly early unless I moved many of my troops in Zeugitania elsewhere. Even when they did, they definitely worked towards getting Numidia to switch and achieved that a couple times against me. The AI is definitely a bit more reckless in the early game until they've fought Hannibal some. It's meant to be that way. However, if they start attacking at 1 to 3 odds regularly then it'd be good to check on. I'd be a bit scared to tweak it too much, though, as I like how aggressive it can be at the right moments and wouldn't want that to change.

Miket_nz -> RE: AI Improvements (10/14/2010 10:56:16 AM)
I'm not sure whether early in the game whether the 5 unit minimum needs to be changed for the romans (isn't this supposed to be the standard consular army size in the game?). Up until Hannibal came along the romans had been pretty succesful at defeating larger armies of lower quality troops. I also guess there is an issue of abstraction at this level of detail that also comes into play, armies of this era could easily decline battle or move with relative freedom in the same region as another army given the often "voluntrary" nature of meeting for battle or staying in camp. I like that the game models the fact that the cocky romans get outmanouvered in the field and can't decline battle on occasion. Gives a nice feel to the game.

Thales99 -> RE: AI Improvements (10/14/2010 8:10:07 PM)
quote: ORIGINAL: mercenarius In the meantime, may I ask if you are playing at the Normal or the Hard level? Did you use the Standard First Move? So far I've always played on Hard level without Standard First Move (I like the added strategic options and variability). I agree that one has to be careful with changes, and the AI is certainly challenging as it is. But maybe "teaching" the AI to build large enough armies will prevent it from getting overwhelmed and attrited too easily in some cases. Of course strategic necessities and lack of troops will not always allow this.

PJJ -> RE: AI Improvements (10/15/2010 1:18:18 PM)
I always welcome AI improvements, but I also like it when the AI sometimes makes mistakes - this is what happens all the time in real military operations (or in any human activity), and both sides in the Punic Wars certainly made their fair share of mistakes, sometimes really stupid ones. And of course bad generals should be more prone to making silly things that end up getting their troops slaughtered. But I don't want to see Scipio invade Africa with one or two legions when I have several times that number of troops waiting for him in Carthage - that's suicide and completely out of character for a brilliant general. Romans were quite arrogant and usually for a good reason, being in the habit of always winning their wars with their incredible aggressiveness and relentlessness, but Hannibal definitely taught them some humility. Hannibal already has a brilliant and well-balanced AI opponent (I always play on hard and often lose to the AI), so one must be careful with the improvements.

Toby42 -> RE: AI Improvements (10/15/2010 2:38:53 PM)
quote: ORIGINAL: PJJ I always welcome AI improvements, but I also like it when the AI sometimes makes mistakes - this is what happens all the time in real military operations (or in any human activity), and both sides in the Punic Wars certainly made their fair share of mistakes, sometimes really stupid ones. And of course bad generals should be more prone to making silly things that end up getting their troops slaughtered. But I don't want to see Scipio invade Africa with one or two legions when I have several times that number of troops waiting for him in Carthage - that's suicide and completely out of character for a brilliant general. Romans were quite arrogant and usually for a good reason, being in the habit of always winning their wars with their incredible aggressiveness and relentlessness, but Hannibal definitely taught them some humility. Hannibal already has a brilliant and well-balanced AI opponent (I always play on hard and often lose to the AI), so one must be careful with the improvements. That's right. Don't ruin a "Great" game by trying to make everyone happy!!

mercenarius -> RE: AI Improvements (10/15/2010 5:09:36 PM)
First, I forgot to reply to Ron's question about Punic Tricks cards. There are no differences in frequency of cards at any of the levels of difficulty. He simply experienced the luck of the draw. Naturally, a longer game tends to average out more (in all respects) than a shorter game. And I will remind everyone here that the game does have a certain chance each turn of reshuffling the deck to keep players from counting cards. I wanted to say something about the relatively small armies which the AI uses to "shadow" hostile leaders. This is built into the behavior of the AI in place of an explicit system of disincentives to prevent Rome from "hiding" from Hannibal and your other generals. Such a system would just be too complicated. But if the AI were allowed to ignore Hannibal's entirely, build a fleet, garrison Rome, and then send armies anywhere it wanted, the game would just become impossible. So the AI is programmed to usually "harass" your generals when it can't meet them in the field. And it's true that sometimes you will catch them off guard. This feature works reasonably well, I think. The business in Spain that was cited does show a weakness in how the AI usually goes about this. However, I think that what happened there is that the AI thought it could capture a port but then couldn't. Probably because another leader used the last available "Siege" card. That is a mistake that I allow the AI to make. As for the 12 versus 5 in Cisalpine Gaul, that's a battle that Hannibal can win but unless the Roman force was overbalanced with Cavalry, I think that Thales got lucky not to suffer some losses. Finally, I do agree that there could be some fine tuning of force levels on expeditions to Africa (or Spain or Sicily) especially when Rome doesn't control a port there already. I am going to try to improve this a little and then issue a second beta patch next week. After that I hope to close the book on changes to the AI for a while and concentrate on finalizing 1.0.2 and issuing that to all users.

Ron -> RE: AI Improvements (10/17/2010 3:29:34 PM)
I am of the opinion that until the Roman AI has 'learned' about the danger of Hannibal, sending smaller Armies initially is probably better for them in the long run so they still have some Troops left. I am finishing my second Hard game since the Beta patch, optional start, and both games have generally followed similar courses. Hannibal moves to Cisalpine Gaul, gathers Gaulish recruits, defeats the Army from Placentia or whatever else the AI sends. From there Hannibal moves to Etruria, hopefully capturing Genoa or Pisa, so an Army from Spain or Carthage can follow depending on the Senate, and defeats the Roman Army sent against him. After that Italy and Rome will be wide open. Having the Roman AI concentrate its Armies to face Hannibal in Etruria - usually the second big battle - makes for some fun initially but has resulted in an early ending overall. This last game Hannibal was sitting in Etruria with an Army of 12 and one Punic Trick card. First the Romans recaptured Cisalpine Gaul, apart from Placentia which was garrisoned by Mago, with an Army of 6 followed by a raid in Carthage with an Army of 2. The Romans had concentrated an Army of 14 with their worst General(4!) to square off against Hannibal who of course crushed them totally with a Trick. Now by the fourth turn usually, Rome/Italy are wide open and their smallish Armies are scattered, ripe to be destroyed piecemeal. Hannibal can win a war of attrition favourably even without any Punic Tricks. Even playing conseratively and capturing Umbria to gather those recruits, the game usually will end in a few turns. Perhaps I have been playing too much and need to take a break! As to the frequency of getting Punic Trick Cards on Hard, well I must have bad luck then because I very rarely get them! I get swamped by the 'Forced March' cards though :)

vonRocko -> RE: AI Improvements (10/17/2010 4:00:03 PM)
quote: ORIGINAL: Ron As to the frequency of getting Punic Trick Cards on Hard, well I must have bad luck then because I very rarely get them! I get swamped by the 'Forced March' cards though :) This has been my experience also, after many, many games on hard, I rarely get punic trick cards, outside of what I start with. Plenty of forced marches. I'll take Mercenarius at his word, but I must be one unlucky person, it seems more than random draw. When I play lower difficulties, I get plenty of punic tricks. I'm not complaining, just stating my observations.

PJJ -> RE: AI Improvements (10/19/2010 10:52:55 AM)
You guys don't have any Punic Tricks? Well, I have four in my current game (hard difficulty). Too bad the Romans don't want to play with Hannibal. [;)] [image]http://i766.photobucket.com/albums/xx304/tolokkua/Hannibal.jpg[/image]

Ron -> RE: AI Improvements (10/19/2010 12:55:35 PM)
quote: ORIGINAL: PJJ You guys don't have any Punic Tricks? Well, I have four in my current game (hard difficulty). Too bad the Romans don't want to play with Hannibal. [;)] That is because our games are over by Turn 8! :)

NefariousKoel -> RE: AI Improvements (10/19/2010 8:15:58 PM)
quote: ORIGINAL: PJJ You guys don't have any Punic Tricks? Well, I have four in my current game (hard difficulty). Too bad the Romans don't want to play with Hannibal. [;)] You seem to have things in hand on that game for the moment. [:D]

mercenarius -> RE: AI Improvements (10/20/2010 2:16:59 AM)
Well, I have digested what has been posted here in this thread. And I do see the validity of what people are saying. Here is what I intend to do. 1) We plan to post a beta patch (1.0.2 beta 2) on Wednesday (the 20th) or Thursday the 21st. This patch will have fixes for all known errors. I want to get this completely stable so I can make it official as soon as I can. The patch will have a few minor tweaks to the AI, but nothing dramatic. 2) Once the 1.0.2 patch is released officially, I will start work on version 1.1 of the game with an eye towards making the game more challenging at the "Hard" level. My goal is to make the AI smart enough to deal with the kind of strategies that I see some players are throwing at it. But I will also try to keep the general difficulty of the "Hard" level about the same as it is. 3) I don't think that you will see version 1.1 any earlier than about 6 weeks from today. [:(] Now, this forum is for the players and I certainly don't want anyone to feel that they cannot post honest criticism of the AI. I will certainly try to take everyone's views into account. Naturally, it's hard to make everyone happy. I hope to come close. I realize that waiting 6 weeks (or more) to get a more challenging AI will disappoint some players, but I don't think it can be done sooner and still remain a challenge based on "intelligence" and not greater numbers, etc. Of course I will continue to read what people post here on this (and similar) circumstances.

Evil Steve -> RE: AI Improvements (10/20/2010 9:41:52 AM)
the only comment I have to make is with militia in sieges: a route for a militia is essentially a hit, which make succeeding in a storm of a city, perhaps, easier than it should be? I appreciate that this may have been taken from a tried and tested design, but was wondering whether militia could be: 1. not routable 2. or always "inspired" 3. or when routed (ie hard-pressed) become Att 2 / Def 0.5 or some such. I'm also seeing that the Romans seem to be getting a full complement of militia after taking a city, whereas Carthage gets a militia that "grows with time". however, I may be seeing things here :)

Raidhaennor -> RE: AI Improvements (10/20/2010 11:26:40 AM)
I wouldn't mind seeing an increased difficulty at the hard level, though I haven't tried the latest beta patch yet (I just downloaded it). From my little experience, what tends to make the game easier is the fact there is virtually no penalty for leaving provinces such as Spain and Cisalpine Gaul undefended once I have recruited all I could. Even Carthage I tend to leave without units when I can. There is really no downside for doing that because : 1. In the case of Spain and Northern Italy, I would lose nothing if I lost control of these provinces (if anything, losing Cisalpine Gaul might allow me to use "Gallic Aid" a second time). 2. The additionnal units I get in Italy when I do that put added pressure on the romans, and therefore reduce the likelihood of an expedition abroad. It's a win-win for me : and even if the romans were foolish enough to try something, see point 1 (actually, sometimes they are, at least in the case of Cisalpine Gaul. Maybe the AI should be "taught" not to bother retaking a province if doing so causes me no harm whatsoever). Only once have I seen a roman force in Africa (it was on hard level) : it was a small force, easily countered by the force in west Africa which "unlock" in these circumstances. Leaving a province undefended ends up reducing the likelihood that it will be attacked, and I think it would help if we could find a way to mitigate that. Either by having a requirement to keep a certain number of units at all time (with a leader) in Spain and Carthage. Or maybe what could be more fun and a little less artificial is to make the senate decrees play that role : as it is now, I can pretty much ignore the senate recommendation. Some form of sanction if I do that could make things interesting : for example if the senate tells me to reinforce Spain, I have to have a leader and a minimum number of units present (either at all times, or for a number of turns, I don't know) ; and failing that I am sanctionned. I may have less recruits from Carthage, or less command phases, or a combination of both. That would require more coherence from the senate though (see my post in the "fun game" thread), and should only be applied to some decisions like reinforcing Spain, upgrading defenses in Africa, and maybe to Sicily, but in the form of a temporary decree for that theater (or a requirement to chase the romans off the island).

NefariousKoel -> RE: AI Improvements (10/20/2010 11:08:04 PM)
Giving up your core territories that have been deprived of recruits is a strategy, I guess. Though they do start regenerating more available recruits slowly afterwards. At some point, later in a long game, many territories will be exhausted and the ones that get 'refreshed' most can still make a difference once it gets to this point and casualties start to hurt more.

SteveD64 -> RE: AI Improvements (10/21/2010 12:23:08 AM)
Yes, Spain and Northern Italy start regenerating and there's the victory points to consider- New Carthage is very important if you play the long game (like Hannibal, historically, tried to do). This is a very fun game with ALOT of decisions to be made every turn.

Raidhaennor -> RE: AI Improvements (10/21/2010 12:35:35 PM)
My games tend to end relatively early (around turn 12-14), when I take Rome. So I usually don't need the few additional units I could recruit ; I tend to already have enough to fight back anything the romans have left and march to Rome. So the long game isn't really a concern, and the same goes with the victory points. And anyway, apart from the occasional retaking of the minor cities in Cisalpine Gaul, the provinces/cities I leave defenseless remain under my control until the end, I am not giving them up, that's the point. By leaving a province defenseless and using the units I get to put pressure on Rome, I actually make it less likely that these defenseless provinces get attacked. My games have always finished with New Carthage under my control. My point was that it shouldn't be an effective strategy, or that at least it should have consequences. Right now there is no downside to using that strategy. Granted I could use "house rules", and that's certainly what I will try in my next few games. But I don't like doing that, I prefer when the game itself already contains all the rules necessary, which is why I was making these suggestions. Don't get me wrong, what I'm saying here doesn't take anything away from the game. If anything, the fact that the AI favors defending itself effectively, rather than getting sidetracked into a pointless/hopeless expedition overseas is a point in its favor, a testament to how good it is. And the game is indeed a lot of fun, and very addictive too : I often start a game only planning to play a few turn, and I end up playing it to the end anyway. edit : I have now tried the beta patch (1.02), and well... you can disregard what I said. The AI went much more aggressively after Carthage and Spain than in any previous game, from the very first turn. That game ended early, but not because I won. [:D] The difficulty is definitely there, excellent job. [:)]

mercenarius -> RE: AI Improvements (10/22/2010 1:12:19 AM)
Well, as you can see, we didn't post the next version of the 1.0.2 beta patch today. But we expect to do this tomorrow. Now, as for routing defenders during city battles: I have this in the game in part to keep the battle boards consistent. But the likelihood of scoring a rout against defenders is greatly reduced in this type of battle. It's true that you will see city militia get routed fairly often but that is because you usually will have overwhelming force used against minor cities with no additional garrison. The next patch will have some minor play balance tweaks and a couple of bug fixes. After it's final I will start in earnest on version 1.1. And I am sorry about the delay with the 1.0.2 beta patch but you should have it in time for this weekend. Thanks for all of the feedback.

mercenarius -> RE: AI Improvements (10/22/2010 10:19:21 PM)
The 1.0.2 beta 2 patch is out and I wanted to say a couple of things about it. It is a beta, but I do hope to get it finalized very soon. People who are happy with things as they are might as well wait. Any dissatisfied players will probably want to try the beta. I think that I have said all of that before. [:D] The small rules changes that this beta patch makes are not set in stone. I hope to get some feedback from interested players in the next week. Reducing the command bonus to 15 percent is very experimental. That is a change that may be changed itself. [:)] The other play balance tweaks that I have made seemed to work well when I tested them myself. I think that they won't ruin the game for people who liked the current play balance in the game. But they should help the AI adapt if things go especially wrong at the beginning. We'll see how it goes. And I will keep everyone's comments in mind as I finalize this patch and then start thinking about a version 1.1. Thanks for your interest.

Toby42 -> RE: AI Improvements (10/22/2010 11:20:19 PM)
quote: ORIGINAL: mercenarius The 1.0.2 beta 2 patch is out and I wanted to say a couple of things about it. It is a beta, but I do hope to get it finalized very soon. People who are happy with things as they are might as well wait. Any dissatisfied players will probably want to try the beta. I think that I have said all of that before. [:D] The small rules changes that this beta patch makes are not set in stone. I hope to get some feedback from interested players in the next week. Reducing the command bonus to 15 percent is very experimental. That is a change that may be changed itself. [:)] The other play balance tweaks that I have made seemed to work well when I tested them myself. I think that they won't ruin the game for people who liked the current play balance in the game. But they should help the AI adapt if things go especially wrong at the beginning. We'll see how it goes. And I will keep everyone's comments in mind as I finalize this patch and then start thinking about a version 1.1. Thanks for your interest. So, reducing the command bonus makes Hannibal even less potent?

SteveD64 -> RE: AI Improvements (10/23/2010 2:09:02 AM)
From my small sample of games played Hannibal has beaten everyone he's faced without the card play. I've been hoarding the Punic Tricks cards until I face a 7+ Roman leader. At the beginning the card comes out in order to give H. some breathing room but the middle game sees him beating Romans with ease without the card. I mean it hasn't been close, the Romans have been slaughtered. I still haven't won, but that's because the Romans won't fight H. after they've learned about his ability. Luck plays a big part obiviously, which as a boardgame port it should.

mercenarius -> RE: AI Improvements (10/23/2010 2:20:42 AM)
I'll take that as a vote for changing it back. [;)] I wanted to experiment with this patch a little before it goes final. But a couple of the changes may not work out. I wanted to get some feedback and you are providing that. This change (from 20 to 15) and the change in command bonus for camp battles are two changes that could easily be reversed if they don't make sense to the players. Why don't you try a couple of games and tell me how it is for you?

Toby42 -> RE: AI Improvements (10/23/2010 3:08:08 AM)
quote: ORIGINAL: CLEVELAND From my small sample of games played Hannibal has beaten everyone he's faced without the card play. I've been hoarding the Punic Tricks cards until I face a 7+ Roman leader. At the beginning the card comes out in order to give H. some breathing room but the middle game sees him beating Romans with ease without the card. I mean it hasn't been close, the Romans have been slaughtered. I still haven't won, but that's because the Romans won't fight H. after they've learned about his ability. Luck plays a big part obiviously, which as a boardgame port it should. I guess that I'm confused? How does reducing the command bonus improve Hannibal's fighting ability? I would think that it would hurt him in battles....

Page: [1] 2 next > >>

Forum Software © ASPPlayground.NET Advanced Edition 2.4.5 ANSI

2.125