Experiment to see how hard game is...

[Didz]

In my game you can start re-training units after one turn in a newly captured city so that isn't really an issue.

Mercenaries obviously help in raising powerful armies quickly but they are relatively expensive anyway and would be harder to recruit if sacking did not boost your treasury funds. They also form a pretty fundemental part of the game and so I whilst I think the whole system for how they work needs to be reviewed I would not want them removed completely from the game just to prevent an unrelated player expliot.

[Kobal2fr]

In my game you can start re-training units after one turn in a newly captured city so that isn't really an issue.

That all depends on how you capture it, the replenishment rate of the pools, and how the folks you captured it from left it. If you occupy, the existing pools are left untouched, sacking cuts them by 1/3rd and extermination empties them. If the AI has recruited everything in sight before you arrived, then even occupy will leave you with empty pools.

Meaning that if you lengthen the time needed for one new "unit" to appear in the pool, and/or reduce the maximum pool, and couple that with increased public order problems, then even a sacked city can become impossible to hold if you are to move your whole army away next turn, as witnessed in LTC.

[Didz]

Even so, all that you are doing is putting speed bumps under the path of the blitz, you are not actually removing the motivation for using this expliot.

The removal of the sacking bonus and introduction of attrition and maintenance penalties would not only do this but bring the game play more into line with the real problems faced by real commanders when keeping medieval armies in the field. Likewise, operational goals would need to be much more focussed and limited if the treasury was not topped up instantly with the spoils of war and if armies suffered heavily on prolonged campaigns.

[Ulstan]

I like the troop selection and such for the crusades, but I feel overall they are too reliable and easy for the player to utilize.

The troops should definitely have to be paid for, and I believe they should also suffer attrition every turn until they reach their destination.

[Kruniac]

Blah blah-blah blah-blah blah blah. Geeze guys. So many long winded posts "debating" over nothing in particular. The AI sucks, the game is too easy for MOST people, and playing a "happy merchant game" just doesnt make up for the CRAP of the AI.

The selling point of this game was supposed to be the 3d battles. Real strategic goodness. Yeah, great. The AI FAILS to battle effectively, and that causes the game to suck right there.

Add in the fact that compared to games such as Europa Universalis III, the Strategic Map area of the game is the most empty, lifeless, and generally hindering aspect of life for the next five thousand years, and you have a game that does one thing very well...


Fails.

An improved AI, a working and complex diplomacy system, a complete overhaul of the trade system (yeah my merchant is going to walk over to Yorkshire, and stand around some grapes or something. Try having merchants compete in a MARKETPLACE, not in the middle of nowhere), and a general change to the "feel" of the game is the only thing that will really fix it.

I keep installing the bugger for two reasons. One - I like the graphics, especially of Militia units and archery units. Two - I bought the game instead of warez-ing it, so I have the CDs for when I get a craving. Trust me - if this game were a download, the peice of crap would have been deleted like Space Force Rogue Universe (THAT is a horrible game. UGH)

Anyways... Add more DEPTH to the game. Add more TRADE Depth. Add more PROVINCIAL Depth. (Holy Roman Empire anyone? What, its a gigantic nation, spamming armies? No, its a political body of many nations, whom sometimes slapped each other around, and generally were very interesting.). Finally, add more COMBAT Depth.

If im recieving a charge from the enemies calvary line (which could very well be a head on charge - nothing really wrong with that, just a bit suicidal.), the calvary should NOT turn around and run off after receiving a few volleys of arrow fire. Note I am not speaking about morale - im talking about literally being given a double-move(run) order to run in the opposite direction...

Oh it gets better.

Guess what happens next?

...The calvary comes back towards my archers, full speed. After taking another volley, they turn around again(!) and head back to their own front lines. This continues until the enemy is massacred and routing.

Dont even get me started on seiges.

This lack of intelligence in the combat department is enough to bury this game forever. It isnt the ONLY thing wrong, but if you are trying to put out a game whose entire point is the "awesome 3d strategic battles!", then at least make it worth playing. Make it a challenge.

Hey, you want a good example of how to stop people from "bum rushing" the AI? MAKE A BETTER AI. No excuses, no "Well use house rules"... JUST FIX THE GAME.

Yeah, end rant.

[SoxSexSax]

Soxsexsax,

Something you may wish to do as a mental exercise is to write out a script that does what you say. Once you do that, you can then give a credible example of what should be changed and we can discuss the merits of it.

Having written scripts for programs in the past, I've learned that some times what seems "Obvious" isn't so easy when you have to write out the steps you take to make a comparison / plan. I DO think that if some people give example scripts CA may be able to use a collected brain storming session from the players as an example of what they want to face.

I'm a computer programmer by profession and have been for years. While I have better things to do with my valuable free time than write a whole bunch of conceptual scripts, I assure you that I consider it to be relatively trivial to do the 5 items I said:

1) 1 is utterly trivial, no need to explain this, simple number comparisons

2) Trivial, the autocalc knows roughly how powerful a stack is, the AI should have a minimum power value before using a stack to siege. I forsee no difficulty here

3) Again, trivial. If City.NumberOfDefenders < 8 Then IncreaseDefensePriority(City). How's that for a conceptual script?

4) I fail to see any difficulty with limiting fleet building until a militray rating sufficient to its neighbour count is acheived. Trivial.

5) This one is less trivial in terms of work but I would only envision fairly straightforward coding. Might be time-consuming but not hard.

Every suggestion I made is straightforward from a coding point of view, and hence difficulty of making changes is just not a valid argument against them doing it.

[John_Longarrow]

SoxSexSax

1) Please define hopeless then. Its easy to write "If situation = hopeless", but what would the the actual comparison? Would it be hopeless to have 3 provinces lets, 6 stacks, and be faced by an army of 20 stacks from a 20 province empire? Gotta add in what else is going on on the map. Some times yes its hopeless. Other times, it would cost the human 15 provinces if they tried to go after those 3.

Plus, if the person who specs it out isn't very very good at taking into account what the AI / player can do then you wind up with people still complaining that its broken.

2) Would require your script to be dependant on a known bug. Autocalc doesn't work correctly during siege fights. It compares the fight as an open field battle, not a siege battle. Any script working based on this would give exactly the same problem as now if the human waits for the attack or sallies.

3) Your script is almost as good as the current one. The problem is what is required for a garrison is dependant on local enemy forces and what can be used against the target. That suggestion would result in a worse situation for the AI than the current one. Just as useless while being more expensive.

4) You are not including a strategic picture. As Sicily I'm not sure I need to have an army as large as the HRE before I build a fleet. That would result in the game being a bit easier as some of the smaller coastal nations wouldn't have the fleet to take rebel islands.

5) Correct, there should be a script that identifies what needs to be put where, but the logic isn't trivial as it would require a lot of input based on what is around and what the factions goals are.

From a coding stand point none of these are that hard to do, conceptually. What we need sample scipts for is to develop a specification that nails down the boundary conditions. How you play and how others play are not the same. As such what you would consider a "Trivial" task to make it more of a challenge for you would be seen by others as a major bug that makes the game too easy for them.

I'd suggest you put on your project manager's hat or client liason hat rather than your coders hat. What we really need to do is figure out first and for most how to identify different play styles that require different AI reactions. That would be the hard part.

[Marius Dynamite]

I agree with Kroniac 100%.

Doesn't CA have programmers and game designers who have been making TW games for a while now? Why do we need to give example scripts? The entire point of a programmer is to change english (or any other language) into programming language.

M2TW AI is just bad work. Graphically and in concept the game is brilliant, AI is poor in all fields.

What annoys me most is that Mods are turning out to have better AI than the original.

[Derventio]

Well I enjoyed the generally intelligent discussion here, but it left me totally depressed.
You see I am a pretty intelligent man (IQ 149) I have an interest and understanding of military tactics. BUT I CANT WIN THIS GAME!
I'm great at CIV and I beat Shogun but stuff since I suck at.

Ah well the siege of Dublin awaits ---again.

[John_Longarrow]

Marius Dynamite

I don't think its the programmers who are making the AI do dumb things. I think its the person who's supposed to be generating the spec for how the AI handles conditions that is doing it. I drop it dead in the lap of who ever is supposed to be in charge of the AI team / development. That's the person who has to be shown "Here is what we wan't, not what you are giving us".

Telling someone "It sucks, fix it" gets interpited as "I'm whining and I've got no clue what I want". That is from sitting on the other side of the table and asking "OK, so show me the problem" and having the other person not be able to. If you can't tell someone how it is broken, give examples of how it is broken, and show what you are looking for, they will NEVER be able to produce what you are asking for.

[SoxSexSax]

SoxSexSax

1) Please define hopeless then. Its easy to write "If situation = hopeless", but what would the the actual comparison? Would it be hopeless to have 3 provinces lets, 6 stacks, and be faced by an army of 20 stacks from a 20 province empire? Gotta add in what else is going on on the map. Some times yes its hopeless. Other times, it would cost the human 15 provinces if they tried to go after those 3.

(Code is abstract)

If (CitiesLostInWar > 0 and CitiesTakenInWar < CitiesLostInWar) _
And (MyMilitary < TheirMilitary / 2) Then AcceptCeasefire.

If I have lost more cities and my military is less than half theirs, accept a ceasefire. Acceptable?

2) Would require your script to be dependant on a known bug. Autocalc doesn't work correctly during siege fights. It compares the fight as an open field battle, not a siege battle. Any script working based on this would give exactly the same problem as now if the human waits for the attack or sallies.

This is just plain pedantic. When PLANNING, you ASSUME everything else is bug free, otherwise what's the point? ASSUMING the bug was fixed, my original method stands. Even if it wasn't, divising a noddy but workable strength value from a stack would be trivial.

3) Your script is almost as good as the current one. The problem is what is required for a garrison is dependant on local enemy forces and what can be used against the target. That suggestion would result in a worse situation for the AI than the current one. Just as useless while being more expensive.

No, my script is vastly BETTER than the original, as the original will not prioritise defence for frontline cities for multiple turns even when they have less than 4 units, which is just wrong. Mine might be wrong in certain cases too, granted, but it is better to have and not need than need or not have. And as for being more expensive, some of my other changes would recoup the money.

4) You are not including a strategic picture. As Sicily I'm not sure I need to have an army as large as the HRE before I build a fleet. That would result in the game being a bit easier as some of the smaller coastal nations wouldn't have the fleet to take rebel islands.

Thinking primarily about the strategic side of things gets AI nations killed. Furthermore, my formula would take all land border sharing nation's militaries into account, meaning sicily would need less defence than the HRE before building boats...but still MORE THAN NOW!

5) Correct, there should be a script that identifies what needs to be put where, but the logic isn't trivial as it would require a lot of input based on what is around and what the factions goals are.

No, the logic is (relatively) trivial. Seriously, all it would need are some hardcoded values per faction (probably percentage based) to determine how many castles of each type to devote to each unit class. Nations that rely heavily on horse archers would lean towards more stable based castles, archery heavy nations would have more bowyer based castles and all round nations would have evenly distributed types. Perfection isn't required, but it simply has to spend less per castle/city than it currently does if it wants to keep up.

(BTW, when I say something is trivial, I mean compared to, for instance, writing an 8 way A* pathfinding algorithm, or an efficient alpha blending routine...trivial to a professional coder, basically)

From a coding stand point none of these are that hard to do, conceptually. What we need sample scipts for is to develop a specification that nails down the boundary conditions. How you play and how others play are not the same. As such what you would consider a "Trivial" task to make it more of a challenge for you would be seen by others as a major bug that makes the game too easy for them.

I'm sorry, are you suggesting that my changes (properly implemented) could make the game easier for somebody? Are we playing the same game, where winning a short campaign with any nation on VH/VH is EASILY doable in 50 turns without losing a battle? I refute this point entirely.

I'd suggest you put on your project manager's hat or client liason hat rather than your coders hat. What we really need to do is figure out first and for most how to identify different play styles that require different AI reactions. That would be the hard part.

You're making a relatively easy task much harder than it needs to be, for no good reason. Maybe that's because you're a project manager, not a coder?

(EDIT: fixed about 10 typos and finished a half written sentence)

[alpaca]

Well I enjoyed the generally intelligent discussion here, but it left me totally depressed.
You see I am a pretty intelligent man (IQ 149) I have an interest and understanding of military tactics. BUT I CANT WIN THIS GAME!
I'm great at CIV and I beat Shogun but stuff since I suck at.

Ah well the siege of Dublin awaits ---again.

Are you kidding me? If the game is really hard for you (and you're pretty intelligent as you claim ) I daresay you never thought about it properly. Once you sit down to figure out how it works it's damn easy (especially the battle part)

[Didz]

Are you kidding me? If the game is really hard for you (and you're pretty intelligent as you claim ) I daresay you never thought about it properly. Once you sit down to figure out how it works it's damn easy (especially the battle part)

Actually, when I read Derventio's post my first thought was that this demonstrates the difference between having an understanding of history and tactic's and being a game player.

A gameplayer will work out how to expliot the weakenesses of the game in order to win, an intellient person with an understanding of tactic's will tend to do what historically and logically ought to work and trust the game to reward him for doing the right thing.

A classic case of too much knowledge being a bad thing.

[John_Longarrow]

SoxSexSax

(Code is abstract)

If (CitiesLostInWar > 0 and CitiesTakenInWar < CitiesLostInWar) _
And (MyMilitary < TheirMilitary / 2) Then AcceptCeasefire.

If I have lost more cities and my military is less than half theirs, accept a ceasefire. Acceptable?

As it doesn't take into account either factions other war commitments, I'd have to say no. If you go up to Milan after you've just taken a city or two of theirs, and you offer a cease fire, I'd say they would be nuts to accept it if you've just been excommunicated and your capitol is now the target of a crusade. More would need to go into this decision than just a simple calculation of "Us VS them". That is why I think we, as a community, would need to work out what we think would be a good way to handle these situations. If we can create an actual specification, the coding would be very easy.

This is just plain pedantic. When PLANNING, you ASSUME everything else is bug free, otherwise what's the point? ASSUMING the bug was fixed, my original method stands. Even if it wasn't, divising a noddy but workable strength value from a stack would be trivial.

Other issues also arise. What is the religious situation in the city? What other armies are around? What are the real chances of holding it once you've taken it is the better question. If I'm the player and I have only two units of town guard in the place, the AI would be very dumb to attack with 4 units of town guards if they can't hold it after taking it. We'd have the same disappointment as now, but the AI would be further limited because it would waste more resources in trying to take the city and losing it.

No, my script is vastly BETTER than the original, as the original will not prioritise defence for frontline cities for multiple turns even when they have less than 4 units, which is just wrong. Mine might be wrong in certain cases too, granted, but it is better to have and not need than need or not have. And as for being more expensive, some of my other changes would recoup the money.

For reference to number 4, please see your own post on number 2. You are stating "We can make the AI spend more, but thats OK because we will fix it some where else". The AI shouldn't be forced to increase its expenses when it has a hard enough time handling a player as is. This is why we would need to actually work out how the AI should garrison a city by taking in to account what could be used against it. The Moors should never have a large garrison in Timbuktu unless there is a reason for it. Deciding that they should always have 8 units in it once they've take it is less useful than the current situation. There is no way to say an arbitrary increase is better in all cases than working out a good solution. That sounds like something a customer would come up with, followed by them being pissed off when it doesn't result in their expectations.

Thinking primarily about the strategic side of things gets AI nations killed. Furthermore, my formula would take all land border sharing nation's militaries into account, meaning sicily would need less defence than the HRE before building boats...but still MORE THAN NOW!

Odd, I've never heard someone state that planning strategicly isn't the best way to handle decision making in a strategic game before. It is not something I've ever encountered, but I've seen many people who concentrate on tactical situations lose horribly in strategic games.

So it would be a good thing for the player to have unrestricted access to all AI boarders without the AI being able to counter? So playing as the Danes, I SHOULD be able to attack any other nation that boarders the Atlantic secure in the knowledge that they won't build a fleet that can oppose me for a looooong time? Hmm. I don't think you have thought through the player's options for this.

(BTW, when I say something is trivial, I mean compared to, for instance, writing an 8 way A* pathfinding algorithm, or an efficient alpha blending routine...trivial to a professional coder, basically)

Ah, I see the problem now. You are thinking like a programmer, not a developer. Now I'm not sure how long you've been coding for (any suggestions for keeping proper track of punch cards besides writing sequence numbers on them?) but I've met too many "I can write the code" guys who have no understanding of how to actually meet the clients demands instead of the clients needs.

I may have only been doing coding for about 12 years, but I have learned that giving the client what they need instead of what they ask for often has made me a lot more money in a lot less time. Since I do know that asking a developer (or a coder) to simple "Make it work" or "Fix it" often results in the exact same arguements I read on this board, I'd prefere to tell CA WHAT we think is broken, HOW we think its broken, and WHAT we really want them to change.

I also know that I've gotten very tired of these types of meetings because too few of the individuals involved were versed well enought in the clients business, how businesses need to work, what the employees in the business were actually doing, and what can reasonably be expected from programs that I've gone on to a much more enjoyable career. Now I only have to talk folks currently serving in or recently returned from combat deployments into staying in the military. Much less stressful for me.

[Kadagar_AV]

John_Longarrow, I have developed a game (nothing fancy, peaked at some 40k users), and been an avid strategic games player since, well, the early 90's.

And I must say that much, if not most, of the problems are due to sloppy code, non-existant beta-testing and general bad management from the devs.

I mean, geez, they released the game with a 100% passive AI and game breaking mechanics... *I seriosly doubt any beta tester could have missed the passive AI, hence my claim of non-existing beta testing*

There is, in my view, only one problem, as mentioned before. The game is simply to complex for this particular gaming company to handle.

The more options you give a player, the harder the AI has to struggle to keep up.

Rake the board game of GO as an example... The rules takes about 10 seconds to learn, and it's EXTREMLY straightforward. Yet they have yet to build a computer able to beat a champion.

It was only rather recently a computer could beat a chess champion, and chess is an EXTREMLY dumbed down game (dont get me wrong, of course chess is challenging, I'm just comparing it to medieval 2).



Basicly, this means the AI will have NO chance in medieval, unless it cheats, and a cheating computer isnt very fun.


I guess the conclusion is that the devs wanted more than they could handle, and THAT is just poor management.

I wouldnt mind if they removed the diplomacy alltogether, as an example. Why? Because I'd rather have a game designed around non-existing diplomacy, than as now, a game designed around broken and bugged diplomacy.

Like today.... I offered HRE cease fire, 30k gold and 2 provinces, and they just wouldnt take it. The very next turn (nothing had changed), they ask for peace and give me a province as thanks...

Is it any wonder many experience the diplomacy as rather random?

Solution would hence be to dumb down the game to a level where the computer AI with only a modest bit of cheating can play evenly with the community.





/rant

[Didz]

What seems to be lacking amongst the AI controlled factions is situational awareness. They don't seem to take into account their relative strength to each other or the player controlled faction, or the threat currently posed by other factions to each of their settlements. Their actions are purely reactive to local opportunies and events no matter how dumb that action is given their overall situation.

There is obvioulsy some sort of mechanism in the game assessing the relative power of factions because we actually get a graph showing precisely that information. There is also a constant montoring of factions to determine how close to victory they are because we are warned when a computer controlled faction gets close to victory. So, in theory at least it should be possible to modify AI behaviour to take into account where its faction is in the pecking order relative to the others and to modify its strategy and goals to take this into account.

e.g. Do I need more income?, Do I need more troops?, Do I need more settlements? Do I need more Allies?

The risk is that what results is a Settlers IV type game where the AI faction remains totally passive until it has an overwhelming advantage and then swamps the human player. This produces a rather poor gaming expereince where the AI does not challenge the player at all, unless it can win the game and is then virtually unstoppable.

However, with so many factions it seems unlikely that this sort of result would occur. Most factions would find themselves neighbours to at least one faction which was weaker than they are and one which is stronger and should be coming up with sensible strategies for dealing with all of its neighbours, not as individuals but as a group. The idea being to try and reduce the threat these neighbours pose whilst increasing the threat it poses to each of them.

The following is my attempt to come up with a foundation model for a strategic decision making routine based upon relative power, using four neighbouring factions.

North (6) bordered by East (11) and West (12) total threat (23)
South (5) bordered by East (11) and West (12) total threat (23)
East (11) bordered by North (6) and West (5) total threat (11)
West (12) bordered by North (6) and West (5) total threat (11)

North and South cannot hope to survive an attack by either East or West and so must avoid any direct hostility towards their neighbours. However, were they to form an alliance they would significant reduce the overall threat posed by either East or West by creating a power block of (11).

Both East and West are reasonably confident of victory over either of their immediate neighbours North or South, but only if they can be sure that the other remains neutral and does not intervene. Therefore, they will be seeking some sort of alliance with either North or South which will ensure that they remain passive whilst a war is pursued against the other.

Both West and East need to increase their power base before their factions come into direct border contact, because whoever is the stronger at this point will be in the best position to win. Merely, annexing North or South will not ensure victory, unless you can also prevent the opposition to annexing the other smaller faction, the goal must be to secure one or both of the other factions whilst denying either to your main opponent.

Therefore, if East were to invade the North, then West must react immediately to protect the North either by forming an alliance with it, by moving troops into the North to counter the Eastern invasion unilaterally, or by persuading the South to launch a counter invasion of the East.

An alliance between West and South would produce a power block of (17) to oppose the Easts (11) which means that the East would need to break off hostilities with the North immediately and seek a ceasefire in order to prepare to meet the new threat. The most obvious counter to a West-South threat would be an alliance between East and North producing a power block of (17) to match that of the West-South alliance. The North currently has a power block(17) and a power (11) block to deal with and so has little option but to agree. If it doesn't the most likely result is an invasion by the West which has just succeeded in neutralising the threat from the South by allying with them. The only other option it has is to try and break the alliance between the West and South thus returning the situation to its initial state.

[John_Longarrow]

Kadagar_AV

I agree that the current AI logic isn't up to snuff. I don't think the Dev team has enough resources to handle what the playing community can come up with to throw at them. The ONLY way I can see for the Dev team to work out a better AI would be for us, the players, to specify what is going wrong and how we believe it needs to be fixed. Just as the playing community has has identified problems with the game for CA to fix, we need to identify what the AI is doing wrong, what the AI needs to look at, and how the AI should react.

P.S. What game did you work on? The only strategic game I've had input into (so far) has been Federation An Empire by ADB.

Didz,

North can also side with West/South and work with them to remove East. If North can gain enough from a war with East, it could wind up an equal parter to West. If they are already at war with East, that makes more sense than making an alliance with them.

This would result in one faction (East) being ganged up on by everyone else. It also means that both North and South would want to carve out big chunks of East so that they can (as non-bordering allies) keep West in check.

[Didz]

North can also side with West/South and work with them to remove East. If North can gain enough from a war with East, it could wind up an equal parter to West. If they are already at war with East, that makes more sense than making an alliance with them.

This would result in one faction (East) being ganged up on by everyone else. It also means that both North and South would want to carve out big chunks of East so that they can (as non-bordering allies) keep West in check.

Absolutely right....the key point I was trying to make the about example is that MTW2 (and any other similar wargame) needs to be able to model this sort of 'strategic' planning and manoeuvring BEFORE it even begins to consider where and what to attack.

What I was trying to demonstrate was the need for strategy games to adhere to step 1 on Sun Tzu's doctrines of war.

e.g. Generally the best policy is to attack the enemy's strategy.

If the strategy of the West(12) and South(5) is to form a West/South power block (17) to attack either the North or East. Then defeating that strategy is the easiest way to undermine their plans either by forming a similar power North/East power block (17) or working to break the West/South alliance.

Likewise, if the North's strategy was to join the West/South power block, this would produce an almost unstoppable alliance North/West/South(23) v East(11). In this instance by far the best defence East has is to undermine the strategy of the North/West/South alliance by destroying their alliance, and this should be its first and second priority.

1st: Attack the enemies strategy.
2nd: Disrupt his alliances by diplomacy.

Only if these steps fail should East find itself having to resort to option 3 and attack the enemy armies in the field.

[Jambo]

All TW games are easy. The only way to lose a TW game is if the objective is to lose. You see, TW games are unique in that it's actually harder to lose a game than to win.

I now have my game modded through scripts to give the AI money and city garrisons upon being besieged by the human, and even now I still win 95% of my battles; the ones I lose tend to be from the odd unexpected naval skirmish.

The problem is incurable. One can endeavour to make the campaign as hard as you can through modding, but ultimately you still win every tactical battle you play. In effect all you're doing is giving yourself twice the number of dough-ball AI stacks to fight. Remember the Hojo "Stack of DOom" from Shogun? The only way to nullify a human player is to bore it into submission through sheer weight of AI numbers and the enormity of time it then takes to complete each turn. ;)

I remember when CA removed the AI's ability to outspend its money in Shogun. With the Hojo hoarde removed the challenge disappeared immediately. I actually preferred Shogun with the cheating horde, and so did many others once they realised their complaints of "AI cheating" resulted in an easy game.

On the battlefield there's only ever one winner. The reason for this is the formula for winning on the battlefield is so simple. Shoot missiles as the AI approaches, engage infantry, then once engaged use cavalry to flank and initiate the chain rout. This could be made more difficult to achieve if the AI acted in a similar way. Unfortunately, it doesn't. It likes to charge its cavalry first!

So, make mental note everybody - don't buy a TW game for a challenge. Buy it for immersion (sometimes), graphics and ... um... fun...

Edit: It would seem that CA's way of introducing the challenge for Med II revolves around the greatly reduced timescale to complete the campaign. 225 turns instead of 400+ in Rome.

[John_Longarrow]

Jambo,

You could also mod the game so that all AI factions are Mongol hordes. They seem to know how to shoot first and charge later!

[Jambo]

John,

I think one of the best ways to make the harder might be to use a script that spawns a Generals unit for the AI every 5 turns or so as long as they're underneath a certain amount cf to their city number. The AI using captains to lead big armies does a lot of harm.

[Didz]

So, make mental note everybody - don't buy a TW game for a challenge. Buy it for immersion (sometimes), graphics and ... um... fun...

Whilst I don't deny that what you say is true, do you really think that its inevitable and can't be fixed?

[Jambo]

Whilst I don't deny that what you say is true, do you really think that its inevitable and can't be fixed?

I'd love to say no Didz, but I really believe the current format of TW games makes them impossible to make difficult per s&#233;. They're just too complex with too many variables to enable competent AI to be designed. Maybe under the "Risk" style of campaign map, as seen in Shogun and Med I, the campaign element could have been made more difficult?

However, the Battle AI hasn't improved one iota from the days of Shogun and Med I so it's difficult to believe it's going to improve drastically in subsequent titles as the maps, units, etc, get more complicated in design and function.... What I'm alluding to here with respect to Med II is that the complexity of siege warfare has far surpassed the ability to program competent AI to deal with it.

Where did they go wrong? Well, for starters I believe the more advanced campaign map was a bad move. While there's no doubting that it adds to immersion and realism, I'd argue it's far too complex for the AI. I'd have much preferred a more traditional and simplified boardgame style of map layout; one which has less freedom and the AI can utilise choke points, river crossings, high ground, etc to its full potential. Secondly, the idea of captains... urgh. Why o' why captains? They're the bane of any AI army. All armies should automatically contain a general unit thereby eliminating this particular AI weakness in a oner.

[John_Longarrow]

Jambo,

I do think the AI can be improved to the point that it will challenge any player. Half of the problem is that they don't have any way to push updates for the AI scripts on a regular basis.

If you take a couple hundred players who like to mod and let them regularly update the campaign AI to counter existing player tactics and strategies, you will eventually get an AI that can be a challenge. I'm not sure how long it would take to do so, but I'm pretty confident in the abilities of the players to come up with continual tweaks to make the game better.

One behavior that I personally think needs to be changed is the predictability of the AI. I personally think there should be multiple different AI scripts to reflect different leadership styles. One for an isolationist AI, one for an expansive AI, etc. By having varied responses to a given player generated situation you will find out that the player has a hard time coming up with a "Best solution", especially if two scripts behave radically different to the same player actions.

Think of how a player would react to an AI that follows AskThePizzaGuy's blitzing strategy. That could result in just enough of a stalemate for the player to allow several other AI factions to really get ahead.

[Dracula(Romanian Vlad Tepes)]

Why everyone is complaining about the AI.Just don't play this game if you don't like it,go and play Bomberman and Packman.

[Didz]

@Jambo

Well I would like to say your wrong but I have yet to play any strategy game other than chess and backgammon where the AI managed to provide a challenge without using the standard 'mass unit production' and 'universal knowledge' cheats.

I do think that the AI could be improved though, particularly at the strategic level where it does make a lot of really dumb mistakes. Battlefield AI is much more of a problem as its difficult to isolate the poor how much of the poor performance is down to the strategic AI and how much just to poor tactical handling.

In my current 0.5 years per turn game, I've noticed that the biggest problem the AI has is in army composition. With all its cities and castles fully developed one would have expected it to have a full choice of troops. But instead of using this to generate balanced armies that are engineered to maximise its performance against mine, it has developed a fetish for Venetian and Hungarian armies heavy in pavise crossbowmen, trebuchet's and hussars. The result is usually a turkey shoot for my horse archers against the hussars followed by a mounted raid to wipe out their exposed trebuchets and then a mass charge to overrun their crossbows.

However, Egypt did manage to pull off one surprise victory last night when an army heavy in Royal Marmalukes destroy my fully stacked Turkish army on its way to beseige Gaza. That army still had far too many trebuchets wasting units slots, but the combination of heavy infantry and elite mounted archers was enough to wear down my battleline and eventually cause it to rout. Lots of flanking by the marmalukes too, which is something I've not noticed since RTW.

Certainly, good army composition (e.g. Armies designed to counter the strengths of their opponent) by the strategic AI would make a big difference to battlefield performance as would a clear strategic goal that avoided wandering army syndrome.

[Kadagar_AV]

Why everyone is complaining about the AI.Just don't play this game if you don't like it,go and play Bomberman and Packman.

Very constructive...

If we wouldnt like the game we wouldnt be here, what we are trying to do is make a good game better.

Basicly, the biggest frustration the game has is how good it COULD be if it functioned correctly compared to the current state of the game.

If the games functions all worked correctly this would be an outstanding game of the century, but in the current state you will have to be very forgiving to enjoy the game.

[SadCat]

[QUOTE=Kadagar_AV]John_Longarrow,

"I wouldnt mind if they removed the diplomacy altogether, as an example. Why? Because I'd rather have a game designed around non-existing diplomacy, than as now, a game designed around broken and bugged diplomacy.

Like today.... I offered HRE cease fire, 30k gold and 2 provinces, and they just wouldnt take it. The very next turn (nothing had changed), they ask for peace and give me a province as thanks...

Is it any wonder many experience the diplomacy as rather random? "


Question did something change? It is now their turn. This brings up the question does the AI use old data? Half way through my turn I ask the AI if it wants a cease fire, does it looks at data from the start of its last turn for the answer? Then at start of it's next turn recalculate and want a cease fire? If looking at old data it would not know that I have taken 3 cities and broken its army this turn. It may still think it is on the offensive about to take one of my cities. SadCat

[diotavelli]

Question did something change? It is now their turn. This brings up the question does the AI use old data? Half way through my turn I ask the AI if it wants a cease fire, does it looks at data from the start of its last turn for the answer? Then at start of it's next turn recalculate and want a cease fire? If looking at old data it would not know that I have taken 3 cities and broken its army this turn. It may still think it is on the offensive about to take one of my cities. SadCat

I suspect it has less to do with old data and more to do with the AI being imperfect at providing a "human" perspective on diplomacy.

If you offer to give/sell a province or two to another faction (even at odds very favourable to them), they'll often refuse - with the rationale that they can't trust you (not to take the province(s) back soon after). This also applies when you attempt to initiate a ceasefire and offer provinces.

However, when they initiate ceasefire discussions, the AI does not seem to consider that your proposals may be dishonest and takes them at face value - and is therefore prepared to give up provinces in return for a ceasefire.

This seems crazy and is - but it's also a reflection of how complex the issues the AI designers were dealing with. If you initiate diplomacy and are less than Reliable, the M2TW AI will automatically suspect you of chicanery because you initiated the dialogue and may therefore be planning something tricksy. However, if the AI initiates the discussion, it will give credence to your position so long as it feels you are responding fairly to its requests - and doesn't consider that you're double-crossing them because they made the initial approach.

What's the alternative? That the AI never trusts you? No point having diplomacy at all at that point. That the AI always trusts you? Too easy to exploit and, again, a case for not having any AI.

So CA had to tread a path between these two extremes and it chose this one. It's not ideal but I don't think it's too bad. I regularly have to negotiate in my job and know that concessions are far more palatable if they made seemingly on your terms, not those of your opposite number; the current approach reflects that.

It could be better but it could be much, much worse.

[Didz]

What's the alternative? That the AI never trusts you? No point having diplomacy at all at that point. That the AI always trusts you? Too easy to exploit and, again, a case for not having any AI.

It shouldn't be about trust in the first place in my opinion. The AI should base its decision on whether it is in its best interests to accept or not. If its a good deal for that particular AI faction why should it reject it just because it thinks your a lying toad, and if its a bad deal why would it accept even if it thinks your a saint.

As for not accepting a gift becuase it thinks you intend to take it back again....why the hell should that matter?

Surely, what matters is whether the AI faction feels confident that it can hold onto the gift long enough to make it worthwhile to accept.