Battle AI Dev Diary

Nice! I hope we will see good battle AI as well as good campaign AI.

Thanks for the insight Jack. It is appreciated.

Very promising news, looking forward to the video development diary on the Battle AI.

Looks promising. If there was one thing in the total war series that discouraged getting totally hooked on it, it was the AI for me. Get the AI right and it will increase the time I will spent with ETW exponentially.

Originally Posted by Razor

Great news! I do hope Lusted has good knowledge of 18th-19th century battlefield tactics otherwise things still might get weird. But I do have confidence in this tester-AI programmer meetings!

We have a lovely wide selection of books at the office.

Originally Posted by Chow_Total_War

forthcoming AI video development diary means an in_game_video or someone_speaking_about_IA's video?

It's an interview with the battle AI programmer.

Originally Posted by Jack Lusted

We have a lovely wide selection of books at the office.



It's an interview with the battle AI programmer.

aw just an interview?is their at least screens or footage to show us the progress of the game?

Originally Posted by Jack Lusted

We have a lovely wide selection of books at the office.

Great to hear that!

Great blog

A thing I noticed was that I played Total War longer than Lusted and I found that very satisfying

Well anyway cant wait for that diary. Keep the info coming

looks like CA is picking up the pace on advertisement now we're getting nearer to the release.

wooT! thats good for one day lets get some more soon =)

Much thanks for the tidbit of info, Sensei and Jack (and CA and Sega). Very promising stuff.

You do have to forgive me for my "I'll see it myself"-attitude. Everyone promises stellar AI these days and none can deliver on their own over-hyped promises. Once I have taken E:TW for a spin and seen the new AI myself, I'll sing more praise.

Good work Jack

Interesting. Looking forward to the video.

Uh huh. these people might believe you but you said that MTW2s AI would be a challenge even to the best players on VH and it wasn't. it was even worse than RTWs AI and was passive to start with for ages till the patch sort of fixed it. I'll believe the AI is actually better when you stop hyping the graphics and naval battles and show people some proof.

Originally Posted by TrueForlornhope

Uh huh. these people might believe you but you said that MTW2s AI would be a challenge even to the best players on VH and it wasn't. it was even worse than RTWs AI and was passive to start with for ages till the patch sort of fixed it. I'll believe the AI is actually better when you stop hyping the graphics and naval battles and show people some proof.

While the Medieval AI wasn't stellar, it was better compared to Rome. Not ground-breaking, but "decent". Ever since CA became a part of the monolith that is called Sega, I feel that they have more room, more time and more money to spend on their products. The delay of E:TW shows this. Showing us more of the AI-building also is a clue of more "care". So, we can expect an improved AI compared to R:TW and M2:TW.

However, these revelations concerning the AI can backfire. I agree with you concerning the hype. These diaries are creating great expectations concerning the AI and that can be very dangerous. The more you show how you are taking your time with the AI, how you use new techniques and so on, the greater the hype and expectations. It may not be their intention, but once E:TW hits, (almost) everyone will expect an AI that can actually whoop the behinds of veteran TW players. We all know that is impossible and yet we hope.

I also hope to God that I'm wrong and I hope that CA will create an AI so lethal I will be begging for an "easy" mode. But right now, I'm gonna remain sitting on the fence.

Originally Posted by Faenaris

It may not be their intention, but once E:TW hits, (almost) everyone will expect an AI that can actually whoop the behinds of veteran TW players. We all know that is impossible and yet we hope. I also hope to God that I'm wrong that CA will create an AI so lethal I will be begging for an "easy" mode.

If CA make the AI merely competent, then its only too possible. Nobody actually knows how good they are as a TW player because the AI is so weak, and weak through every level of the game. Veteran or not, if the AI managed each faction in a merely sensible way, getting just the basic things right at each level, then most players would find the game difficult to beat.

Think for a moment of how many things the AI gets wrong. Politically its totally incoherent, each faction constantly changing from ally to enemy and back again. The players finds it frustrating to deal with. How must it be for the factions themselves? Their whole strategic posture is effected, with the recruitment and development infrastructure getting different messages every few turns as the AI flips form peace to war, peace to war, and worse, the armies they have scraped together sent backward and forward. Watch a game without fog of war on. The AI armies go round and around in circles most of the time, changing target again and again, getting nowhere and achieving nothing.

Then of course there's the recruitment and development process itself. Imagine, for a moment, what the AI factions would look like if they followed the same pattern as the player, ie they maximise the economy of each settlement, screwing as much money out of them as possible, except of course for one settlement in five or six where they emphasise growth and military development, giving them the best quality troops with the best equipment in the game and loads of money to buy them. Faced with AI factions that recruit/develop like a player, or if the recruitment and development systems can't be manipulated like that, then the players massive organisational advantage would evaporate.

Then of course there's the AI's totally incoherent operational approach, my personal favourite. Watch the operations of the Hordes in RTW-BI. They have no excuse for not being mutually supporting stacks of 20. After all they have no home to defend, and are all heading in the one direction. Yet there they are, in little groups of 4 and 5 units, sometimes fifteen, picked off one by one by the player as they head off in lots of different directions. Imagine for a moment if the AI refused to attack anywhere without having a strong and balanced force that was at least 50% of the size of the opponents overall strength, and then only operated in mutually supporting stacks of 20, all aimed at one target at a time. It's all entirely possible in a properly written AI because it is about how that faction conducts itself in its turn, not about its interaction with others. Cut out the weak, incoherent operations by the AI and 90% of the players advantage over it goes down the toilet.

Then of course there's the battlefield AI. Assuming at least the current level of coordination from the AI armies, what will happen if the AI places infantry and cavalry on the wings, making a flanking move very difficult, maybe impossible? What will happen if their general is simply a morale boosting figure who's all but unkillable, not a buffoon who still gets too close to the action and gets himself slaughtered, giving the player the easy win? And what if the army he's leading isn't a collection of third rate militia units but solid regulars all the way through, with no weak units to pick and kill to create a quick hole? And what if his army waited for reinforcements to arrive before engaging the player, moving back to those reinforcements if need be, rather than each army attacking one at a time, and being beaten one at a time?

There at least a dozen things the AI gets horribly wrong, layers of incompetence that have a cascading effect through the game. If CA provide even partial fixes for all of them, if they take away those big fat mistakes, the player will actually have to play to win. At which point a lot of us might be struggling.

Originally Posted by Canterbury

<snipped to avoid massive quote>

I can only agree completely with you, Canterbury. If CA "fixes" the flaws on the 4 levels you mentioned (and the dozens others), it will be a giant leap for the gameplay. While veteran players will still beat that AI (because of sheer experience, blitzing or other reasons), it would make the overal game longer, more challenging and more real.

P.S. I made a rather horrible grammar mistake in my previous post. I really hope that CA can deliver a kick-ass AI. My wording in the previous post made it seem I didn't want that.

Originally Posted by Canterbury

If CA make the AI merely competent, then its only too possible. Nobody actually knows how good they are as a TW player because the AI is so weak, and weak through every level of the game. Veteran or not, if the AI managed each faction in a merely sensible way, getting just the basic things right at each level, then most players would find the game difficult to beat.

Think for a moment of how many things the AI gets wrong. Politically its totally incoherent, each faction constantly changing from ally to enemy and back again. The players finds it frustrating to deal with. How must it be for the factions themselves? Their whole strategic posture is effected, with the recruitment and development infrastructure getting different messages every few turns as the AI flips form peace to war, peace to war, and worse, the armies they have scraped together sent backward and forward. Watch a game without fog of war on. The AI armies go round and around in circles most of the time, changing target again and again, getting nowhere and achieving nothing.

Then of course there's the recruitment and development process itself. Imagine, for a moment, what the AI factions would look like if they followed the same pattern as the player, ie they maximise the economy of each settlement, screwing as much money out of them as possible, except of course for one settlement in five or six where they emphasise growth and military development, giving them the best quality troops with the best equipment in the game and loads of money to buy them. Faced with AI factions that recruit/develop like a player, or if the recruitment and development systems can't be manipulated like that, then the players massive organisational advantage would evaporate.

Then of course there's the AI's totally incoherent operational approach, my personal favourite. Watch the operations of the Hordes in RTW-BI. They have no excuse for not being mutually supporting stacks of 20. After all they have no home to defend, and are all heading in the one direction. Yet there they are, in little groups of 4 and 5 units, sometimes fifteen, picked off one by one by the player as they head off in lots of different directions. Imagine for a moment if the AI refused to attack anywhere without having a strong and balanced force that was at least 50% of the size of the opponents overall strength, and then only operated in mutually supporting stacks of 20, all aimed at one target at a time. It's all entirely possible in a properly written AI because it is about how that faction conducts itself in its turn, not about its interaction with others. Cut out the weak, incoherent operations by the AI and 90% of the players advantage over it goes down the toilet.

Then of course there's the battlefield AI. Assuming at least the current level of coordination from the AI armies, what will happen if the AI places infantry and cavalry on the wings, making a flanking move very difficult, maybe impossible? What will happen if their general is simply a morale boosting figure who's all but unkillable, not a buffoon who still gets too close to the action and gets himself slaughtered, giving the player the easy win? And what if the army he's leading isn't a collection of third rate militia units but solid regulars all the way through, with no weak units to pick and kill to create a quick hole? And what if his army waited for reinforcements to arrive before engaging the player, moving back to those reinforcements if need be, rather than each army attacking one at a time, and being beaten one at a time?

There at least a dozen things the AI gets horribly wrong, layers of incompetence that have a cascading effect through the game. If CA provide even partial fixes for all of them, if they take away those big fat mistakes, the player will actually have to play to win. At which point a lot of us might be struggling.

AWSOME FREAKIN POST DUDE!!! Rep man

Thanks Jack, Reiksmarshal and Stradar1. I try and be constructive, even if it does involve pointing out what I think are mistakes.

I read a lot of strategic analysis, documentation on weapons procurement and military history in general and what comes through is that most military disasters are self inflicted, and most successes come purely from doing your own stuff right and the exploitation of your opponents incompetence. This is scarcely an original thought. There is a quote to that effect from Herodotus or Thucidydes or Pericles on the RTW loading screen. I just don't think its appreciated how true it is. Get the AI to stop committing suicide, make the player actually beat it, and the game will be transformed.

It may not always come through in my posts but I really, really want ETW to be a good game. What little time I have for gaming I spend on this forum because I think the TW franchise has the potential to produce a genuinely great game. Nothing else comes close to it in terms of size, character and visual appeal [i'm as much a graphics whore as anyone] It just needs a brain...

One thing that might be possible now that the AI has been integrated and turned abstract, ie no diplomats, just screens, is on the upper levels of difficulty to give diplomacy and political bonuses to the AI nations when they get into a fight with the player, and of course penalties on the lower levels of difficulty. Thus on VH campaign a nation that's attacked by the player will find it a lot easier to make peace with any other nations it's at war with, will find it a lot easier to get allies/trade deals etc, will be immune to war with others while its at war with the player, will have a lot lower levels of internal dissent, making it easier to raise taxes.

I may be wrong but at the moment the difficulty levels in relation AI seems to only to work negatively, in that it stops the player from getting peace deals and flings enemies at him, rather than positively for the AI nations. The latter will force the player to realise that attacking an enemy will result in him fighting a fully focused opponent who might be able to drag in others from the neighbourhood. On the upper levels of difficulty even when the player is attacked by the other nation, or when the player attempts to carry a defensive war into the other nations territory, the other should still have diplomacy bonuses. Again, hopefully, with a new, integrated AI and abstract diplomacy system the player won't feel like he's operating in a vacuum and won't be able to take the whole world on and not feel the consequences.

At any difficulty level if you get too big, too quick somebody bigger, even the rest of the world if necessary, should be out there, all their rivalries put aside, all their internal problems suppressed, ready to knock you down. The good thing is that it would be entirely realistic. Just ask Napoleon, or Frederick, or the British at the time of the American War of Independence. They all ended up at war with everyone, and none of them came out the winners, even though their enemies were neither clever nor quick.