Theoretical discussion: 2 UpT

[Louis XXIV]

The flaw with the current AI is it still has problems protecting ranged units. 2 unit per turn would absolutely accomplish this, but it does so at the entire expense of the combat system. It turns it from a fairly flexible tactical situation that depends on positioning and terrain and turns it into a formula where the combination unit stack is the only point.

 

[nokmirt]

The flaw with the current AI is it still has problems protecting ranged units. 2 unit per turn would absolutely accomplish this, but it does so at the entire expense of the combat system. It turns it from a fairly flexible tactical situation that depends on positioning and terrain and turns it into a formula where the combination unit stack is the only point.

I realize that. That is why I hope they fix it in the next patch. I see cannons and catapults running around like so many chickens, it is insane.

I happened to put CiIII in the other day just to see how the AI did things. In about 50 turns they attacked me with 3 seaborne invasions with plenty of units. I had trouble holding them off. I lost two cities I had to get back. See, thats what I want is an AI that comes to kill you, not foot around like some old tot chasing old grandma's shirttails. I want a beast with hell in his eye to fight with. The thing with CiIII is that the AI is an ICS freak, I had to block my territory with workers or any units I could find, so settlers from Denmark and England could not pass and build cities in land I set aside for myself. Other than that the game is fast paced and fun. CiV deserves to be that good. I hope soon it is.

 

[glider1]

And anyway 2UPT, is a one line change in the code to try, it's not a total rewrite and there might be things to learn about it. I wouldn't put a lot of work into it Nokmirt unless you find something that is interesting about how the AI behaves. I totally agree that 1UPT is probably better, but Firaxis have been promising to release the AI code for a long time now and nothing is happening and so we need options. I'm starting to wonder if Nokmirt is right, that most of the AI tactical code is borrowed from Civ4 (not the strategic code which is new), and that this is somewhat embarrassing to reveal. This cannot be entirely true either, because the tactical AI does get into formations of a very limited nature. I had a situation where a naval formation was at sea, dead like a ghost navy. I harrassed them with a single frigate killing one unit at a time. The formation simply moved a few tiles away as if it was a single unit, but did not attack. So I assume that the formation logic is there, but the formation did not have any mission code assigned to it, which left it in it's zombie state.
Cheers

 

[Arilian]

BTW what happens you you attack a field with two units on it? Is it the same as in CIV4 (stronger defends) or always the one who moved to that field sooner/later?

If the first one it shows that firaxis was testing 2/3up.

 

[filli_noctus]

2upt will not, imo, help the AI. In the short term you may see the AI perform better because your playstyle isn't suited to handling that. In the long term you'll figure out optimal unit pairings and use them consistently (presumably one ranged/siege + one melee). The AI will not know about optimal combinations and just mash together any old pairs it happens to have lying around. Without adding awareness of unit combos to the AI 2upt (or indeed anything more than 1upt) is just going to give the player another tool the AI can't use to full effect.

 

[Trias]

Initially to me, the AI as horrible as it is, does perform remarkably better being able to use 2UPT. Until I see otherwise I will use 2UPT rules. I am sick of the AI not being able to fight its way out of a wet paper bag with 1UPT rules. In one game I played the AI had two of my cities surrounded with tons of cannons and rifles, and would never take the cities, though it could with ease. It was, to put it bluntly, rather pathetic, as if McClellans ghost was in charge of their army. So thats what I think about the current 1UPT rules in the game. They are disastrous!

The main effect that you are seeing is that in 2 upt, numerical advantage is more relevant than in 1 UPT. As a consequence, the AI actually is benefitting from its bonuses giving it a huge army. In 1upt this big army is probably doing the AI more harm than good, because its units get completely jammed. The lesson to learn from this, is that the AI production/upkeep bonuses should be dialed back a bit and replaced by a (small) strength bonus. (As has been argued by many members of this forum.)

 

[nokmirt]

The main effect that you are seeing is that in 2 upt, numerical advantage is more relevant than in 1 UPT. As a consequence, the AI actually is benefitting from its bonuses giving it a huge army. In 1upt this big army is probably doing the AI more harm than good, because its units get completely jammed. The lesson to learn from this, is that the AI production/upkeep bonuses should be dialed back a bit and replaced by a (small) strength bonus. (As has been argued by many members of this forum.)

I agree the AI is having major unit congestion problems with 1 UPT. In one of my games I gave the AI units 20 hit points and increased the max city defense to 50. The max city defense did not help the AI take cities any easier, in fact it had trouble taking them period. Even with the normal rules the AI does not seem aggressive enough when it comes to taking cities, at least in the games I have played. As far as increasing unit hit points, it tended to help battles last longer. However as long as the human player is persistent and has equal to or one level higher tech, say rifles vs muskets or pikes, the AI is going to lose eventually, even if it has superiority in numbers. Their armies get bogged down behind the front lines and it does not seem to use its artillery effectively. For instance, I have had rifle units stand in a grassland flanked by two hills each with an emeny cannon, and an enemy musket in the middle. Many times the AI fails to attack with both cannons to reduce my rifles before engaging with its muskets. Many times it comes at you with these insane melee charges without any artillery support and gets slaughtered. The fact is that the AI needs some programming help. Changing xml code or adding an extra unit in the hex is not going to solve the problems with this AI. It is simply stupid and will be until it is programmed to be smarter.

One thing I was thinking of doing is giving units more movement, perhaps that would help the AI. I am not sure because many time the AI does not move or shoot from nearly all units that can do damage to an enemy. I have seen it surround a city, and take pot shots at that city from a cannon or two, every turn. Then like 25-50 turns later it decides to assault and take the city over. The AI in this game is horribly indecisive. It needs to be programmed to come and kick your butt. Like in Civ III, that AI is ruthless and it comes to get you over and over. I do not have proof, but I think CiV never had its own programming for the AI complete. It was not new programming, it was left over from CiIV. CiV should have had a complete unique programming overhaul for the 1 UPT system. I believe that is why the AI is so bad off in this game. They applied what old code they could from 4 and pretty much did a shoddy job piecing it together (probably due to money and time constraints, or both). It is quite obvious this is the case due to how derelict the AI performs on a hex based map. If anyone has any other reasoning why they believe the AI could be so horrible. I'd love to read your thoughts on the matter. This is just my assumption.

 

[Soryn Arkayn]

1UpT is without a doubt the primary reason for the AI's utter ineptitude in warfare.

I think Firaxis' decision to use 1UpT is the single worst blunder they committed in Civ5. They've had to radically change virtually everything else either directly or indirectly because of it. And because they've had to change the game so much the blundered into unknown territory for the Civ franchise, which is the reason why the game has been changed so drastically since it's release -- and it's still pretty bad. Playable, yes -- but playable like playing baseball in an empty field; or for my fellow Canadians, how road hockey is playable. My point is that it's not the way the game is intended to be played.

I wonder how long it will take Firaxis to finally admit that they made a terrible mistake and remove the 1UpT restriction. I've played mods with multiple or unlimited units per tile, and it works fine. It's not perfect, but it's better.

 

[_Calyx]

1UpT is without a doubt the primary reason for the AI's utter ineptitude in warfare.

I think Firaxis' decision to use 1UpT is the single worst blunder they committed in Civ5. They've had to radically change virtually everything else either directly or indirectly because of it. And because they've had to change the game so much the blundered into unknown territory for the Civ franchise, which is the reason why the game has been changed so drastically since it's release -- and it's still pretty bad. Playable, yes -- but playable like playing baseball in an empty field; or for my fellow Canadians, how road hockey is playable. My point is that it's not the way the game is intended to be played.

I wonder how long it will take Firaxis to finally admit that they made a terrible mistake and remove the 1UpT restriction. I've played mods with multiple or unlimited units per tile, and it works fine. It's not perfect, but it's better.

I can't agree with this - with all of the faults in Civ V, even considering the terrible combat decisions made by the AI, the 1 UPT combat system is the games saving grace.

Warfare is just a lot more fun in Civ V than it has been in any other iteration of the franchise, especially compared with the purely economic warfare in Civ 4.

If it weren't for 1 UPT I wouldn't be playing Civ V.

 

[nokmirt]

1UpT is without a doubt the primary reason for the AI's utter ineptitude in warfare.

I think Firaxis' decision to use 1UpT is the single worst blunder they committed in Civ5. They've had to radically change virtually everything else either directly or indirectly because of it. And because they've had to change the game so much the blundered into unknown territory for the Civ franchise, which is the reason why the game has been changed so drastically since it's release -- and it's still pretty bad. Playable, yes -- but playable like playing baseball in an empty field; or for my fellow Canadians, how road hockey is playable. My point is that it's not the way the game is intended to be played.

I wonder how long it will take Firaxis to finally admit that they made a terrible mistake and remove the 1UpT restriction. I've played mods with multiple or unlimited units per tile, and it works fine. It's not perfect, but it's better.

With 2 or more units per hex does the AI escort embarked units? It seems to me they are trying to fix a broken system 1 UPT to begin with. On top of that they never researched the intricacies of hex based warfare. Civ has always been about stacking. Maybe they should limit it, but at least allow it. If they went to 2 UPT then they would have to come up with new rules for combat. These could include ZOC, supply, retreat, and rules for artillery, cavalry, and infantry. To install the system properly the game would have to evolve and become more complicated. This game needs some complication to make it better. Simple to me equals less fun and a giant headache. That statement sums up CiV so far since last year. They have been trying to balance the rules over and over, with mostly unreliable results. They should concentrate getting this game back on track with traditional civ games (Those that have played these games forever know what I am talkng about). Shortly, the game needs more options all around. It needs that fun addictive feel added in like we saw in Civ 3 and 4. Even Civ II, which I have for the PS2, is a better designed game than CiV. Sure not in graphics perhaps, but it still has that addictive feel to it. CiV needs that badly!

 

[CivOasis]

The problem with 1upt isn't 1upt, it's that the AI doesn't treat different unit classes seperately.
Thus, 2(+)upt beats around the bush, in that it doesn't fix that problem.
That's not a 1upt issue, it just gets covered up with MUPT.

 

[glider1]

I am a theoretical supporter of 1UPT and there is no problem with it in theory. Here are some of my very favourite quotes from this thread in the last few days

• Nokmirt: "Until I see otherwise I will use 2UPT rules. I am sick of the AI not being able to fight its way out of a wet paper bag with 1UPT rules!"
• "Nokmirt: "I realize that. That is why I hope they fix it in the next patch. I see cannons and catapults running around like so many chickens, it is insane"
• "Soryn Arkayn: Playable, yes -- but playable like playing baseball in an empty field"

Nice analogies and the chicken/catapult analogy is actually very funny . I actually read the last quote in my mind as "playable yes, but like playing basketball on an empty court" I dunno why. Makes me realize that despite the sad reality that I am spending hours on Civ as a self-admitted fanatical addict, having fun with fluff, embarrassed to admit to myself that essentially I am playing 1UPT with a Civ4 tactical AI where each unit is endlessly thrown at me as a Civ4 stack of one unit until I get bored or tired, with some flimsy formation code included to give me the impression I am playing a new tactical AI, that keeps me forever sucked into playing this game like a pigeon being fed bread in the park , despite all that, people are still able to come up with analogies like those in the list above.

Thank you.

Cheers

 

[CivOasis]

Well, the good news about the above: As long as its the AI over the concept, it can (and, IMO, should) be fixed.
Not that its as bad as everyone says, they made significant improvements (see the Korea patch)

 

[nokmirt]

The problem with 1upt isn't 1upt, it's that the AI doesn't treat different unit classes seperately.
Thus, 2(+)upt beats around the bush, in that it doesn't fix that problem.
That's not a 1upt issue, it just gets covered up with MUPT.

So what should be done to fix it then? Is there a way to do so? Evidently it is quite a problem to patch, because it has been over a year and the AI has not improved all that much. I am beginning to wonder if the AI can be fixed. I realize that it is possible, but concerned that the games developers will never strive to fix it. Really the problem is that the AI has never been prorammed to understand hex based tactical strategy and yes that includes enemy unit identification. It needs to know what type of units it is facing and how to deal with defeating them. Also, it cannot be expected to use units properly without knowing the capabilities of the units it controls. So your right it is no a 1 UPT problem. Like I said it is an always has been a programming issue. It is from shoddy hurried old programming, because of time constraints to package the game quickly so they could begin making money. That is why the game came out half finished. They left a lot out that was supposed to be added to the game. It was mentioned in threads some months back. These points keep being brought back up like a revolving door, with just enough time in between for people to forget they were mentioned, but I have not forgotten.

They should replace the catapults and cannon with a damn chicken! A chicken that just runs around for laughs and fires an egg at you every once in a while.

 

[Camikaze]

Moderator Action: Moved to Ideas & Suggestions.

 

[James T]

For me, 1UPT is the best part of this game. In IV I did not enjoy war nearly as much as I do in V.

My experience seems to be different than a lot of posters, though. I play on all levels, but only win about 1/3 of my Deity games. On Deity, I lose cities to the "hopeless" AI. I have seen AI carpets wander around seemingly aimlessly, but have also seen efficient attacks. I have lost cities in blitzkrieg attacks (1 turn after DOW) and in longer campaigns (3-10 turns).

I would love to see a more strategically sound AI with less bonuses, but I would hate to see 1UPT go.

The one change I would absolutely love to see to 1UPT is allowing pass-through for non-combat characters. My military units can pass through other civs' units, I can't understand why it should be different for non-combatants.

 

[Nefliqus]

With 2 or more units per hex does the AI escort embarked units? It seems to me they are trying to fix a broken system 1 UPT to begin with. On top of that they never researched the intricacies of hex based warfare. Civ has always been about stacking. Maybe they should limit it, but at least allow it. If they went to 2 UPT then they would have to come up with new rules for combat. These could include ZOC, supply, retreat, and rules for artillery, cavalry, and infantry. To install the system properly the game would have to evolve and become more complicated. This game needs some complication to make it better. Simple to me equals less fun and a giant headache. That statement sums up CiV so far since last year. They have been trying to balance the rules over and over, with mostly unreliable results. They should concentrate getting this game back on track with traditional civ games (Those that have played these games forever know what I am talkng about). Shortly, the game needs more options all around. It needs that fun addictive feel added in like we saw in Civ 3 and 4. Even Civ II, which I have for the PS2, is a better designed game than CiV. Sure not in graphics perhaps, but it still has that addictive feel to it. CiV needs that badly!

I agree with you but 1UpT is not only one problem. It is connection of several gameplay elements:

1)hex title system (only 6 possible way in comparison to 8 in square title maps),
2)1UpT rule,
3)low average movement point per units (less roads, river/forest/hill movement stopping rules, only 2 movement points for most of units)
4) small maps (less than 300 titles per player)

... causes reduction of possible path for units movement...

...which result in:

1) iterate units micromanagement path finding
2) damage units withdraw blocking
3) easy blockade of attacking effort
4) problems with AI coding

 

[glider1]

Nicely analysed Nefliqus that is about it. As for 2UPT, I honestly think it has potential if 1UPT never matures properly. Nokmirt asked if it will be possible to code the changes. Absolutely because the basic mechanics are most likely already in the AI code. If Firaxis do not release the code, that will make it a lot more difficult to implement if not impossible.

I think the best idea is to wait until the next patch comes out. Perhaps Firaxis will surprise us with some quality AI tactical code (I doubt it but you never know). After all, ask this question. Let's say that we had only 10,000 civ gamers that would buy a 1UPT AI upgrade DLC that makes reasonable improvements to the tactical AI all round. Not spectacular but solid improvements. Let's say these 10,000 players would pay 50$ for that (I would). This generates $500,000 revenue. Now if you employ a single computer scientist, young but with plenty of talent, for a single year to improve the AI code. How could that not be financially sensible? Even after you take all the overheads of employing an AI coder, combined with the risk that improvements will not work, you still cannot loose. Why? Because if nothing else, the AI coder could at least remove all of the stupid simple logic errors that exist in the code today, and I would still pay 50$ for it and the company would still make some profit or at least break even. And if you think that that a single computer scientist working on their own could not improve the AI code, think again. Look what Jdog5000 did basically on his own to the Better BTS AI and he did it for free, for nothing but the love of it and to give some joy to us civfanatics. Jdog5000 is highly educated at Phd level and is now working in the robotics industry and actually said that the experience he got from working on the Better BTS AI is actually complementary to his career in robotics. Would people like him knock back a one year contract to have on your job resume that you were the primary AI developer for the biggest strategic turn based computer game on the planet?

Whatever. If after the next patch the AI code is still not improved, we then have to wait for Firaxis to release the AI code. If they do the later, I honestly think that 2UPT could work very well and even 1UPT could probably be somewhat improved by volunteer programmers. It will be interesting to see what Norkmit finds play-testing 2UPT. It resolves most of all of Nefliqus's observations above, and the AI could be quite readily be taught some basic stacking rules. The promotion system would have to change, but after all that is not that hard to do (just very time consuming). The AI would have tight, densly packed armies where it could more easily shuffle it's injured units through to the rear of the army for healing. The AI could be quite readily taught some principles of how to organise a 2UPT army. The points to study with this 2UPT are:
1) Embarking behaviour
2) Stacking behaviour
3) Formation behavior with the rudimentary formation code that appears to be in the Civ5 AI

The other option is that perhaps Firaxis will improve the 1UPT code just sufficiently but not release the AI code. Then we can concentrate on the compromise alternative that I am playing consistently even on an old computer:
1) Inter city spacing 5
2) Huge map (more tiles more space) and with only standard numbers of cities and units, the turn times are still very good and huge maps are finally completely playable on older machines.

We could then:
1) Explore a modified map script that does not create overly complex terrain for the AI but that still is fun. Perfect world is amazing, but it's terrain is too complex. The continents+ map pack is great but it is not as interesting as Perfect World......
2) Explore increasing unit movements and hit points.
3) Explore adding combat flavours to particular units in the Lua code. For example the highly talented Thalassicus over at the VEM project, has already noticed that it is possible to add combat flavours to units in the Lua code, and he was thinking about adding a unique combat flavour to seige, and the internal AI code could still work with it. The disadvantage is probably turn times as well as other unforseen issues because we cannot get access to the internal code, but it still is an option if Firaxis keep sitting on their AI code like a hen on a cold egg.

Cheers

 

[Gatsby]

The 1upt-Stack of Doom dichotomy is a good example of how a virtue is not the opposite of one vice but the midpoint between two vices. The 2-upt idea is a good move back toward an optimal midpoint between 1upt and Stack of Doom, but I think the game could probably afford to have a few more upt. The Call to Power games had good upt limits imo, especially because players had strategic options for using combined arms.

 

[nokmirt]

What I am going to try to do to remove unit congestion in the game is add a movement point to Medieval land units on up.

Ancient/Classical units will have original movement of two, four for horse, five for companion cavalry etc. all those early units stay the same. Settlers and workers will remain at two respectively.

Medieval era on up all land units get +1 movement.

The other thing I did was make the ironclad ocean going with 40 combat strength and 22 ranged combat.

Default Naval Unit AIs for Frigate, Destroyer, and Ironclad have been changed from UNITAI_RESERVE_SEA to UNITAI_ESCORT_SEA. Perhaps this will help the AI escort embarked units. I do not think this is going to work though. The unitai_escort_sea means the escort stacks with the transport as in CiIV, in CiV I do not believe the AI realizes to escort means to guard the embarked unit in the adjacent hex. The AI has not been coded to do so. I am not sure yet, but the only way for the AI to escort embarked units is through 2 UPT rules, meaning it could then stack an escort warship with an embarked unit. I will try this in 1 UPT first and see what happens.

I am going to try these changes in a prince level test game to see what kind of results I get.

So for this first game. I am using 1 UPT rules to gather some data and screenshots about how the AI uses the changes I've made.

My 2nd game which I have saved uses 2 UPT rules, I will finish that as soon as this first game is over. This way I can compare the results from both 1 UPT and 2 UPT.

So far in the first game with 1 UPT rules. The AI hit me fairly hard in an early war and almost took my northern city of New Orleans. I made some xml adjustments to the tactical AI for capturing cities, among others to try to get it to be more aggressive when it attacks a city and fights in general. So far it has bombarded my city with its two slingers every turn, and even assaulted the city with its only available warrior. My city was redlined. I was able to make some troops in time to push the Incas back, but it was a hard fight, and this my friends on Prince level (which I am using for the tests). I am happy from what I have seen thus far from the AI. Hopefully this continues.