Theoretical discussion: 2 UpT

[WimpyTheWarrior]

I would love UPT to be per Civ and not per world, and to increase as certain military techs are researched. Start at 1UPT, +1 with Military Science, +1 with radio (or electronics or computers).

You learn the ability to fight with more complex tactics.

 

[glider1]

Yeah I guess you could do variable XUPT in the game. I think we have to remember that 2UPT is kind of an insurance policy in case the 1UPT tactical AI is never fixed properly. With 2UPT The basic thing is that there has to be incentive to put ranged at the rear of the formation and not at the front of the formation. Two ways I can think of doing that are:

1) Probability of the weaker unit in the stack being hit is greatest at one tile range attack, and drops off until at three tile range the probability is very low. It may have an exploit where you build much higher number of ranged units to compensate the probability of being hit, but perhaps something could be done about that.

2) Reintroduce collateral damage. At one tile range attack, the collateral damage to both units in the stack is greatest, with c-damage trailing off from there with distance. This could work because collateral damage has more relative effect on the weaker unit in the stack than the stronger. I think that is not a bad idea (ok so where is the flaw!?)

Ironically, these ideas need access to the AI code to implement. Essentially 1UPT is the most difficult option to code the AI with. I think it worked for Civ5 because what happens with 1UPT is that if you throw enough units at a player, it becomes like a puzzle problem. The AI is incompetent, but it doesn't really matter because the player is still having fun trying to solve the optimum puzzle to beat the AI army in each particular situation. So 1UPT tactics with a weak AI is essentially a one side puzzle problem, a bit like a mini-game light weight puzzle for humans to solve, but because the AI is almost incompetent, the puzzles have nothing really to do with warfare tactics in terms of cause and effect in relationship between the two opposing armies.

The assumption I am making is whether it might be YEARS before we ever see an AI that makes 1UPT a bit more of a contest between two opposing armies which introduces more complex and interrelated cause and effect tactics.....

2UPT is an insurance policy in case the 1UPT computer science doesn't happen for a long time. Modders could get into the core and change the mechanics to make it work. 2UPT would have elements of both Civ4 and Civ5 in it, because 2UPT does force armies to spread laterally into the terrain. 2UPT is more complex to code the AI than Civ4, but not as complex as 1UPT. This is because in 2UPT, damaged units can shuffle to the rear of the army under protection and without other units having to move out of the way. So 2UPT is probably not as good as pure 1UPT all round, but 2UPT compensates the human for this with more prospect of a programmable better AI to face up to.

It just needs some lateral thinking to make it work, but more importantly it needs Firaxis to continue the tradition of releasing the AI code. There is a possibility that what will happen in Civ6, is that we see a slightly better 1UPT AI on maps with much higher tile resolutions. Unfortunately that will be years away and requires patience.

Cheers

 

[krc]

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.

My 2nd game which I have saved uses 2 UPT rules.

(I edited the quote to just list the changes.) A general comment. if you tweak too many things at once, it will be almost impossible to figure out which ones are responsible for changes in the outcome.... I'd be interested in hearing (with more test games) -- and possibly trying -- the change to 2UPT alone before mixing in a variety of other rule changes.

 

[nokmirt]

(I edited the quote to just list the changes.) A general comment. if you tweak too many things at once, it will be almost impossible to figure out which ones are responsible for changes in the outcome.... I'd be interested in hearing (with more test games) -- and possibly trying -- the change to 2UPT alone before mixing in a variety of other rule changes.

I am going to check a few things during the game. The ironclad I have had that way for awhile. Another thing to look into is AI naval combat. The naval tests are not going to occur all that often during the game. My main concern is 1 UPT. There is a mod for improving it. I was thinking of trying it for a game.

In the meantime, I am going to go back and play a bit of civ 4. Hopefully a new patch for 5 comes out soon. I'll keep my eyes open. Hopefully it gets some big changes along with a much needed expansion. Yesterday when I went back to looking at civ 4. I was like man there is so much more to love here. It is so much more in depth it is like night and day. I believe by playing CiV for so long its actually ******** my mind, it is so dumbed down!

 

[nokmirt]

CiV keeps crashing. It can't take it captain! I have messed with my xml a bit too much. Something is not working right. In any case, I will reinstall and try a new game. I was having fun in these wars with the Inca. From what I have seen the Inca seem to know how to use their slingers quite well. The AI at this point is trying to stay on the defensive. It has warriors and slingers vs. my swords, pike, spears, and one archer. They build enough of those slingers to really protect their perimeter. I had two triremes coming up the coast to add some bombardment against them. These slingers also know how to retreat, but they are running out of hexes to back up into. Space would have been their undoing, too bad I have to start a new game. What fun!

 

[Nefliqus]

dreams, dreams...
What I want it is just small function AddMinorCiv()... nothing more.

I think that it is easer to add HEXs to civ4 than make better AI in civ5. I don't understand why moders prefer to mode civ5 without SDK than civ4 with tons of ready staff.

 

[glider1]

Hi Nefliqus
Yeah I come from a Civ4 tradition myself. I see your point about modding. Basically us modders have either left civ altogether while waiting for Firaxis to release the code, or we are still sitting around waiting on a hope....Civ5 has some important features. It is multiprocessor enabled, it has AI code with city specialization logic and strategic aims and I do like the maps. The tactical AI is poor, but even if modders or Firaxis could fix the 50-100 small logic errors in the code right now, the tactical AI would improve quite a bit.

In theory 1UPT is very nice. If Firaxis do not follow up on the release of the code, Modders will return to Civ4 in smaller numbers. Keep an eye out for modders of Civ4 returning to that project it is always a possibility. The very best modders out there are people like Afforess. He already included some modifications to reduce stack sizes to a predefinable limit in Civ4 which could be turned into a genuine 2UPT or even 1UPT concept for Civ4. My mod RevolutionDCM is basically ready to go for further work on a solid ranged bombard/archer AI. The difficulty is that Civ4 only runs on one core! Basically it's a lot of work to mod! The hope is that the Civ5 code that Firaxis might release will be top quality code that is easy to modify to make better. In the mean time, many modders of civ have gone elsewhere to work on other projects on the web. This includes myself.

That said, I am keeping an eye out on Civ5. Perhaps the next patch makes significant improvements! (fingers crossed). It might be possible to improve the AI sufficiently for interesting games, just with some simple changes. I am playing Civ5 just for fun, but only with modification changes. Without modification, Civ5 is too simple for long term players. Like a few people have said already, I think it was a bold decision to implement 1UPT, but it broke a long tradition and set back the game AI quite a bit (I think personally). Civ5 tactical AI was designed around the simple premise that if the AI builds enough units and throws them at the player with a bit of formation code that only partially works, that should be good enough to make it interesting for most people in the short to medium term (I only think this). I'm sure some people inside Firaxis did not agree with this (I think that it was the wrong way to go to reinvent the wheel with Civ5), but I cannot stress how big a project it was to build Civ5 and make it financially successful, and the project managers probably had to make quick decisions about what to do with the tactical AI.

There still are some nice features with Civ5! A great modding community that hopefully are still active or waiting in the wings, and the hope that the code will be released soon. I am deeply addicted to civ after 5 years of play and modding, and it's difficult to break the habit.

Cheers

 

[CivOasis]

There are some large, in depth mods for CivV, both complete and in progress. I don't know a ton about civ IV modding, but my understanding is that lua has more potential than python.

 

[krc]

I am going to check a few things during the game. The ironclad I have had that way for awhile. Another thing to look into is AI naval combat. The naval tests are not going to occur all that often during the game. My main concern is 1 UPT. There is a mod for improving it. I was thinking of trying it for a game.

In the meantime, I am going to go back and play a bit of civ 4. Hopefully a new patch for 5 comes out soon. I'll keep my eyes open. Hopefully it gets some big changes along with a much needed expansion. Yesterday when I went back to looking at civ 4. I was like man there is so much more to love here. It is so much more in depth it is like night and day. I believe by playing CiV for so long its actually ******** my mind, it is so dumbed down!

During the period after Civ5 was announced but before it appeared, I had the idea that faster movement rates in Civ4 would help the AI there. So I basically doubled all of the movement rates in Civ4, with infantry having a base of 2 and cavalry a base of 4. This had several effects. First, Civ4 barbarians became deadly in the early game; most workers required a military escort if they were to work any tile at the border (since barbs could lurk two hexes away in the fog of war and would quickly swoop down and pick off workers). This change slowed down growth in the early game (even at lower difficulty levels) because you simply had to build a bigger military. Promotions that extended the line of sight also became much more useful. In the long run, however, I think it gave humans a bigger advantage over the AI because of the lack of zones of control (which I didn't have the patience to figure out if I could mod back in). Still, it makes for an interesting change to Civ4, similar perhaps to the change that 2UPT makes for Civ5.

 

[krc]

CiV keeps crashing. It can't take it captain! I have messed with my xml a bit too much. Something is not working right. In any case, I will reinstall and try a new game.

Are you using the mod tools to do this, or directly editing the XML that came with the game? If you build a mod, then you don't have to reinstall....

 

[nokmirt]

Are you using the mod tools to do this, or directly editing the XML that came with the game? If you build a mod, then you don't have to reinstall....

I did because my game was crashing. It could be because of some mods I was using as well. I just backed up and copied some main xml files. Then I made a few changes. I am just testing things, if they work out then I could add them into a mod.

 

[krc]

So, I made my own mini-mod that only included the change from 1UPT to 2UPT. I've now played a couple of games with this setting, and I am not convinced that it helps the AI at all. This reaction may, in part, be due to the random civilizations I got for those two games (Greece and Persia), without particularly aggressive neighbors. But I did watch Spain try without success to conduct a war against Arabia, and she still couldn't capture the CS of Sidon in spite of an overwhelming advantage in troops..

Instead, I think the change is much more powerful for a human player, especially one bent on conquest. You can put twice as many troops on the front line, bringing more force to bear on the points that matter. Cities that would have been difficult to take because they are ringed by hills or forests become much easier targets. Having a single flatland hex next to the target city allows you to stack four ranged units to fire through the gap....

I haven't figured out exactly how the AI decides which of two units in a hex gets attacked, so I do not yet know if there is any benefit to stalking a ranged unit with a melee unit. Does anyone have any idea what the mechanism is for deciding which of two units defends?

 

[glider1]

Thanks for doing that Krc. I think you could not find a pattern probably because there isn't one. I think that the AI currently works entirely as a single unit AI with some initial unit formation logic that disperses quickly and becomes ineffective. After that point, it's each unit for themselves I think. Therefore you could possible get all sorts of combinations.

For now while we cannot get access to the code to see for sure. So yes, I agree that 2UPT probably doesn't work.
Cheers

 

[krc]

I agree, based on these two games, that 2UPT does not seem to help the AI.

It does, however, definitely ease some of the congestion that you can run into when trying to move troops around.

I checked that, even though you can stack two workers, they cannot both be actively constructing something on a tile at the same time. In other words, you cannot speed up road construction by using two workers on the same tile.

I also found it startling when a Spanish scout moved into the same hex as my warrior during the early game exploration phase; I'm no longer used to sharing tiles....

 

[krc]

Based on the two previous games I played, I wasn't completely convinced that 2UPT did not help the AI. (Mainly because I had neighbors who seemed uninterested in going to war, so I ended up having to DoW them to get things going.)

So, I tried one more time. Played as Babylon on a tiny pangaea map. Selected Greece, Aztecs, and Mongolia as my neighbors. (Rather a fun setup. You know they are coming for you sooner or later.) Got attacked repeatedly by all of them, and saw absolutely no indication that 2UPT changed anything for the AI. Monty made the usual jaguar rush and came through the bowman gauntlet in single-file. Alex put together two different mixed troop attacks (first with hoplites and archers, second with pikemen, archers, and companion cavalry). Both attacks came across a broad front, with no stacking. Genghis made his attack before getting keshiks, and did nothing special. As part of a later war, he had the usual difficulty in taking the city-state of Venice. Went after it with four keshiks, two trebuchets, and one pikeman. Lost the pikeman because he kept attacking with it before wearing down the city defenses, and took six turns to get another in place to actually capture the city. I fought purely defensive wars up until I could raise trebuchets and longswords, and ended up with a domination victory. The AI has even less of a chance defending against a human attack that exploits 2UPT than he does when using 1UPT.

Now I'm convinced. Of course, combining 2UPT with some of the other tweaks that nokmirt suggested may still improve things....

 

[Nefliqus]

Based on the two previous games I played, I wasn't completely convinced that 2UPT did not help the AI. (Mainly because I had neighbors who seemed uninterested in going to war, so I ended up having to DoW them to get things going.)

So, I tried one more time. Played as Babylon on a tiny pangaea map. Selected Greece, Aztecs, and Mongolia as my neighbors. (Rather a fun setup. You know they are coming for you sooner or later.) Got attacked repeatedly by all of them, and saw absolutely no indication that 2UPT changed anything for the AI. Monty made the usual jaguar rush and came through the bowman gauntlet in single-file. Alex put together two different mixed troop attacks (first with hoplites and archers, second with pikemen, archers, and companion cavalry). Both attacks came across a broad front, with no stacking. Genghis made his attack before getting keshiks, and did nothing special. As part of a later war, he had the usual difficulty in taking the city-state of Venice. Went after it with four keshiks, two trebuchets, and one pikeman. Lost the pikeman because he kept attacking with it before wearing down the city defenses, and took six turns to get another in place to actually capture the city. I fought purely defensive wars up until I could raise trebuchets and longswords, and ended up with a domination victory. The AI has even less of a chance defending against a human attack that exploits 2UPT than he does when using 1UPT.

Now I'm convinced. Of course, combining 2UPT with some of the other tweaks that nokmirt suggested may still improve things....

but 2UPT just improves units throughput, nothing more at the moment. Current AI doesn't have stacking logic for sure! So about what AI are you talking about?

 

[Louis XXIV]

Throughput means makes it easier to move units? It does for humans. Does it make the AI more effective at moving units? The moving restriction in some areas was deliberately designed to be a handicap. Humans will be far more effective at taking advantage of these changed rules than the AI. Granted, that doesn't kill the argument for 2 UPT, but it does remove help to the AI as an argument. I still don't see the point. Outside of convenience for transportation, there is very little to gain from 2upt, but you lose all the benefits from 1upt. It strikes me as the worst of both worlds.

 

[binhthuy71]

Great thread! If PLOT_UNIT_LIMIT could be raised for non-military units it would obviate the annoyance of having Worker moves or actions blocked by another Worker or a GP. If I could edit that ability in I'd enjoy the game a bit more.

 

[nokmirt]

Throughput means makes it easier to move units? It does for humans. Does it make the AI more effective at moving units? The moving restriction in some areas was deliberately designed to be a handicap. Humans will be far more effective at taking advantage of these changed rules than the AI. Granted, that doesn't kill the argument for 2 UPT, but it does remove help to the AI as an argument. I still don't see the point. Outside of convenience for transportation, there is very little to gain from 2upt, but you lose all the benefits from 1upt. It strikes me as the worst of both worlds.

So maybe 1 UPT is the best way to go. To tell you the truth Louis I have been thinking that because of your points about 2 UPT, but other than that I realize 2K and Firaxis will not adopt a 2 UPT system. (At least I feel they are going to stick to their guns about their 1 UPT philosophy.) Has anyone tried the improved 1 UPT mod? The main thing is that it does increase unit movement. It also increases city range, which I play at 4 anyway, but the mod has it set at 3. (Meaning a city every 4 hexes, no less) The mod retards road and railroad bonuses for movemnet to balance out the extra movement given to units. To the point, what I am saying is that this could be a great step in the right direction. The second I believe is AI coding.

Great thread! If PLOT_UNIT_LIMIT could be raised for non-military units it would obviate the annoyance of having Worker moves or actions blocked by another Worker or a GP. If I could edit that ability in I'd enjoy the game a bit more.

This could be coded into the game. So do you feel 2 UPT helps a bit or not? Other than the worker issue?

 

[Louis XXIV]

I think, once the DLL is released, people can work on the AI and pool resources to improve it. After all, that's what really helped Civ4 - the Better AI mod got released and Firaxis essentially adopted it.