You are talented, to make an AI as good as Wesnoth on your spare time is quite an achievement,
Playing on hexes: Notes from Battle for Wesnoth
- Thread starter EmpireOfCats
- Start date
You are talented, to make an AI as good as Wesnoth on your spare time is quite an achievement,
Ahriman
Tyrant
Fair enough, the Civ warfare AI has always been bad too, and there are obvious things that could be done to fix it (and sine have been done in many Civ4 AI mods), like getting the AI to realize the probabilistic implications of throwing its stack at a city, and deciding not to if it will lead to obvious defeat.
So maybe I have rose-tinted glasses of Civ.
I think that linearly increasing production bonuses to increase army size is relatively effective in a Civ4 stack world, once congestion effects start coming into play, linear increases in AI army size have declining marginal returns in AI power.
So you have to have *much* bigger AI armies to remain competitive when congestion is an issue, and I worry that much bigger AI armies "feels" more wrong to the player in a 1 unit per tile game design than it does in a stacked system.
My biggest problem with the BfW AI is that it generally seems unable to make a good defensive line; it is far too easy as a human player to select a defensive line on the boundary of good terrain, and to have the AI throw itself against it inefficently.
In a scenario where the AI is designed to have huge gold reserves, this can still be lots of fun. But it does mean that the AI needs a big troop advantage to be competitive.
I guess my feeling with BfW is that the strongest point of the game is the elegance of the design.
The AI is good enough to have fun with, but maybe I have unrealistic expectations of how easy it would be to go beyond that; I know nothing about the details of AI coding.
But as far as the game mechanics, I cannot think of a single improvement I would make that would make the game unambiguously better; its a perfect demonstration of an engine where there is nothing left that could be taken away, but there is nothing to add that wouldn't just be adding features for the sake of adding features, rather than in actually making the game more fun or more strategic.
Maniac
Apolyton Sage
Sirp, would you allow someone using the Wesnoth universe to make a Civ5 mod?
Absolutely -- it would be very welcome. Though I would have to say that the Wesnoth universe is one of the least interesting parts of Wesnoth imo -- after all it is largely generic Tolkien/D&D inspired stuff.
-Sirp.
Maniac
Apolyton Sage
Thanks! 
The greatest attraction to Wesnoth are of course the game mechanics. Multiple weapon damage types and the day-night cycle (which could be changed to summer-winter for civ purposes) in particular would add to Civ5 combat. But if I'd base a Civ5 mod on the Wesnoth mechanics, I might as well use the universe as well. Plus 1) I suck at creating even generic universes, so using Wesnoth would make things a lot easier for me. 2) Wesnoth has great music and unit portraits (which can be turned into buttons), again saving a lot of work. Which leads me to wonder: would there be any conflicts between Civ5 being a commercial game, and the music and unit portraits being under a "free" license?
Elaborating a little, I figured the game could start when humans and orc first arrive on the continent. Then you can use the classic "start with one settler and warrior" model. The established elves and dwarves could be city states. Though of course there would also be playable elf and dwarf civs, who represent elves/dwarves leaving their traditional homes.
The greatest attraction to Wesnoth are of course the game mechanics. Multiple weapon damage types and the day-night cycle (which could be changed to summer-winter for civ purposes) in particular would add to Civ5 combat. But if I'd base a Civ5 mod on the Wesnoth mechanics, I might as well use the universe as well. Plus 1) I suck at creating even generic universes, so using Wesnoth would make things a lot easier for me. 2) Wesnoth has great music and unit portraits (which can be turned into buttons), again saving a lot of work. Which leads me to wonder: would there be any conflicts between Civ5 being a commercial game, and the music and unit portraits being under a "free" license?
Elaborating a little, I figured the game could start when humans and orc first arrive on the continent. Then you can use the classic "start with one settler and warrior" model. The established elves and dwarves could be city states. Though of course there would also be playable elf and dwarf civs, who represent elves/dwarves leaving their traditional homes.
Steven Aus
Teach The World To Fish For Life
Something interesting for Wesnoth players is the Experimental branch of Wesnoth 1.8. 
Wesnoth-XP allows you to choose damage calculation settings (accessed by the General tab in Preferences). "Split ratio %" is the percentage, based on defense of unit being attacked, of damage that is guaranteed, and "RNG-Smoothing" is the number of simulated rolls for each attack. I find a ratio of 0% and 3 smoothing works well. Basically, the fraction of simulated rolls compared to total rolls determines the fraction of damage done. So with smoothing of 3, you or the AI can do 0 damage, 1/3, 2/3, or 3/3 (ie. all) of the standard damage for an attack. The nice thing about this is when attacking a unit when it is on favourable terrain means you are likely to score some damage rather than all or nothing. But it does mean you can't be extremely lucky and have 6 AI units attack and do no damage at all by missing all their attacks like in mainline.
I think 0% guaranteed damage and 3 smoothing rolls is a nice balance between having all or nothing, or on the other side being completely predictable (a high damage split ratio % and/or smoothing number does change the feel of the game). It's nice to have a mid-range option now in Wesnoth 1.8 and I find having three levels of non-zero damage rather than one to make the game a lot more fun.
Anyway, if you are at all interested, try it out! The Reduce RNG influence thread, and other threads related to proposed Experimental Wesnoth features, can be found in the Experimental forum.
Have fun!
Best regards,
Steven.
Wesnoth-XP allows you to choose damage calculation settings (accessed by the General tab in Preferences). "Split ratio %" is the percentage, based on defense of unit being attacked, of damage that is guaranteed, and "RNG-Smoothing" is the number of simulated rolls for each attack. I find a ratio of 0% and 3 smoothing works well. Basically, the fraction of simulated rolls compared to total rolls determines the fraction of damage done. So with smoothing of 3, you or the AI can do 0 damage, 1/3, 2/3, or 3/3 (ie. all) of the standard damage for an attack. The nice thing about this is when attacking a unit when it is on favourable terrain means you are likely to score some damage rather than all or nothing. But it does mean you can't be extremely lucky and have 6 AI units attack and do no damage at all by missing all their attacks like in mainline.
I think 0% guaranteed damage and 3 smoothing rolls is a nice balance between having all or nothing, or on the other side being completely predictable (a high damage split ratio % and/or smoothing number does change the feel of the game). It's nice to have a mid-range option now in Wesnoth 1.8 and I find having three levels of non-zero damage rather than one to make the game a lot more fun. Anyway, if you are at all interested, try it out!
The Reduce RNG influence thread, and other threads related to proposed Experimental Wesnoth features, can be found in the Experimental forum.Have fun!
Best regards,
Steven.
I think the people complaining about AI cheating need to realize that 3 pounds of meat is still vastly superior to a few ounces of silicon, and will be for a long time.
Despite what Ray Kurzweil says, computers are a lot worse then humans at making plans. If you want a difficult game, the AI has to cheat for the time being.
Anyone who thinks I'm being defeatist on this, try this exercise: Multiply the number of operations per second that Deep Blue is capable of by the number of ways CiV is more complicated than chess, and divide by your CPU speed. That's the time between turns for a smart AI.
Despite what Ray Kurzweil says, computers are a lot worse then humans at making plans. If you want a difficult game, the AI has to cheat for the time being.
Anyone who thinks I'm being defeatist on this, try this exercise: Multiply the number of operations per second that Deep Blue is capable of by the number of ways CiV is more complicated than chess, and divide by your CPU speed. That's the time between turns for a smart AI.
Granted as fact, but when the AI cheats in obvious ways, it's an element that pulls you out of the immersion of the game. Cheating will be necessary as you move the difficulty slider towards the top, but ideally a good AI should be able to challenge the player at a normal difficulty level without cheating so obviously that the player is aware of it.
Ahriman
Tyrant
Its a nice ideal, but its hopelessly unrealistic.
Peng Qi
Emperor
Combat had better be nothing like BFW. I hate how absurdly random combat is in that game.
Lorteungen
Warlord
I've never played Wesnoth, but could you (or Sirp) explain the reasoning behind this design? It doesn't seem intuitive to me (and apparently to OP as well).
Peng Qi
Emperor
It's not intuitive. It's terrible. It results in strategically brilliant moves being rendered totally meaningless if you hit two bad rolls in a row. Let me put it this way. Here are the stats of two units that have nearly opposite functionalities:
Elvish Fighter-
HP: 33
Attacks:
5 - 4 melee (blade)
3 - 3 ranged (pierce)
The "attacks" mean "4 chances to hit at 5 damage each hit" and "3 chances to hit at 3 damage each hit."
Dwarvish Thunderer-
HP: 34
Attacks:
6 - 2 melee (blade)
18 - 1 ranged (pierce)
Now, chance to hit is based on the tile the defending unit is standing in compared to a chart provided for each individual unit. For example, the above elf standing in an open field has a 60% chance to be hit. Standing in a forest, he only has a 40% chance to be hit. The dwarf, meanwhile, will have a 70% chance to be hit in both a plains tile AND a forest tile.
Here's where the problem comes in: look at that dwarf's ranged damage. He has a single chance to hit and deals a massive 18 damage if he does. Imagine how much it would destroy your strategy if you just happen to get unlucky three turns in a row with his rifle. This isn't even particularly unlikely; even with a 60% chance to hit, your chances of missing three times in a row are 6.5%. If the game decides that it's 6.5% time at an inopportune moment, your entire strategy goes out the window and even a dramatically inferior unit on a dramatically inferior tile can lose to a superior one. This is compounded by the fact that when a unit fights or kills another unit, it gains XP, and when it gains enough XP to level, it fully heals and becomes stronger. Ergo, an unlucky string of attacks against one unit you had a 95% chance to kill could suddenly lead to an insurmountable situation for you if that enemy unit gets lucky and kills one or two of your units. This is not a far fetched situation; I have seen it happen dozens of times.
Honestly, I think Wesnoth is an excellent concept and game that's crippled by its combat mechanics. I can't even play it for more than a few games before something irritatingly unlikely happens that reminds me why I stopped playing it.
Ahriman
Tyrant
BfW is a fantasy-type game, with dungeons & dragons type creatures. Creatures can have different levels; a lowly dwarf warrior at level 1 can become a powerful dwarf lord at level 3.
Most fantasy games have some kind of "weapon skill" statistic, where skilled warriors have a higher chance to hit than low-skill warriors, and also get more attacks or do more damage.
And most tactical wargames have terrain modifiers; good terrain allows better defense.
BfW realizes that all of these aren't necessary to get the same result.
The probability that any weapon strikes a defender is solely determined by the terrain the defender is on; if they're in a river, they have 20% defense, so attacks have 80% chance to hit. If they're in open ground, they have 30% defense, so attacks have 70% chance to hit. If they're in forest or hills they have 50% defense, so attacks have 50% chance to hit.
[And different unit types can have different % chance on different terrains, so elves are actually 70% defense in forest no 30%, merfolk are 60% defense in rivers, and heavy infantry get damage resistance but worse defense.]
So, rather than high skill units needing some kind of weapon skill, they just have more attacks per turn, and each attack does more damage.
And the chance for each strike to hit (and thus the probability of various outcomes) is incredibly simple and transparent; its just dependent on the defenders terrain.
This terrain dependence has strong tactical implications; for example I have incentives to set up my defensive line on the edge of the forest, so that my units get the forest defense bonus while your units must stand on the lower defense open terrain in order to attack me.
I'm not saying this design would directly translate well into Civ, but it works perfectly in BfW
* * *
@Peng Qi
I think you just don't like probability. When you know the probabilities, and can work them into your plan (am I willing to risk this unit here? What if it gets unlucky?) then it adds a further level of depth that purely deterministic systems can't achieve. You have to plan not just on what will probably happen, but on what might happen, and adjust your strategy on the fly.
The thunderer for example is a high-risk high-reward unit - the most extreme case in the game.
The advantage of having that big up-front shot is that if it hits and kills the defender, the defender doesn't get to counterattack.
Some people like this, some people don't.
Most fantasy games have some kind of "weapon skill" statistic, where skilled warriors have a higher chance to hit than low-skill warriors, and also get more attacks or do more damage.
And most tactical wargames have terrain modifiers; good terrain allows better defense.
BfW realizes that all of these aren't necessary to get the same result.
The probability that any weapon strikes a defender is solely determined by the terrain the defender is on; if they're in a river, they have 20% defense, so attacks have 80% chance to hit. If they're in open ground, they have 30% defense, so attacks have 70% chance to hit. If they're in forest or hills they have 50% defense, so attacks have 50% chance to hit.
[And different unit types can have different % chance on different terrains, so elves are actually 70% defense in forest no 30%, merfolk are 60% defense in rivers, and heavy infantry get damage resistance but worse defense.]
So, rather than high skill units needing some kind of weapon skill, they just have more attacks per turn, and each attack does more damage.
And the chance for each strike to hit (and thus the probability of various outcomes) is incredibly simple and transparent; its just dependent on the defenders terrain.
This terrain dependence has strong tactical implications; for example I have incentives to set up my defensive line on the edge of the forest, so that my units get the forest defense bonus while your units must stand on the lower defense open terrain in order to attack me.
I'm not saying this design would directly translate well into Civ, but it works perfectly in BfW
* * *
@Peng Qi
I think you just don't like probability. When you know the probabilities, and can work them into your plan (am I willing to risk this unit here? What if it gets unlucky?) then it adds a further level of depth that purely deterministic systems can't achieve. You have to plan not just on what will probably happen, but on what might happen, and adjust your strategy on the fly.
The thunderer for example is a high-risk high-reward unit - the most extreme case in the game.
The advantage of having that big up-front shot is that if it hits and kills the defender, the defender doesn't get to counterattack.
Some people like this, some people don't.
And then there is the cuttlefish, TEN ATTACKS of three damage apiece, for playable actors it's the Duelist/Master-at-Arms
BfW mixes units with few attacks and big damage (dwarven thunderer, troll slingers...) and units with many attacks dealing less (elves). If you dislike risk, then picking the latter kind of units will workk better for you as more rolls mean less randomness. Also, magic having a flat-out 70% chance to hit whatever the terrain makes choosing mages important.
Versus the ai, in campaigns, randomness can be painful as you want to level certain units, and sometimes you just can't protect them all and may lose an invaluable unit, meaning you'll want to reload.
In MP, good players trump weaker ones and I'd say that luck is only a minor factor in deciding who wins.
Still, using the exact same system as BfW with units with less moves and more attacks would probably give good results for Civ and what its combat would look like (except the lack of support and ranged combat).
Versus the ai, in campaigns, randomness can be painful as you want to level certain units, and sometimes you just can't protect them all and may lose an invaluable unit, meaning you'll want to reload.
In MP, good players trump weaker ones and I'd say that luck is only a minor factor in deciding who wins.
Still, using the exact same system as BfW with units with less moves and more attacks would probably give good results for Civ and what its combat would look like (except the lack of support and ranged combat).
Peng Qi
Emperor
I tried to get into competitive PVP Wesnoth but literally every single game was determined by some random unlucky thing happening to the loser, whether that was me or my opponent.
That's what you must do in wesnoth. To minimize how much luck will decide who wins. How? You have extra units to attack if the unit doesn't die. You attack during day or night depending on which ones better. You stay on favorable terrain and attack enemies on grassland. If you don't manage to kill the enemy this turn okay. On the next turn you kill him. Don't forget that sometimes your own units survive near impossible odds.
Peng Qi
Emperor
If I wanted to play a game that revolved entirely around attempting to manage and mitigate random chance, I'd go play Texas hold 'em. At least then I could win some money.
All the perfect strategy in the world won't save you if the RNG decides it's time to make you lose the game.
No, it's really not, already even accomplished in civ4, by a few small tweaks to make the AI extra aggressive - probably on an average/not especially cheating difficulty this AI "challenges" 90%+ of civ players. So when civ4 can be fixed in this sense by simply making the AI slightly better at "winning" the game, at the very least civ5 shouldn't begin by introducing purposefully obscured and annoying mechanics, and in turn handicapping the AI with stupid limitations that don't affect the human.
"The AI get massive bonuses but are crippled by terrible diplomacy/always have to warmonger/religious crusade" just gets annoying. And it really can be done either way - let the AI work towards "winning the game" and they will challenge many players, and reduce bonuses needed all around even at highest levels, or let the immersion be realistic and more complete.
Anyway, a lot of challenge in a random civ game simply comes from inequal starts and luck, so even on a relatively equal footing a human player could still find themselves in a tough game. In short, I think they could remove the multiple free settlers/workers and those types of things and still have a challenging AI; or, they could go another way and make an AI far, far better with immersion and realism. A bit of effort, yes, but probably less than the time spent of fancy graphics and stuff anyway.
Actually it probably would. Unless RNG made you lose every single time you fought. Doesn't Civ also have lots of luck too?? Haven't your units in Civ lost against a 99.97% chance??
