Humble opinion about mod

[SpankmyMetroid]

Since we're throwing out ideas, here's my two cents. As mentioned the biggest difference between the AI and human play is the human can turtle units carefully until they become super-promoted behemoths and then roll over the AI with superior units in a 1UPT world, logistics and range+ being the biggest offenders which effectively multiplies your units' efficiency by 1.5x or makes them immune to counterattacks. Promotions swing the game. This rewards defensive gameplay, which is boring and also hard for the AI to follow. Players also only need to invest a less amount of gold/hammers to make the core army that they'll use the rest of the game, which, also as mentioned, is something the AI can't do as they need to constantly recycle units to be able to make up for losing them due to poorer strategy. Once a fighting army is reached, that equates to the human being able to pour more resources into infrastructures while the AI needs to spend time reinforcing their units, which will lead to a further disparity between empire effectiveness; being able to build libraries instead of spearmen to replace lost spearmen will obviously get you to pikemen faster. This is the grind, for better or for worse. But what if turtling like this was discouraged, and to balance this a more aggressive fighting style encouraged? And what if Authority (which has been seen as weak lately) can be buffed at the same time to further reward being more aggressive?

I think promotions should come rapidly but units can decay back to a certain level. Either greatly reduce the amount of experience needed to level up or make it so one kill = one level. This can either be over a certain amount of turns, and maybe XP gets halved upon promotion as well (This intuitively makes sense; a squadron of spearmen that fought in 2000BC shouldn't keep all their knowledge sitting as a garrison for a couple thousand years).

To spitball some completely arbitrary numbers, 15 exp is now needed to level up no matter the level, and every 15 (standard) turns after a unit's creation, their level drops down by 1. Melee units on the offense gain 7 xp when attacking, 3 on defense; a unit that's attacked, defended, and attacked again will receive a promotion. Ranged units receive less (3) on attack to make up for potentially firing with impunity. Attacks on cities and defense from city attacks use these numbers too. This accomplished a few things in my eyes:

• Wars become more momentum based; instead of throwing your best units at each other in an even grind, based on smart play units rapidly become stronger and can start to overwhelm enemy reinforcements that are comparatively less powerful. Intuitively, a company of soldiers that has seen lots of victories in a war is going to intimidate anyone who tries to stand against them next.
• For warmongering civs, constantly attacking to keep your units elite is now imperative, which rewards some of the other authority bonuses like yields on kill. If your army is victorious, you have a choice to jump right into the next war while your units still have their promotions, or take some cooldown time to shore up your empire and army but be at an equal playing field with your next target.
• This gives the player a chance to, instead of trying to make their whole army a bunch of super soldiers, make a few linchpin units and the rest more expendable (you're probably going to have to sacrifice some units to be able to enable the others to win their fights). It also removes a bit of the skill gap of keeping units alive for the entire game that the AI lacks.
• Ideally this would be enough of a bonus for the AI that they would no longer need "extra" promotions on harder difficulties to keep up with the player, which has always felt more unfun and cheaty than other difficulty buffs for the AI to me. Once you win your first war against the AI that have units 2 levels higher than you by default, the xp disparity is more balanced, making that advantage moot; it's really just a hump for the first one or two wars.
• Overall, making promotions easier to attain but hard to keep encourages offensive, fast gameplay and makes combat more rewarding in the short term instead of being a game long grind.

Couple more ideas along these lines, especially regarding Authority:

• Another option is to straight up award a promotion on destroying a unit (or effectively doing this by lowering the xp per combat and giving bonus exp to destroying a unit) to discourage tanking. I can see a cheese for the player that puts a unit with cover buffs in a chokepoint and guzzles ranged attacks without actually putting that unit at risk.
• XP promotion buildings grant XP but can raise the floor of XP that a unit can decay down to; units trained with a barracks will never drop below level 1, armory 2, etc. This makes these buildings still valuable and gives an incentive to retrain units once a new building is unlocked. If you don't do this but your enemy does, they will start the war being a level higher than you.
• If XP is lost on upgrade instead of decaying over time, this now introduces a choice for the player between upgrading or not, which was always strictly better given the gold was available. Choice is the lifeblood of 4X.
• Another buff that can be given to authority would be rewarding yields on death as a policy bonus (sort of like a culture finding honor in death). This encourages smart sacrifice of units as well as making it not a total wash to have to refresh an army as needed. Yes, this cheapens some religious tenets but gives some much needed yields to AI civs going auth and throwing away hammers from expendable units that could have been going into infrastructure.
• Defense is typically much easier than offense, for good reason. Defensive bonuses from terrain and fortification as well as city attacks provide a huge advantage to the defender. Unless I'm mistaken, typically two units that fight each other are awarded the same amount of exp (5 and 5 for melee? not sure about a ranged attacker and their defender). As I mentioned in the arbitrary numbers part, what if the attacking unit is awarded slightly more and the defending unit slightly less? This creates an xp advantage to the attacker and encourages the momentum mentioned above; even if you send some units to their deaths, your army is still becoming progressively stronger than theirs. This also is less of an issue for Auth since their war weariness bonuses lets them grind through units without losing as much happiness.
• On that note, authority should be constantly making units and be more willing to throw them at the enemy to break through their defenses. What if, as another authority policy buff, an authority unit dying gives a small combat buff (5%) for the next turn to every friendly unit on adjacent tiles? This further encourages sacrifice of units and gives attackers an advantage.

Issues with this:

• If a unit decays to a prior level, what determines which promotion is lost? Should it be up to the player to decide (if this is possible to code) or should it default to the highest promotion in the tree that is has? What if it has several in the same tier (like shock 1/ drill 1)?
• There needs to be a good way to heal melee units you decide you want to make the linchpins of your army so they can constantly be fighting. Ideally there would be a way to reinforce a unit with another adjacent unit (ala advance wars) but I don't know if that's functionally possible, or the best approach. Maybe being promoted should heal more than 10?
• Medic units will decay if not in combat, and you don't want them in combat. Should units with medic gain a small amount of exp based on how much HP surrounding units heal? I think a better solution would be to have a dedicated medic support unit like in Civ 6.
• Should scouts (or other units that get exp from noncombat) decay?
• This is probably too much change for people to be comfortable with.

Please let me know if this is crazy or not.

 

[CppMaster]

The AI wins because it has far better reflexes. It is not even slightly outplaying a human in terms of thinking, it is just much much faster. The starcraft alphago really struggled once they prevented it from doing super human things and in some cases things that were literally impossible for humans.

All these sentences are false. You were talking about SC2, Dota 2 and Go and me too, but also chess.

It is better than at the start with limited APM and not microing stalkers across multiple screens but it still spikes to 1500 apm in battles, something no human can do. 5:50 timestamp.

This a histogram of APM per player with mean APM. APM of AlphaStar is lower in general, except for some spikes, and TLO has even bigger spikes. It's a similar micro that pro players do regularly:

Quotes from pro players that played against AlphaStar:
"I was surprised by how strong the agent was, AlphaStar takes well-known strategies and turns them on their head. The agent demonstrated strategies I hadn’t thought of before, which means there may still be new ways of playing the game that we haven’t fully explored yet."
"I was impressed to see AlphaStar pull off advanced moves and different strategies across almost every game, using a very human style of gameplay I wouldn’t have expected. I’ve realised how much my gameplay relies on forcing mistakes and being able to exploit human reactions, so this has put the game in a whole new light for me. We’re all excited to see what comes next."

So either you are wrong or these pro players, authors of AlphaStar, professional commentators and basically everyone interested in the topic. You're the only one who think it only wins, because of how fast it is, not because of strategic thinking. My money is that on that you're the one who is wrong.

So that's StarCraft 2, but there also are Dota 2, Go and Chess. Do you want to dispute these too?

 

[N.Core]

One of the biggest problems with the tactical AI (unless I'm greatly mistaken) is that it doesn't remember the positions of units. If a unit vanishes into tiles that aren't visible, the AI forgets about it, while a human can recognize where that unit likely went and adapt accordingly. This is why it gets extra sight on Immortal and Deity.

The tactical AI works similarly to a chess AI, in that the "combat sim" runs through different possibilities and selects the one which it thinks will result in the best outcome. It has a maximum search depth for the sake of performance, however, so turn times don't take forever.

And there's also memory constraints. You don't want to make a feature where AI can "remember" something but heavily tanking your memory.
While a chess board is only 8*8 tile, a visible part of a civ map is way bigger than that.

 

[ilteroi]

If I were to do that, I'll probably do a separate AI model for different aspects of the game, like tactics, what to produce etc. Tactics are interesting to start with. I could try to train an agent that tries to beat Iteroi's algorithms.

I do ML for a living and RL as a hobby. I've play with some toy tasks and Atari games: https://github.com/CppMaster/openai-gym-playground and I've managed to beat the hardest difficulty of StarCraft 2: https://github.com/CppMaster/SC2-AI
Of course, my AI model is way simpler than AlphaStar and would struggle against good players, because I'm just one person doing it after hours on my PC as opposed to an entire experienced team that were doing it for several months on a cluster of very powerful machines

now we're talking!

actually when i started tinkering with civ in 2014, i thought about whether this isn't a case for AI. but the engine just isn't very automation friendly, and the ecosystem wasn't where it is now ... so i told myself that i'd just do some quick fixes the oldfashioned way. and look at me now, arguing with strangers on the internet

seriously, if you're interested we can probably rig up an easy "game state exfiltration" with zeromq + protobuf. this would send each player's map view to some other process (eg python for AI convenience) on turn start. then you can watch and learn tactics. next step would be to receive back missions for AI units.

there is one giant problem though, we cannot run the gamecore standalone and with the UI everything is sloooooow and cumbersome. how many games do you reckon you need to observe that way?

 

[stii]

All these sentences are false. You were talking about SC2, Dota 2 and Go and me too, but also chess.

View attachment 646867
This a histogram of APM per player with mean APM. APM of AlphaStar is lower in general, except for some spikes, and TLO has even bigger spikes. It's a similar micro that pro players do regularly:

Quotes from pro players that played against AlphaStar:
"I was surprised by how strong the agent was, AlphaStar takes well-known strategies and turns them on their head. The agent demonstrated strategies I hadn’t thought of before, which means there may still be new ways of playing the game that we haven’t fully explored yet."
"I was impressed to see AlphaStar pull off advanced moves and different strategies across almost every game, using a very human style of gameplay I wouldn’t have expected. I’ve realised how much my gameplay relies on forcing mistakes and being able to exploit human reactions, so this has put the game in a whole new light for me. We’re all excited to see what comes next."

So either you are wrong or these pro players, authors of AlphaStar, professional commentators and basically everyone interested in the topic. You're the only one who think it only wins, because of how fast it is, not because of strategic thinking. My money is that on that you're the one who is wrong.

So that's StarCraft 2, but there also are Dota 2, Go and Chess. Do you want to dispute these too?

You know what I was talking about? Man that is an impressive skill.

Lets be clear I WAS NOT. Don't tell me what I WAS TALKING ABOUT. You are free to ask but it is utterly absurd that you can decide what I was saying and then argue against it. It isn't even a strawman at this point is is just pointless.

I don't agree with the other things you are saying either but what is the point? You aren't replying to me you are making things up and then replying to those things. If you want more context or explanation then ask but otherwise go talk to yourself.



All this APM talk is also pointless because yet again you ignore what I said and reply to something else. It spiked to 1500. That is not possible for a human. I'm literally just repeating myself but you utterly ignore it so apparently I need to do this?

Do I want to dispute other things I haven't said. Well no but who knows what you might make up next?

 

[ma_kuh]

One of the biggest problems with the tactical AI (unless I'm greatly mistaken) is that it doesn't remember the positions of units. If a unit vanishes into tiles that aren't visible, the AI forgets about it, while a human can recognize where that unit likely went and adapt accordingly. This is why it gets extra sight on Immortal and Deity.

Does the AI use recon units for start-of-turn scouting, like a human might?

 

[CppMaster]

All this APM talk is also pointless because yet again you ignore what I said and reply to something else. It spiked to 1500. That is not possible for a human. I'm literally just repeating myself but you utterly ignore it so apparently I need to do this?

It is literally possible, because of how the game calculates APM and it's in the graph that TLO has much higher APM spike.

Lets be clear I WAS NOT. Don't tell me what I WAS TALKING ABOUT. You are free to ask but it is utterly absurd that you can decide what I was saying and then argue against it. It isn't even a strawman at this point is is just pointless.

Don't lie, you were talking Dota 2 and Go too:

Yeah no those games are nothing like Civ (at least in terms of why the AI wins). The AI wins because it has far better reflexes. It is not even slightly outplaying a human in terms of thinking, it is just much much faster. The starcraft alphago really struggled once they prevented it from doing super human things and in some cases things that were literally impossible for humans.

"Those games" refer to Dota 2 and StarCraft 2, AlphaGo is Go.

Do I want to dispute other things I haven't said. Well no but who knows what you might make up next?

I asked, because even if it wasn't true for StarCraft 2 (it is), the other games (Dota 2, Chess, Go) still prove that AI can beat human in strategic decisions.

 

[Recursive]

Moderator Action: Cool down the temperature, and please refrain from personal attacks. Your disagreements about AI effectiveness are fine, but your hostility towards each other is not. Further hostility will result in moderation action.

 

[CppMaster]

seriously, if you're interested we can probably rig up an easy "game state exfiltration" with zeromq + protobuf. this would send each player's map view to some other process (eg python for AI convenience) on turn start. then you can watch and learn tactics. next step would be to receive back missions for AI units.

there is one giant problem though, we cannot run the gamecore standalone and with the UI everything is sloooooow and cumbersome. how many games do you reckon you need to observe that way?

I had the same issue with StarCraft 2 on Windows that I couldn't train without rendering. It made training about twice longer.

As for number of games, there is no way to know until I design observation space, action space and reward structure. Even then it's hard to guess. The more the better. It'd be best to produce data in real time and train on it at the same time. Then I'd just run it until it learns.

 

[Recursive]

Does the AI use recon units for start-of-turn scouting, like a human might?

I don't know, ask @ilteroi.

 

[stii]

One of the biggest problems with the tactical AI (unless I'm greatly mistaken) is that it doesn't remember the positions of units. If a unit vanishes into tiles that aren't visible, the AI forgets about it, while a human can recognize where that unit likely went and adapt accordingly. This is why it gets extra sight on Immortal and Deity.

The tactical AI works similarly to a chess AI, in that the "combat sim" runs through different possibilities and selects the one which it thinks will result in the best outcome. It has a maximum search depth for the sake of performance, however, so turn times don't take forever.

I seem to remember the AI from earlier civ games the AI would just still see units that moved out of vision for 2 turns. Which is a bit of a hack but mirrors how a human works pretty well, you know pretty much exactly where a unit is after one turn but five turns later you don't really have a clue.

 

[t99KJwEd]

Trying out @balparmak's "Dynamic Battle Lines" makes the AI's lack of object permanence obvious. It will gladly attack my "exposed" units and push into my backline to die as long as it loses vision between turns.

I know this is ugly, but maybe a better solution to the warring problem is to make AI units even more overpowered. It makes superior player tactics more important and mitigates the lowered difficulty from cutting down on AI unit spam. I think the AI having magically strong units is more satisfying to play against than them magically buying armies and forcing more grind. Less units also means more room for AI improvements on both logistical and processing-practicality levels.

I think some of the continued reliance on unit supply is tied up with how monotonous naval combat is. Limited unit types for the deciding part of the game, no terrain, and limited alternatives for defending colonies, means the solution is just spam a lot of boats. If that part of the game were more tactically complex it might justify cutting supply more drastically.

 

[DizzKneeLand33]

You guys are amazing -- if the Civ 6 team had 1/10th the dedication all of you have to making a game as good as possible then it would be worth playing. Instead, time to cross the 6000 hour mark on Civ 5, again thanks to all of you.

 

[SpankmyMetroid]

Since we're throwing out ideas, here's my two cents. As mentioned the biggest difference between the AI and human play is the human can turtle units carefully until they become super-promoted behemoths and then roll over the AI with superior units in a 1UPT world, logistics and range+ being the biggest offenders which effectively multiplies your units' efficiency by 1.5x or makes them immune to counterattacks. Promotions swing the game. This rewards defensive gameplay, which is boring and also hard for the AI to follow. Players also only need to invest a less amount of gold/hammers to make the core army that they'll use the rest of the game, which, also as mentioned, is something the AI can't do as they need to constantly recycle units to be able to make up for losing them due to poorer strategy. Once a fighting army is reached, that equates to the human being able to pour more resources into infrastructures while the AI needs to spend time reinforcing their units, which will lead to a further disparity between empire effectiveness; being able to build libraries instead of spearmen to replace lost spearmen will obviously get you to pikemen faster. This is the grind, for better or for worse. But what if turtling like this was discouraged, and to balance this a more aggressive fighting style encouraged? And what if Authority (which has been seen as weak lately) can be buffed at the same time to further reward being more aggressive?

I think promotions should come rapidly but units can decay back to a certain level. Either greatly reduce the amount of experience needed to level up or make it so one kill = one level. This can either be over a certain amount of turns, and maybe XP gets halved upon promotion as well (This intuitively makes sense; a squadron of spearmen that fought in 2000BC shouldn't keep all their knowledge sitting as a garrison for a couple thousand years).

To spitball some completely arbitrary numbers, 15 exp is now needed to level up no matter the level, and every 15 (standard) turns after a unit's creation, their level drops down by 1. Melee units on the offense gain 7 xp when attacking, 3 on defense; a unit that's attacked, defended, and attacked again will receive a promotion. Ranged units receive less (3) on attack to make up for potentially firing with impunity. Attacks on cities and defense from city attacks use these numbers too. This accomplished a few things in my eyes:

• Wars become more momentum based; instead of throwing your best units at each other in an even grind, based on smart play units rapidly become stronger and can start to overwhelm enemy reinforcements that are comparatively less powerful. Intuitively, a company of soldiers that has seen lots of victories in a war is going to intimidate anyone who tries to stand against them next.
• For warmongering civs, constantly attacking to keep your units elite is now imperative, which rewards some of the other authority bonuses like yields on kill. If your army is victorious, you have a choice to jump right into the next war while your units still have their promotions, or take some cooldown time to shore up your empire and army but be at an equal playing field with your next target.
• This gives the player a chance to, instead of trying to make their whole army a bunch of super soldiers, make a few linchpin units and the rest more expendable (you're probably going to have to sacrifice some units to be able to enable the others to win their fights). It also removes a bit of the skill gap of keeping units alive for the entire game that the AI lacks.
• Ideally this would be enough of a bonus for the AI that they would no longer need "extra" promotions on harder difficulties to keep up with the player, which has always felt more unfun and cheaty than other difficulty buffs for the AI to me. Once you win your first war against the AI that have units 2 levels higher than you by default, the xp disparity is more balanced, making that advantage moot; it's really just a hump for the first one or two wars.
• Overall, making promotions easier to attain but hard to keep encourages offensive, fast gameplay and makes combat more rewarding in the short term instead of being a game long grind.

Couple more ideas along these lines, especially regarding Authority:

• Another option is to straight up award a promotion on destroying a unit (or effectively doing this by lowering the xp per combat and giving bonus exp to destroying a unit) to discourage tanking. I can see a cheese for the player that puts a unit with cover buffs in a chokepoint and guzzles ranged attacks without actually putting that unit at risk.
• XP promotion buildings grant XP but can raise the floor of XP that a unit can decay down to; units trained with a barracks will never drop below level 1, armory 2, etc. This makes these buildings still valuable and gives an incentive to retrain units once a new building is unlocked. If you don't do this but your enemy does, they will start the war being a level higher than you.
• If XP is lost on upgrade instead of decaying over time, this now introduces a choice for the player between upgrading or not, which was always strictly better given the gold was available. Choice is the lifeblood of 4X.
• Another buff that can be given to authority would be rewarding yields on death as a policy bonus (sort of like a culture finding honor in death). This encourages smart sacrifice of units as well as making it not a total wash to have to refresh an army as needed. Yes, this cheapens some religious tenets but gives some much needed yields to AI civs going auth and throwing away hammers from expendable units that could have been going into infrastructure.
• Defense is typically much easier than offense, for good reason. Defensive bonuses from terrain and fortification as well as city attacks provide a huge advantage to the defender. Unless I'm mistaken, typically two units that fight each other are awarded the same amount of exp (5 and 5 for melee? not sure about a ranged attacker and their defender). As I mentioned in the arbitrary numbers part, what if the attacking unit is awarded slightly more and the defending unit slightly less? This creates an xp advantage to the attacker and encourages the momentum mentioned above; even if you send some units to their deaths, your army is still becoming progressively stronger than theirs. This also is less of an issue for Auth since their war weariness bonuses lets them grind through units without losing as much happiness.
• On that note, authority should be constantly making units and be more willing to throw them at the enemy to break through their defenses. What if, as another authority policy buff, an authority unit dying gives a small combat buff (5%) for the next turn to every friendly unit on adjacent tiles? This further encourages sacrifice of units and gives attackers an advantage.

Issues with this:

• If a unit decays to a prior level, what determines which promotion is lost? Should it be up to the player to decide (if this is possible to code) or should it default to the highest promotion in the tree that is has? What if it has several in the same tier (like shock 1/ drill 1)?
• There needs to be a good way to heal melee units you decide you want to make the linchpins of your army so they can constantly be fighting. Ideally there would be a way to reinforce a unit with another adjacent unit (ala advance wars) but I don't know if that's functionally possible, or the best approach. Maybe being promoted should heal more than 10?
• Medic units will decay if not in combat, and you don't want them in combat. Should units with medic gain a small amount of exp based on how much HP surrounding units heal? I think a better solution would be to have a dedicated medic support unit like in Civ 6.
• Should scouts (or other units that get exp from noncombat) decay?
• This is probably too much change for people to be comfortable with.

Please let me know if this is crazy or not.

why is it a bad idea

 

[wabab]

The "1 unit per tile" rule that Civ 5 has does not actually eliminate the unit stacking issue; it only limits stacking to a units movement points. With ranged units, stacking is now effectively a function of attack range and number of tiles that simultaneously fire on a single unit.

I'm not sure what a good fix for this is. Letting melee units attack after moving into position may help (or could be a terrible idea; who knows) since it would force the front lines to move more. A more ideal solution (for me, at least) would be putting different supply caps on different units. Units with lower caps would be (substantially) more powerful than units with higher supply caps. Even when surrounded by high supply cap units, low supply cap units should be able to defeat a fair number of them before dying.

 

[Recursive]

It is unlikely that any major overhauls to fundamental combat mechanics will be added into VP at this stage. XP and promotions might be revised as there seems to be considerable interest in doing so, but anything involving one UPT is probably gonna stay the same.

 

[rkkn]

Since we're throwing out ideas, here's my two cents. As mentioned the biggest difference between the AI and human play is the human can turtle units carefully until they become super-promoted behemoths and then roll over the AI with superior units in a 1UPT world, logistics and range+ being the biggest offenders which effectively multiplies your units' efficiency by 1.5x or makes them immune to counterattacks. Promotions swing the game. This rewards defensive gameplay, which is boring and also hard for the AI to follow. Players also only need to invest a less amount of gold/hammers to make the core army that they'll use the rest of the game, which, also as mentioned, is something the AI can't do as they need to constantly recycle units to be able to make up for losing them due to poorer strategy. Once a fighting army is reached, that equates to the human being able to pour more resources into infrastructures while the AI needs to spend time reinforcing their units, which will lead to a further disparity between empire effectiveness; being able to build libraries instead of spearmen to replace lost spearmen will obviously get you to pikemen faster. This is the grind, for better or for worse. But what if turtling like this was discouraged, and to balance this a more aggressive fighting style encouraged? And what if Authority (which has been seen as weak lately) can be buffed at the same time to further reward being more aggressive?

I think promotions should come rapidly but units can decay back to a certain level. Either greatly reduce the amount of experience needed to level up or make it so one kill = one level. This can either be over a certain amount of turns, and maybe XP gets halved upon promotion as well (This intuitively makes sense; a squadron of spearmen that fought in 2000BC shouldn't keep all their knowledge sitting as a garrison for a couple thousand years).

To spitball some completely arbitrary numbers, 15 exp is now needed to level up no matter the level, and every 15 (standard) turns after a unit's creation, their level drops down by 1. Melee units on the offense gain 7 xp when attacking, 3 on defense; a unit that's attacked, defended, and attacked again will receive a promotion. Ranged units receive less (3) on attack to make up for potentially firing with impunity. Attacks on cities and defense from city attacks use these numbers too. This accomplished a few things in my eyes:

• Wars become more momentum based; instead of throwing your best units at each other in an even grind, based on smart play units rapidly become stronger and can start to overwhelm enemy reinforcements that are comparatively less powerful. Intuitively, a company of soldiers that has seen lots of victories in a war is going to intimidate anyone who tries to stand against them next.
• For warmongering civs, constantly attacking to keep your units elite is now imperative, which rewards some of the other authority bonuses like yields on kill. If your army is victorious, you have a choice to jump right into the next war while your units still have their promotions, or take some cooldown time to shore up your empire and army but be at an equal playing field with your next target.
• This gives the player a chance to, instead of trying to make their whole army a bunch of super soldiers, make a few linchpin units and the rest more expendable (you're probably going to have to sacrifice some units to be able to enable the others to win their fights). It also removes a bit of the skill gap of keeping units alive for the entire game that the AI lacks.
• Ideally this would be enough of a bonus for the AI that they would no longer need "extra" promotions on harder difficulties to keep up with the player, which has always felt more unfun and cheaty than other difficulty buffs for the AI to me. Once you win your first war against the AI that have units 2 levels higher than you by default, the xp disparity is more balanced, making that advantage moot; it's really just a hump for the first one or two wars.
• Overall, making promotions easier to attain but hard to keep encourages offensive, fast gameplay and makes combat more rewarding in the short term instead of being a game long grind.

Couple more ideas along these lines, especially regarding Authority:

• Another option is to straight up award a promotion on destroying a unit (or effectively doing this by lowering the xp per combat and giving bonus exp to destroying a unit) to discourage tanking. I can see a cheese for the player that puts a unit with cover buffs in a chokepoint and guzzles ranged attacks without actually putting that unit at risk.
• XP promotion buildings grant XP but can raise the floor of XP that a unit can decay down to; units trained with a barracks will never drop below level 1, armory 2, etc. This makes these buildings still valuable and gives an incentive to retrain units once a new building is unlocked. If you don't do this but your enemy does, they will start the war being a level higher than you.
• If XP is lost on upgrade instead of decaying over time, this now introduces a choice for the player between upgrading or not, which was always strictly better given the gold was available. Choice is the lifeblood of 4X.
• Another buff that can be given to authority would be rewarding yields on death as a policy bonus (sort of like a culture finding honor in death). This encourages smart sacrifice of units as well as making it not a total wash to have to refresh an army as needed. Yes, this cheapens some religious tenets but gives some much needed yields to AI civs going auth and throwing away hammers from expendable units that could have been going into infrastructure.
• Defense is typically much easier than offense, for good reason. Defensive bonuses from terrain and fortification as well as city attacks provide a huge advantage to the defender. Unless I'm mistaken, typically two units that fight each other are awarded the same amount of exp (5 and 5 for melee? not sure about a ranged attacker and their defender). As I mentioned in the arbitrary numbers part, what if the attacking unit is awarded slightly more and the defending unit slightly less? This creates an xp advantage to the attacker and encourages the momentum mentioned above; even if you send some units to their deaths, your army is still becoming progressively stronger than theirs. This also is less of an issue for Auth since their war weariness bonuses lets them grind through units without losing as much happiness.
• On that note, authority should be constantly making units and be more willing to throw them at the enemy to break through their defenses. What if, as another authority policy buff, an authority unit dying gives a small combat buff (5%) for the next turn to every friendly unit on adjacent tiles? This further encourages sacrifice of units and gives attackers an advantage.

Issues with this:

• If a unit decays to a prior level, what determines which promotion is lost? Should it be up to the player to decide (if this is possible to code) or should it default to the highest promotion in the tree that is has? What if it has several in the same tier (like shock 1/ drill 1)?
• There needs to be a good way to heal melee units you decide you want to make the linchpins of your army so they can constantly be fighting. Ideally there would be a way to reinforce a unit with another adjacent unit (ala advance wars) but I don't know if that's functionally possible, or the best approach. Maybe being promoted should heal more than 10?
• Medic units will decay if not in combat, and you don't want them in combat. Should units with medic gain a small amount of exp based on how much HP surrounding units heal? I think a better solution would be to have a dedicated medic support unit like in Civ 6.
• Should scouts (or other units that get exp from noncombat) decay?
• This is probably too much change for people to be comfortable with.

Please let me know if this is crazy or not.

This is a very interesting idea. I think it would probably give war more strategic depth than it currently has. It's certainly something I'd want to try out.
And on a more minor note, it also makes sense from a realism perspective. A military unit an era later isn't going to be made up of the same individual people. So indeed some of that hands on knowledge and wisdom (aka experience) is going to be lost over time.


It's a very dramatic shift, though, which is a big negative on its own, as per your very last point. It begins approaching the "this is a different game entirely" line. And...

encourages offensive, fast gameplay and makes combat more rewarding in the short term instead of being a game long grind.

civ in general is a game mostly about the long grind, so the fact that combat is as well is not out of place. I'd even wager that many people are playing the game for this very reason. The gradual buildup of a mighty unbeatable empire.

 

[stii]

Also if ranged units are the problem (they are) then you can just make them worse. It is much much harder to get high level melee units. Just reducing the CS of all ranged units would make the game quite a bit harder.

 

[Amask]

Letting melee units attack after moving into position may help

Along similar lines, maybe just more movement points for all units.
a) infantry would then actually be able to do something after it moves onto a hill, instead of just dying to ranged fire right away.
b) if it survives, it could retreat more easily, without interfering with the ranged troops and siege
c) if AI is able to save more units, they'll eventually accumulate more scary promotions
d) cavalry becomes more useful. Even with its current movement points it's often not enough to hit and run to safety

I'll probably try to mod this myself at some point and report the results, but it won't be for a while.

Also considering adding 50hp to all units.
a) AI gets to save more units -> more promotions for them
b) if you manage to occupy good positions around an enemy city, it's now harder for the defender to push you away, and he can't just snipe your siege as easily
c) damaged armies take longer to heal, giving the side with the advantage or bigger reserves more of a window to make progress

I'm aware there are a bunch of issues with both of these ideas, so I'm not suggesting we try implementing them, but I think they have some merit and are worth pondering upon.

 

[Stalker0]

One idea I had thought about was lowering the return damage from a melee attack. This would make melee stronger and more aggressive. Often melee units are forced to defend because the attack puts them too low to survive return damage. But if that was lower melee units could push more often