Information about developments at snowballing problem

Currently viewing this thread:

Here all test results since 1.5.4 updated with 16 x 1.5.7 results

nf-lK.png

Actually 1.5.6 and 1.5.7 results are nearly same 60.4 vs 64.8, maybe if we continue collecting 3-4 samples they can be even equal.

So improvements done in 1.5.7 seems effected less while some other developments like longer wars and xp limit changes can be neutralized developments. Its interesting Khuzaits have both 46 average score. Probably this can be related to luck, normally 1.5.7 Khuzait average should be less. Actually there were no big developments in 1.5.7 (even a small number change maybe most positive effected can be reducing cavalry ratio speed bonus to 40%) so it can be normal having similar results with 1.5.6. After that finding something reduces snowball score much will be harder. We can do this of course with small additions like having more militia at settlements or less hostile actions / less wars or adding cheat money to poor clans so they will defect less but gameplay can be effected badly and I have no control over troop trees for suggestions at there. Probably I cannot change these and I do not think they can have major effect.
I would actually say that a 4.4 in difference is quite big once you take in considerations that you had less people doing the 1.5.7 test than the 1.5.6 one.

I personally see it as an issue that the Khuzaits are guranteed to survive in either their original or expanded form while never losing too much.
The Southern and Northern Empire are always underperforming except for a few fringe cases. Which I understand because of the civil war, but it's never because one defeats the other and the Khuzaits destroy the weaker one. It seems to always be that they weaken eachother and the Khuzaits wrecks both.
The Western Empire I got no issues with. They're a hit or miss faction, which I'm fine with.
Sturgia is a faction that's hit or miss as well. I'm fine with this too.
Aserai seem to always stay in their dessert (As can be seen in their average number of mid 20's). This is a pretty big ssue since it makes them a very stale faction.
Battania and Vlandia is bound to kill one another. This makes sense lorewise, but it makes whoever wins too strong for the other kingdoms to handle. The ideal would be either kingdom gets split between 2 different kingdoms, but due to their locations this is very rarely the case.

Yeah, but make sure you label it, so people don't confuse a modded test run for vanilla.



Stronger factions get stronger. Weaker factions get weaker. It builds slowly, and from very little, like a snowball but it is the long-term effect of too many deterministic outcomes in a game.

I understand that's what happens, but my issue comes from when it's only 2 factions that are snowballing. The ideal in my world would be 3-4 factions in the end game. All of them fairly equal in strength. As it stands right now it seems more to be 1-2 strong factions. 2-3 moderate factions and the rest are too weak to become a moderately sized faction. It gets worse when you know nearly 100% which factions are gonna start snowballing (Like the Khuzaits and whoever is winning in the western part of the continent)
 

VVyrd

Knight at Arms
WBWF&SM&B
We have one more problem by the way. Longer peace time means poorer clans currently. Because 35% of clan income is wars and 65% of clan income is taxes currently (as average). So if AI vs AI wars decreases this will increase poor clans especially if a kingdom lost half of their territory and if they have no enemy they will have big problems. Loot from looters do not bring that much gold. You can suggest making this ratio 20% loot / 80% taxes by increasing taxes and decreasing loots but this can make some players unhappy. I already reduced loots a bit in 1.5.7. Maybe can reduce more later and increase taxes a bit. However this time still kingdoms lost territories suffer much because losing territory means lower tax income also.
Maybe a revanchism bonus would help? As in, when a faction loses a war, they get a temporary buff to help them get back on their feet. The more they lost in the war, the bigger the buff, with it diminishing with time until it fades. Shouldn't be anything too crazy though or it'd be unfair.
 

Dabos37

Sergeant Knight
@mexxico @Blood Gryphon @Apocal I have had to stop the current test because while some kingdoms have been actively fighting (especially WE, NE, and Vlandia have been at war almost all days - 600 days or so - ), Sturgia for some reason has been 80-100 days at peace. This is especially an issue due to the passive cheat for the AI, and Sturgia now has +8K strenght due to passive XP and Raganvad just have 3000-3500 at day +500 denars xD. SO yes, AI needs wars to get money, especially because they amass an insane amount of elite troops as you can see below:



So yes, while having less wars feels amazing IMO and a slower pace would be much better for the game, with the current mechanics it is not viable. It could work if some changes would be added, for example:

- Making the AI troops XP cheat only active if they are at war, or just making it getting deactivated after X days of peace (some like: if kingdomDaysPeace > 10, then Cheat = 0). The reason behind do not deactivating the cheat just after a kingdom making peace, is that it would be bad for kingdoms making peace because they are loosing a war and need decent troops for the next war to try to come back.
- Another option could be deactivate the passive AI cheat for +Tier2 or +Tier3 units.
- Some kingdoms being at peace for 80-100 days would be a problem, especially if lacking more content. This won't be a problem if we would have more ways to initiate wars like quests, or something else, and not just if you use influence as vassal.

Mexxico, I think that you are probably thinking "I have tons of work for making such kind of rebalancing and these ideas maybe won't get accepted", and this is ok. Just keep in mind that if you have some time in the future, I think that the bigger part of the player base would like a slower pace where endless wars would not be a thing.
 
Last edited:

Old-Bull

Sergeant at Arms
We have one more problem by the way. Longer peace time means poorer clans currently. Because 35% of clan income is wars and 65% of clan income is taxes currently (as average). So if AI vs AI wars decreases this will increase poor clans especially if a kingdom lost half of their territory and if they have no enemy they will have big problems. Loot from looters do not bring that much gold. You can suggest making this ratio 20% loot / 80% taxes by increasing taxes and decreasing loots but this can make some players unhappy. I already reduced loots a bit in 1.5.7. Maybe can reduce more later and increase taxes a bit. However this time still kingdoms lost territories suffer much because losing territory means lower tax income also.

I still think armies at war are making too much money.

In reality, funding an army at war is a gigantic money sink. You have to feed them, equip them, and house them. You have to have a supply line or the army starves and slows down.

In this game, food is cheap, equipping the army and housing the army is free. You don't have to feed the horses.You don't have to have a supply line. The army doesn't have to sleep so they never have to stop to set up camp, eat, sleep, and break camp. Huge armies move incredibly fast. The profits from looting is insanely high. It results in snowballing because the only way to pay for rapid expansion is to go to war and stay at war.

Maybe some of the loot should be subtracted to equip the army. Maybe some of the loot should be subtracted to pay for tents. Maybe some of the loot should be subtracted to feed all the horses. Maybe armies should have to occasionally stop to eat, sleep, and resupply.. A lot of medieval wars ended because the country being invaded would burn all the food and sack their own city before the army got there.

Maybe army cohesion should decay faster.

Without a supply line mechanic in the game, armies are just too fast and too profitable for reasonable balance.

I think the games economy is way out of balance. Wars have historically been funded through taxes but in this game fiefs actually cost you money. They don't make enough in taxes to pay for themselves. You should have to have a profitable kingdom to fund expansion.

At this point snowballing is a result of economic balance, not op cultural bonuses. A faction from the economically poor steppes shouldn't dominate every game, but they do. They're the richest faction in game because of the low cost of army upkeep and the insanely high profitability of war.
 
Last edited:

Dabos37

Sergeant Knight
@mexxico @Blood Gryphon @Apocal I have had to stop the current test because while some kingdoms have been actively fighting (especially WE, NE, and Vlandia have been at war almost all days - 600 days or so - ), Sturgia for some reason has been 80-100 days at peace. This is especially an issue due to the passive cheat for the AI, and Sturgia now has +8K strenght due to passive XP and Raganvad just have 3000-3500 at day +500 denars xD. SO yes, AI needs wars to get money, especially because they amass an insane amount of elite troops as you can see below:



So yes, while having less wars feels amazing IMO and a slower pace would be much better for the game, with the current mechanics it is not viable. It could work if some changes would be added, for example:

- Making the AI troops XP cheat only active if they are at war, or just making it getting deactivated after X days of peace (some like: if kingdomDaysPeace > 10, then Cheat = 0). The reason behind do not deactivating the cheat just after a kingdom making peace, is that it would be bad for kingdoms making peace because they are loosing a war and need decent troops for the next war to try to come back.
- Another option could be deactivate the passive AI cheat for +Tier2 or +Tier3 units.
- Some kingdoms being at peace for 80-100 days would be a problem, especially if lacking more content. This won't be a problem if we would have more ways to initiate wars like quests, or something else, and not just if you use influence as vassal.

Mexxico, I think that you are probably thinking "I have tons of work for making such kind of rebalancing and these ideas maybe won't get accepted", and this is ok. Just keep in mind that if you have some time in the future, I think that the bigger part of the player base would like a slower pace where endless wars would not be a thing.

Concerning this failed test, I am thinking about trying it again, but this time removing the AI troops XP cheat (just to check if this is the main reason because long peace is bad). Does someone know an easy way to remove it?
 

Sarissofoi

Veteran
M&BWBWF&S
95VcA2O.png
Control
Aserai
30​
Battania
9​
Khuzait
28​
NIMP
27​
SIMP
4​
Strugia
28​
Vlandia
29​
WIMP
16​

20 years test
Looks pretty balanced(outside SIMP and Battanians who have 2k and 3k troops respectively all other factions have 5.5k to 6.5k troops.
Its modded.
I lowered manually Tiers for all troops outside of Noble Line by one(2 down for mercenary line), which make easier to catch up in EXP for losing faction and its less financial drain also less prisoners and loot I guess.
Need further testing but for now it looks interesting, isn't it?

Also Jaculan is Vlandian territory its just their banner color didn't change for some reason.
 

Dabos37

Sergeant Knight
95VcA2O.png
Control
Aserai
30​
Battania
9​
Khuzait
28​
NIMP
27​
SIMP
4​
Strugia
28​
Vlandia
29​
WIMP
16​

20 years test
Looks pretty balanced(outside SIMP and Battanians who have 2k and 3k troops respectively all other factions have 5.5k to 6.5k troops.
Its modded.
I lowered manually Tiers for all troops outside of Noble Line by one(2 down for mercenary line), which make easier to catch up in EXP for losing faction and its less financial drain also less prisoners and loot I guess.
Need further testing but for now it looks interesting, isn't it?

Also Jaculan is Vlandian territory its just their banner color didn't change for some reason.

I think it is one of the reasons because snowballing is worse in 1.5.7 Before this patch, the AI was able to get high tier units insanely fast due to some perks giving too much XP. Now it has been balanced, and losing faction is having a harder time to get high tier units, while winning factions just keeps improving units and having elite armies.

Anyway, I am going to help you with this test but in my case, I am going to reduce upgrade XP cost for all tiers drastically, just to check if the map is more balanced at year 20.
 

Sarissofoi

Veteran
M&BWBWF&S
I will run 4 more tests tonight.So we will see if it was just fluke or not.
This make war loot less a factor but it also lowered influence/cash gain for ransoming.
Which mean that peace is less devastating as influx of income for winning side is less.
Still I think that the flat steal from enemy treasure after defeating lord should be limited. Its what a ransoming lord is for(or at last it should be.
Also I think player should get only few items(armor/weapons) as a lot (but higher worth per item) when rest should be converted into gold like in NPC lord case. Currently its too much annoying to get tons of trash items that are good only for selling and zero loot that is worth wearing.
Over all I am pretty interesting in digging it and see how it affected whole balance.
 

mexxico

Developer
I think it is one of the reasons because snowballing is worse in 1.5.7 Before this patch, the AI was able to get high tier units insanely fast due to some perks giving too much XP. Now it has been balanced, and losing faction is having a harder time to get high tier units, while winning factions just keeps improving units and having elite armies.

Anyway, I am going to help you with this test but in my case, I am going to reduce upgrade XP cost for all tiers drastically, just to check if the map is more balanced at year 20.

I still think if we have more samples from both 1.5.6 and 1.5.7 they will have similar results maybe even 1.5.7 can have better results. Currently there are very less average snowball score difference between 1.5.6 and 1.5.7. So I stopped saying snowballing is worse in 1.5.7.

By the way to prove any development's good effect you need to make so many tests (maybe 20), game map ends up in a state which is so random at each test run. So any less number of samples can mislead you. Even @Apocal's one 116 score test result increases average by 5 points only itself.

For example currently I am making some tests on 1.5.8 with some minor financial developments + minor campaign AI developments + improving mercenary hiring algorithm + settlements value calculation fix at war / peace AI. Results are 35, 39, 108, 26, 38 as you see even in 5 tests I have one 100+ result even other 4 are balanced. So results can differ so much, we have high deviation there. By luck you can hit 100+ score in a test. If you do not hit that 100+ you can feel that you have a score of 40 but with one 100+ test result your average can be 50+ again.

As I said previously I think having 50 average is ok and even it is fun to deal with this problem I cannot spend more time on this issue because so many things to do. At 1.5.6 and 1.5.7 I do not think snowballing is major problem anymore.

Increasing needed XP limits for tiers probably lower snowball score because all kingdoms will have more low tier troops and all mobile party & garrison wages will reduce and clans at kingdoms which lost territory will have less financial problems and they will defect less so when defection reduces snowball score also reduces. However I want to keep troop tier distribution at in average 20%-20%-20%-20%-20% and I do not want to have money inflation at game and I want to keep some clans having financial problems at late game (so player will find & recruit them for his kingdom) so that kind of solution will damage this distribution and create money inflation and player won't be able to find any poor clans. So actually my first aim is not only lowering snowball score but also keeping gameplay fun.

Thank you all for spending time on solving this problem. Please keep above informations on mind.
 
Last edited:

Sarissofoi

Veteran
M&BWBWF&S
TBH balancing now its kind of pointless anyway. Its good to have extra information on how what affect what etc but with adding new features to game and new options for AI or changing mechanics(to some extent) it will be need re balanced every time.
So I guess its fine now.
Still I hope that AI will need horses for upgrading cavalry soon.

Also I think its still pretty important to get some information with some radical changes in gameplay.
 

Dabos37

Sergeant Knight
I still think if we have more samples from both 1.5.6 and 1.5.7 they will have similar results maybe even 1.5.7 can have better results. Currently there are very less average snowball score difference between 1.5.6 and 1.5.7. So I stopped saying snowballing is worse in 1.5.7.

By the way to prove any development's good effect you need to make so many tests (maybe 20), game map ends up in a state which is so random at each test run. So any less number of samples can mislead you. Even @Apocal's one 116 score test result increases average by 5 points only itself.

For example currently I am making some tests on 1.5.8 with some minor financial developments + minor campaign AI developments + improving mercenary hiring algorithm + settlements value calculation fix at war / peace AI. Results are 35, 39, 108, 26, 38 as you see even in 5 tests I have one 100+ result even other 4 are balanced. So results can differ so much, we have high deviation there. By luck you can hit 100+ score in a test. If you do not hit that 100+ you can feel that you have a score of 40 but with one 100+ test result your average can be 50+ again.

As I said previously I think having 50 average is ok and even it is fun to deal with this problem I cannot spend more time on this issue because so many things to do. At 1.5.6 and 1.5.7 I do not think snowballing is major problem anymore.

Increasing needed XP limits for tiers probably lower snowball score because all kingdoms will have more low tier troops and all mobile party & garrison wages will reduce and clans at kingdoms which lost territory will have less financial problems and they will defect less so when defection reduces snowball score also reduces. However I want to keep troop tier distribution at in average 20%-20%-20%-20%-20% and I do not want to have money inflation at game and I want to keep some clans having financial problems at late game (so player will find & recruit them for his kingdom) so that kind of solution will damage this distribution and create money inflation and player won't be able to find any poor clans. So actually my first aim is not only lowering snowball score but also keeping gameplay fun.

Thank you all for spending time on solving this problem. Please keep above informations on mind.


Yes, you are right. I have made a fast test drastically reducing upgrade XP cost for all tiers and this is the result xD:


So yes, it looks like it is actually worse.

Concerning 1.5.8 tests, yes, they look pretty good and I do agree with snowballing is not a major problem anymore. Plus it is actually good for the game having some campaigns with snowballing to feel different (as long as it does not happen most of the time). Thanks for all the effort and work.
 

Blood Gryphon

Master Knight
WBVC
@mexxico good job 1.5.8 looks promising. Thanks for giving us the opportunity to help out, been another great thread

Looking forward to the new stuff you’re working on!
 

Sarissofoi

Veteran
M&BWBWF&S
So all 5 moded runs.
The one that reduce all common troops/bandits/caravan guards by one tier and mercenaries by 2.Nobles troops(and militia) stay the same.
This affect increased mortality for common troops, less money from loot and ransoms for them(same for bandits), but also cheaper upkeep and recruitment/upgrades for them. obviously Militia now fare better in Auto calculated sieges which help defender and rebels. Noble troops are the elite having higher chances in auto calculated combat now(both for win and for survival).
War loot have lesser effect on snowballing, and peace times are easier to survive.
What I noticed was that victorious side have less troops compared to non moded version and fluctuation in troop count(after 20 years) was much lower. So weaker faction usually have more troops(4-5k) than weaker faction in unmoded version(2-3k) and stronger factions have less troops(6-7k) compared to unmodded version(8-10k). The losing factions also usually keep more of their possession than losing factions in unmodded version.
Starting dateMod AMod BMod CMod DMod E
25​
Aserai
30​
23​
29​
26​
21​
18​
Battania
9​
27​
21​
25​
8​
21​
Khuzait
28​
41​
28​
30​
22​
21​
NIMP
27​
12​
20​
9​
19​
22​
SIMP
4​
11​
16​
24​
27​
22​
Strugia
28​
23​
21​
28​
17​
24​
Vlandia
29​
20​
18​
18​
47​
20​
WIMP
16​
13​
14​
12​
11​
rebels
2​
2​
nskH82n.png
Khuzaits snowball with far largest score(41) and 7,5k troops. Vlandia with only 20 score have the same amount of troops.Aserai and Battania both ahve around 6k troops. rest factions have from 4 to 5k troops.
8qAgbfW.png
Battanains look like they control plenty but they don't. Castles are still under Vladian rule.
Largest factions Aseri and Khuzaits have 7k troops. Same for Sturgia. Rest factions have from 3 to 5k troops.
4zzvtU3.png
Khuzaits are the largest faction but they have only 5.5k troops. For comparison Aserai, SIMP, NIMP have bigger armies(6,5k, 6k, 5,5k). rest factions have 3,5 to 4,5k troops. Strugia who is second largest(in control score) have only 4,5k troops.
Frh9PKp.png
The Vladains started snowballing. They not only have most settlements but also the biggest army 10k that I seen in this series of test.(Usually biggest faction gathered like 7k in this series). Next are SIMP and Aserai with 7 and 6k troops. Weakest faction Battanians still have 4k troops.

Overall interesting experiment.
 
Last edited:

Apocal

Master Knight
Test 4Test 5Test 6
Aserai222930
Battania524944
Khuzait524245
Northern Empire059
Southern Empire003
Sturgia6015
Vlandia91212
Western Empire323615
 

Pejot

Knight at Arms
WBVC
95VcA2O.png
Control
Aserai
30​
Battania
9​
Khuzait
28​
NIMP
27​
SIMP
4​
Strugia
28​
Vlandia
29​
WIMP
16​

20 years test
Looks pretty balanced(outside SIMP and Battanians who have 2k and 3k troops respectively all other factions have 5.5k to 6.5k troops.
Its modded.
I lowered manually Tiers for all troops outside of Noble Line by one(2 down for mercenary line), which make easier to catch up in EXP for losing faction and its less financial drain also less prisoners and loot I guess.
Need further testing but for now it looks interesting, isn't it?

Also Jaculan is Vlandian territory its just their banner color didn't change for some reason.

Maybe try increasing upgrade cos of troops especially cav since AI doesn't need horses to upgrade them. Footmen can be twice the price and horsemen twice the price + average required horse price.
 

Dabos37

Sergeant Knight
So all 5 moded runs.
The one that reduce all common troops/bandits/caravan guards by one tier and mercenaries by 2.Nobles troops(and militia) stay the same.
This affect increased mortality for common troops, less money from loot and ransoms for them(same for bandits), but also cheaper upkeep and recruitment/upgrades for them. obviously Militia now fare better in Auto calculated sieges which help defender and rebels. Noble troops are the elite having higher chances in auto calculated combat now(both for win and for survival).
War loot have lesser effect on snowballing, and peace times are easier to survive.
What I noticed was that victorious side have less troops compared to non moded version and fluctuation in troop count(after 20 years) was much lower. So weaker faction usually have more troops(4-5k) than weaker faction in unmoded version(2-3k) and stronger factions have less troops(6-7k) compared to unmodded version(8-10k). The losing factions also usually keep more of their possession than losing factions in unmodded version.
Starting dateMod AMod BMod CMod DMod E
25​
Aserai
30​
23​
29​
26​
21​
18​
Battania
9​
27​
21​
25​
8​
21​
Khuzait
28​
41​
28​
30​
22​
21​
NIMP
27​
12​
20​
9​
19​
22​
SIMP
4​
11​
16​
24​
27​
22​
Strugia
28​
23​
21​
28​
17​
24​
Vlandia
29​
20​
18​
18​
47​
20​
WIMP
16​
13​
14​
12​
11​
rebels
2​
2​
nskH82n.png
Khuzaits snowball with far largest score(41) and 7,5k troops. Vlandia with only 20 score have the same amount of troops.Aserai and Battania both ahve around 6k troops. rest factions have from 4 to 5k troops.
8qAgbfW.png
Battanains look like they control plenty but they don't. Castles are still under Vladian rule.
Largest factions Aseri and Khuzaits have 7k troops. Same for Sturgia. Rest factions have from 3 to 5k troops.
4zzvtU3.png
Khuzaits are the largest faction but they have only 5.5k troops. For comparison Aserai, SIMP, NIMP have bigger armies(6,5k, 6k, 5,5k). rest factions have 3,5 to 4,5k troops. Strugia who is second largest(in control score) have only 4,5k troops.
Frh9PKp.png
The Vladains started snowballing. They not only have most settlements but also the biggest army 10k that I seen in this series of test.(Usually biggest faction gathered like 7k in this series). Next are SIMP and Aserai with 7 and 6k troops. Weakest faction Battanians still have 4k troops.

Overall interesting experiment.

It is indeed really interesting, and while applying specifically this change for the game would not be a great idea, we can get some good conclusions from it. I know it is hard to please everyone, but I personally would prefer to see a higher passive income for AI, while battle loot being decreased.

Thanks for testing this mate.

BTW, what you did was just decreasing one tier for every unit in the npccharacters.xml right? If you do not mind, could you please share your file with me?
 
Last edited:

Pejot

Knight at Arms
WBVC

Probably troops templates needs to be edited. As for AVG horse price. I don't know how to make it dynamic based on prices so just add constant values to horseman upgrade cost. Something around 300 when normal horse is needed and 1k when war horse is needed.
 

Sarissofoi

Veteran
M&BWBWF&S
Probably troops templates needs to be edited. As for AVG horse price. I don't know how to make it dynamic based on prices so just add constant values to horseman upgrade cost. Something around 300 when normal horse is needed and 1k when war horse is needed.
I am asking how to make edit(preferably step by step) - not for the values.I have no idea how to do it or when to even look for it.
I will be grateful for instructions.
Also if you know how to increase/decrease upkeep level separate from unit Level it would be nice.

It is indeed really interesting, and while applying specifically this change for the game would not be a great idea, we can get some good conclusions from it. I know it is hard to please everyone, but I personally would prefer to see a higher passive income for AI, while battle loot being decreased.

Thanks for testing this mate.

BTW, what you did was just decreasing one tier for every unit in the npccharacters.xml right? If you do not mind, could you please share your file with me?

Yes.
SPNPCCharacters, Bandits and SPSpecialcharacters.
Lowering bandits Tiers probably also helped peasants because now they can face them head. I also lowered mercenary troops by 2 tiers because they are overvalued compared to their stats and equipment.
Here are files
 
Top Bottom