SP - Battles & Sieges The tactical way: beyond the melee cluster

[Ikea Knight]

Still one of my most favorite threads.
But I didn't have the time to play much recently and when I could I actually focues on Skirmish or Duels in the Multiplayer.
However, I do agree that the Ai hasn't evolved significantly enough.
I still don't see the merrit to most of formations, for example in Viking Conquest ordering the Shield Wall actually made the units with shields stay in the front.
Also the Ai has the ability to give units advanced commands like skirmishing for Archers.
I'd like to have that as an option too for players. You can of course micromanage that, but it can be quite the hassle.
Also while the AI commanders have options like "flank this" or "attack this". While I don't think that it currently works all that well in the game, I'd like to have to give these kind of commands to my units too.
Finally one of my thoughts on Cavalry is that it just splits up too easily.
I'd like the option to adjust the cavalrys speed so that the unit doesn't seperate itself because of the different maneuver and tempo stats in the game.
I am no historian, but I have heard that cavalry charges usually weren't in full galop but rather in canter, which seemed to more comfortable for the rider and enabled the different horses to stay closer together.

Other than that, Terco has made very good statements in his latest post and I very much agree with all of them.

 

[eddiemccandless]

More than a month with the game released I am about to update the thread with a bunch of personal feedback.

As we have all noticed I can already testify that there is something strange going on with AI in combat. In Dev Blog 25/10/18, everything that was mentioned was incredible on paper; however in the game it is not running satisfactorily.

In Bannerlord we have the option to divide troops in battle; a feature that I personally use often and I am very grateful for its implementation.


Infantry:

I've tried everything to carry out different tactics on the battlefield to test the limits of the AI; and I have to say that these are currently low.

I wonder, should a bot leave a formation and thereby ignore a player's order? My answer is a resounding no; the AI must stay within its movement capsule, not make any "tactical" decisions beyond self-protection. Well, this happens absolutely every time in game.

Why do the bots of a line formation which are waiting for an enemy unit to attack suddenly charge within a few meters to it without any charge order being given? This should not be so, or I understand that this should not be so.

For example, I've tried splitting an infantry unit in two by arranging each one in line manually forming a V (inverted wedge tactic) to try to encircle the enemy. As I say, before receiving enemy contact, my bots charge against them without my permission. The most worrying thing is that not all of them are charging, the ones that are farthest from the V-axis do it by intersecting and the ones that are farthest from the V-axis remain motionless without knowing what to do.

When all this escalates badly, is when I decide to split an infantry unit into 3. Three groups in order to "advance" in line formation. When making contact with the enemy, the central unit acts correctly by holding the enemy's attack, however the side units tend to cluster towards the center of the enemy unit. When the two side units are manually signaled their flank position (U arms) they without charging or advancing order ordered charge against the enemy bot units positioned at the sides.

It is absolutely necessary to implement the order (recover from warband) to use weapons. Another case in which the AI decides "too much"; it is the player who should assign how, when and where.

I still think that the passive behavior programmed for the AI in shieldwall formation is not appropriate. If you want to continue with this kind of behavior for Ai, implement width commands (bring back the F8 - Tighter formation/ F9 - Looser formation command) and then be able to articulate high density ranks as an analog shieldwall where the bots do remain aggressive.

Ranged:

In general, they do their job well, but there's a problem that still exists Issues with archery's ai in close formation. It is not always possible to deploy to a loose formation (where all the bots fire normally); therefore there are still formations where the bots in the secondary rows do not fire, thus losing tactical effectiveness. For example, I have an infantry unit in a circle formation and a ranged unit within it.

If I have them in line they shoot the enemy and rotate, unless it is a unit composed of a large number of bots, which leave the circle of protection provided by the infantry. I lose tactical effectiveness.

If I place them in the loose-column, they automatically go outside the limits of infantry protection. I lose tactical effectiveness.

If I set them in shieldwall-wedge, because of the problem mentioned above; the bots in the secondary rows do not fire. I lose tactical effectiveness.

If I arrange them in a circle-square, the only ones that fire are the bots that have an enemy in their focal field. I lose tactical effectiveness.

Cavalry:

Without a weapons use command, this unit in SP does not unleash its full potential. Charges to enemies from behind (anvil and hammer) or to those who retreat are not executed with the forcefulness that this procedure requires.

An example of the shortcomings of the AI. A cavalry unit split in two prepares to face an enemy infantry unit. The cavalry units are placed in a wedge formation at a considerable distance from the enemy (A1---------E------------A2). A1 is then ordered to charge and after a lapse of 5 seconds A2 is ordered the same. Therefore A1 will charge E first and keep it focussed on them while A2 charges from behind. This maneuver should be devastating and surprisingly it is not. Replace A1 with an infantry unit and the result remains the same. Enemy infantry hardly suffers any heavy casualties.

Mounted Ranged:

I like the way the main behavior of this type of unit is programmed. I like that they form a rotating circle by keeping the enemy inside while firing at it. What I don't like is that once the projectiles run out, the AI decides to charge on their own. I think they should be programmed to keep circling with their comrades until they all run out of ammunition. Once this happens, it would be more logical for the formation to look for the player as a "follow me". Assuming that the player has been killed, then yes, carry out the charge

We are back to the point of the imperative need to implement the use of weapons order. For this type of unit it is essential to.

I think it would be interesting to implement a harassment/skirmish order or function to any mounted projectile unit. Just as infantry units equipped with javelins throw them before close combat, so should cavalry. For cavalry equipped with a bow I like the currently programmed behavior of circling around the enemy. However, it should also be possible to have an order to execute circles without keeping the enemy inside. I'm talking about cantabrian circles or manoeuvres in a encircling way like the caracole.

Hot spots:

A statement to be set in stone: The user always decides what, when, where and how in battle command. To make this impossible, impoverishes the player's playable experience.

In short, the bot as an individual must always be subordinated to the structure of the formation. The bot will only leave the formation when it is annihilated, flees or is ordered to retreat.

A fighting bot will always remain aggressive. It is the player who will always have to modify its behavior by positioning with respect to an enemy unit.

A bot on foot should not run unless ordered to charge. The user is the one who decides the pace. Engage Enemy Order (Alternative replacing Advance order)

The follow command keeps too much space between the player and the following bots. I think that this space should be greatly reduced.

The fallback command still doesn't work. The bots still turn their backs to the enemy when this command is executed. This should not be the case under any circumstances. The bots should always fall back facing the enemy. Fallback order does not work properly

Implementing an order to use weapons is vital. Battle Command: Switch Weapon

Implementing a system to " focus fire/ focus attack " should have been considered as a natural evolution for the franchise. Focus fire on enemy unit

The colliders of each bot should be reviewed. The living space of each bot should be increased as we can see today that the bodies of the bot suffer a totally disproportionate overlap. Agent overlap issue in flock movement

Hitboxes and hurtboxes should be checked. A lot of inconsistences with shields Body parts hitboxes leave a lot to be desired / Shield hitboxes are extremely imprecise / Inconsistency damage & hitboxes with overhead guard.

At the moment the game does not offer anything beyond the melee cluster.

AI definitely still has problems. I am wondering though how much of this can be blamed on the wonky physics. Switching to a physics based combat seems to be one of the main features of Bannerlord, as stated by Armagan himself in some interviews. Yet I can't help but feel that there is something wrong with the way the physics of combat work at the moment, with units being able to slide through thin gaps in a formation (which is even more weird when that happens with cavalry on a field battle). I feel that it might be difficult (and perhaps pointless) to properly fix the AI if the underlying basic combat is not working right. Am I alone in noticing this?

 

[abc123456]

AI definitely still has problems. I am wondering though how much of this can be blamed on the wonky physics. Switching to a physics based combat seems to be one of the main features of Bannerlord, as stated by Armagan himself in some interviews. Yet I can't help but feel that there is something wrong with the way the physics of combat work at the moment, with units being able to slide through thin gaps in a formation (which is even more weird when that happens with cavalry on a field battle). I feel that it might be difficult (and perhaps pointless) to properly fix the AI if the underlying basic combat is not working right. Am I alone in noticing this?

I have said it before and I will say it again. Machine learning will fix a lot of the AI problems. TW got money now they can afford some machine learning specliasts.

 

[Terco_Viejo]

@Ikea Knight much appreciated my boi

[...] Am I alone in noticing this?

Not at all, I've noticed it too. Honestly, I've been noticing this since the captain mode was released in the alpha/ open-closed beta period before the EA release. We found practically the same problems in SP since then that "seem" to have been fixed. Far from it.

Then, not less important, the combat system for both the player and the bots should be revised. This one is not in as bad a position as the MP may be; however it also has flaws and inconsistencies.

Regarding the Ai, I would like to assume that the devs (Cem Çimenbiçer aka @Daegoth & partners) have an ace up their sleeve. Is it plausible to think that Ai is currently capped for performance reasons? Could the automated calculations/algorithms that become from Ai (considering the large number of agents in scene) in all its potential constitute actually a performance problem?

 

[eddiemccandless]

@Ikea Knight much appreciated my boi


Not at all, I've noticed it too. Honestly, I've been noticing this since the captain mode was released in the alpha/ open-closed beta period before the EA release. We found practically the same problems in SP since then that "seem" to have been fixed. Far from it.

Then, not less important, the combat system for both the player and the bots should be revised. This one is not in as bad a position as the MP may be; however it also has flaws and inconsistencies.

Regarding the Ai, I would like to assume that the devs (Cem Çimenbiçer aka @Daegoth & partners) have an ace up their sleeve. Is it plausible to think that Ai is currently capped for performance reasons? Could the automated calculations/algorithms that become from Ai (considering the large number of agents in scene) in all its potential constitute actually a performance problem?

Right, I suspected it was the same issue that had been noticed by people during the beta. I think it is the same issue for MP and SP, if the problem is in the physics it stands to reason that it's going to affect both.

@abc123456 , machine learning techniques are indeed a powerful tool, but it's not the solution to every problem (I am a robotics engineer and work as a postdoc in academia, I am by no means a specialist but it is in my general area of expertise). It is also very easy to misuse it and not always easy to understand that that is happening. I would say that it's overkill for this situation, especially since AI in Warband seemed perfectly competent to me. I do hope that @Terco_Viejo is correct, I did read somewhere that they had at some point a very competent AI for spears that was scrapped because it was in fact too good.

 

[Johnmike13]

More than a month with the game released I am about to update the thread with a bunch of personal feedback.

As we have all noticed I can already testify that there is something strange going on with AI in combat. In Dev Blog 25/10/18, everything that was mentioned was incredible on paper; however in the game it is not running satisfactorily.

In Bannerlord we have the option to divide troops in battle; a feature that I personally use often and I am very grateful for its implementation.


Infantry:

I've tried everything to carry out different tactics on the battlefield to test the limits of the AI; and I have to say that these are currently low.

I wonder, should a bot leave a formation and thereby ignore a player's order? My answer is a resounding no; the AI must stay within its movement capsule, not make any "tactical" decisions beyond self-protection. Well, this happens absolutely every time in game.

Why do the bots of a line formation which are waiting for an enemy unit to attack suddenly charge within a few meters to it without any charge order being given? This should not be so, or I understand that this should not be so.

For example, I've tried splitting an infantry unit in two by arranging each one in line manually forming a V (inverted wedge tactic) to try to encircle the enemy. As I say, before receiving enemy contact, my bots charge against them without my permission. The most worrying thing is that not all of them are charging, the ones that are farthest from the V-axis do it by intersecting and the ones that are farthest from the V-axis remain motionless without knowing what to do.

When all this escalates badly, is when I decide to split an infantry unit into 3. Three groups in order to "advance" in line formation. When making contact with the enemy, the central unit acts correctly by holding the enemy's attack, however the side units tend to cluster towards the center of the enemy unit. When the two side units are manually signaled their flank position (U arms) they without charging or advancing order ordered charge against the enemy bot units positioned at the sides.

It is absolutely necessary to implement the order (recover from warband) to use weapons. Another case in which the AI decides "too much"; it is the player who should assign how, when and where.

I still think that the passive behavior programmed for the AI in shieldwall formation is not appropriate. If you want to continue with this kind of behavior for Ai, implement width commands (bring back the F8 - Tighter formation/ F9 - Looser formation command) and then be able to articulate high density ranks as an analog shieldwall where the bots do remain aggressive.

Ranged:

In general, they do their job well, but there's a problem that still exists Issues with archery's ai in close formation. It is not always possible to deploy to a loose formation (where all the bots fire normally); therefore there are still formations where the bots in the secondary rows do not fire, thus losing tactical effectiveness. For example, I have an infantry unit in a circle formation and a ranged unit within it.

If I have them in line they shoot the enemy and rotate, unless it is a unit composed of a large number of bots, which leave the circle of protection provided by the infantry. I lose tactical effectiveness.

If I place them in the loose-column, they automatically go outside the limits of infantry protection. I lose tactical effectiveness.

If I set them in shieldwall-wedge, because of the problem mentioned above; the bots in the secondary rows do not fire. I lose tactical effectiveness.

If I arrange them in a circle-square, the only ones that fire are the bots that have an enemy in their focal field. I lose tactical effectiveness.

Cavalry:

Without a weapons use command, this unit in SP does not unleash its full potential. Charges to enemies from behind (anvil and hammer) or to those who retreat are not executed with the forcefulness that this procedure requires.

An example of the shortcomings of the AI. A cavalry unit split in two prepares to face an enemy infantry unit. The cavalry units are placed in a wedge formation at a considerable distance from the enemy (A1---------E------------A2). A1 is then ordered to charge and after a lapse of 5 seconds A2 is ordered the same. Therefore A1 will charge E first and keep it focussed on them while A2 charges from behind. This maneuver should be devastating and surprisingly it is not. Replace A1 with an infantry unit and the result remains the same. Enemy infantry hardly suffers any heavy casualties.

Mounted Ranged:

I like the way the main behavior of this type of unit is programmed. I like that they form a rotating circle by keeping the enemy inside while firing at it. What I don't like is that once the projectiles run out, the AI decides to charge on their own. I think they should be programmed to keep circling with their comrades until they all run out of ammunition. Once this happens, it would be more logical for the formation to look for the player as a "follow me". Assuming that the player has been killed, then yes, carry out the charge

We are back to the point of the imperative need to implement the use of weapons order. For this type of unit it is essential to.

I think it would be interesting to implement a harassment/skirmish order or function to any mounted projectile unit. Just as infantry units equipped with javelins throw them before close combat, so should cavalry. For cavalry equipped with a bow I like the currently programmed behavior of circling around the enemy. However, it should also be possible to have an order to execute circles without keeping the enemy inside. I'm talking about cantabrian circles or manoeuvres in a encircling way like the caracole.

Hot spots:

A statement to be set in stone: The user always decides what, when, where and how in battle command. To make this impossible, impoverishes the player's playable experience.

In short, the bot as an individual must always be subordinated to the structure of the formation. The bot will only leave the formation when it is annihilated, flees or is ordered to retreat.

A fighting bot will always remain aggressive. It is the player who will always have to modify its behavior by positioning with respect to an enemy unit.

A bot on foot should not run unless ordered to charge. The user is the one who decides the pace. Engage Enemy Order (Alternative replacing Advance order)

The follow command keeps too much space between the player and the following bots. I think that this space should be greatly reduced.

The fallback command still doesn't work. The bots still turn their backs to the enemy when this command is executed. This should not be the case under any circumstances. The bots should always fall back facing the enemy. Fallback order does not work properly

Implementing an order to use weapons is vital. Battle Command: Switch Weapon

Implementing a system to " focus fire/ focus attack " should have been considered as a natural evolution for the franchise. Focus fire on enemy unit

The colliders of each bot should be reviewed. The living space of each bot should be increased as we can see today that the bodies of the bot suffer a totally disproportionate overlap. Agent overlap issue in flock movement

Hitboxes and hurtboxes should be checked. A lot of inconsistences with shields Body parts hitboxes leave a lot to be desired / Shield hitboxes are extremely imprecise / Inconsistency damage & hitboxes with overhead guard.

At the moment the game does not offer anything beyond the melee cluster.

Wow this was well put!

 

[Meko]

sieges are the worst when you join someone else's army. You lose all control of your troops and they just stupid charge to their deaths like lemmings.

I take on 3 enemy lords combining for 180 against my rag tag group of half high tier troops and half recruited prisoners of 170 and I'll lose maybe 0 to 4 and 2 or 3 wounded from using draw out tactics, shield wall pinches and strafing runs with horse archers..

In a siege though... 56 dead 30 wounded...
You have no troops to control.

 

[abc123456]

@abc123456 , machine learning techniques are indeed a powerful tool, but it's not the solution to every problem (I am a robotics engineer and work as a postdoc in academia, I am by no means a specialist but it is in my general area of expertise). It is also very easy to misuse it and not always easy to understand that that is happening. I would say that it's overkill for this situation, especially since AI in Warband seemed perfectly competent to me. I do hope that @Terco_Viejo is correct, I did read somewhere that they had at some point a very competent AI for spears that was scrapped because it was in fact too good.

I will diverge a bit from OP to include campaign AI as well. I think a myriad of AI issues across the board (campaign, battlefield) can be solved without traditional programming techniques.

Bannerlord has a set of well-defined answers and data, we can definitely train models to solve a range of problems in the battlefield AI and campaign AI.

TaleWorlds want the AI to play by the same rules as the player yet maintain a satisfactorily difficult and clever AI. This is not achievable with traditional programming - the code and effort will be enormous with a lot of conditional (if/else) statements.

Yes, it is a lot of effort to use machine learning in Bannerlord, but I believe it is the only way to achieve the results TaleWorld wants, otherwise, they will end up creating a cheating AI to create artificial difficulty. There is a reason why most games out there have dumb AI or cheating AI.

Have a look at some amazing things people have done with deep learning in video games.

AI learns unique tactics in hide and seek


How Ubisoft use AI, pretty cool


Smartbot learning new strategy to beat normal bot

 

[TheShermanator]

Apologies if this is a dumb question, but: What exactly does the infantry's Advance (F1-F4) do in the context of this conversation about the issues with F1-F3?

I agree with the shortcomings of the F1-F3 command. But the existence of F1-F4 makes me wonder if the devs intended for that command to achieve what we are describing - to prompt our contingent to keep formation while attacking.

That said, I've not actually had good luck with that using F1-F4. E.g. If I have infantry in shield wall and Advance them, they do seem to keep their shields up, but the formation becomes ragged in the way described on this thread.

 

[Terco_Viejo]

I have been fiddling with the game parameters and based on this I will frame this new comment as an analysis.

Structural integrity of the formation​

I have been testing a number of formations and have observed two behavioural patterns in the bots. Given a formation command plus a hold position command (stop), we can see that in line, loose and skein formations the bots are taking the initiative to leave the formation in the last few meters and charge on their own, significantly altering the structural integrity of the formation. In contrast, in shieldwall, square and circle the integrity of the formation remains unaltered by an enemy charge.

​

My question is, why does this happen? This behaviour is supposed to be hardcoded and therefore only Taleworlds will have access. In my opinion, that parameter/code that makes the bot not to charge when in the shieldwall, square and circle formations should be incorporated for line, loose and skein behaviours.
(*) I've ignored column and scatter because they are buggy and have a bad performance with high volumes of agents on screen (eventually incorporated too).

Vital space​

In this thread and especially in this one, the problem of the " vital space " of the agents has been commented widely. In the default game we can observe an exacerbated overlap of agents when there is a high number of agents in the melee. This is due to the individual living space and the living space once inside the formation. The parameters that control all this are "body capsule radius" and "Bipedal Radius"; both modified for testing purposes. Here you have a comparative image (The space between rows in the default is too crowded) and a video that tries to simulate the final battle moment of the 2016 siege defence video .

Default	Modified
BipedalRadius" value="0.38" body_capsule radius="0.37"​	BipedalRadius" value="0.50" body_capsule radius="0.41"​

* Despite the slight tweaking of values, there is still jittering, slight but still appreciable.

Battle pace​

Many of the SP players (among whom I am one) have noticed that the pace of battles is too fast, resulting in instantaneous meat grinders. This kind of "fast paced combat" fits more or less in with the game philosophy in the MP (Captain Mode) section; however in SP I would personally lift the foot off the throttle.

One of the first symptoms of over-accelerated pace is the base speed of each agent, both charging and running. In my opinion an infantry charge should be carried out with a light trot, staying away from "speedy gonzales" sprints. I dare say that practically every M&B fan will have seen the film Bravehearth.

What a rush!... and then fight... oof
​

Well, this type of running from so far away is fantasy, the infantry charges in periods before the gunpowder were carried out with a walking approach and in the last 50 meters charging so as not to arrive exhausted to the melee. Here you have a real video (plausible). In my opinion a more realistic jogging ( considering the weight of the equipment + the "virtual stamina" of each agent), base and that "the difference" was determined by the points of athletics skills and with their multipliers.

A comparison of values:

Default	Modified
"bipedal_sprint_speed_ratio" value="0.37" "bipedal_sprint_time_ratio" value="0.9" "bipedal_speed_multiplier" value="6.2"​	"bipedal_sprint_speed_ratio" value="0.10" "bipedal_sprint_time_ratio" value="0.7" "bipedal_speed_multiplier" value="5.0"​

Here you have a comparative (Default-Modified) in video form:

Modified starts 00:13​

On the other hand, and as a result of this comment by @Signalize, we know that a change in the values of the armour is going to take place or is currently being worked on. It is plausible to think that these modifications will bring about a significant change by giving the armour the nature of "durability" and protection thant the game needs.

That said, another parameter I have been messing around with is "hit points" and "absorbed damage ratio".

Default	Modified
hit_points="100" absorbed_damage_ratio="1.0"​	hit_points="150" absorbed_damage_ratio="0.75"​

As a result of their modification along with the alteration of the speed parameter mentioned above, we see a significant increase in battle timing. In a 500vs500 equal to equal scenario, we can measure duration times that are around 50 seconds - 1 minute for the default values and between 2 minutes to 2.5 minutes for the modified values.

Modified starts 01:25​

Therefore, if all formations have a structural integrity where each agent has its living space, the speed of each agent is slightly reduced and the health and damage absorption values are slightly modified; could this improve the playable experience?

I have previously quoted Signalize and if you guys (@Dejan @Duh_TaleWorlds @Daegoth ) have the time and energy to shed some light on this issue, we would all appreciate the information.

 

[Ikea Knight]

As per usual a quality post. Making the formations more controlable and battles longer seems to be the right direction for the singleplayer.

I don't mind slowing down the engagement speed if that time can be used to make strategical decisions on every units position.

 

[Terco_Viejo]

While writing that comment I didn't realize that the modding tools were released. We'll see what it has to offer.

 

[Blood Gryphon]

I have been fiddling with the game parameters and based on this I will frame this new comment as an analysis.

Structural integrity of the formation​

I have been testing a number of formations and have observed two behavioural patterns in the bots. Given a formation command plus a hold position command (stop), we can see that in line, loose and skein formations the bots are taking the initiative to leave the formation in the last few meters and charge on their own, significantly altering the structural integrity of the formation. In contrast, in shieldwall, square and circle the integrity of the formation remains unaltered by an enemy charge.

​

My question is, why does this happen? This behaviour is supposed to be hardcoded and therefore only Taleworlds will have access. In my opinion, that parameter/code that makes the bot not to charge when in the shieldwall, square and circle formations should be incorporated for line, loose and skein behaviours.
(*) I've ignored column and scatter because they are buggy and have a bad performance with high volumes of agents on screen (eventually incorporated too).

Vital space​

In this thread and especially in this one, the problem of the " vital space " of the agents has been commented widely. In the default game we can observe an exacerbated overlap of agents when there is a high number of agents in the melee. This is due to the individual living space and the living space once inside the formation. The parameters that control all this are "body capsule radius" and "Bipedal Radius"; both modified for testing purposes. Here you have a comparative image (The space between rows in the default is too crowded) and a video that tries to simulate the final battle moment of the 2016 siege defence video .

Default	Modified
BipedalRadius" value="0.38" body_capsule radius="0.37"​	BipedalRadius" value="0.50" body_capsule radius="0.41"​

* Despite the slight tweaking of values, there is still jittering, slight but still appreciable.

Battle pace​

Many of the SP players (among whom I am one) have noticed that the pace of battles is too fast, resulting in instantaneous meat grinders. This kind of "fast paced combat" fits more or less in with the game philosophy in the MP (Captain Mode) section; however in SP I would personally lift the foot off the throttle.

One of the first symptoms of over-accelerated pace is the base speed of each agent, both charging and running. In my opinion an infantry charge should be carried out with a light trot, staying away from "speedy gonzales" sprints. I dare say that practically every M&B fan will have seen the film Bravehearth.

What a rush!... and then fight... oof
​

Well, this type of running from so far away is fantasy, the infantry charges in periods before the gunpowder were carried out with a walking approach and in the last 50 meters charging so as not to arrive exhausted to the melee. Here you have a real video (plausible). In my opinion a more realistic jogging ( considering the weight of the equipment + the "virtual stamina" of each agent), base and that "the difference" was determined by the points of athletics skills and with their multipliers.

A comparison of values:

Default	Modified
"bipedal_sprint_speed_ratio" value="0.37" "bipedal_sprint_time_ratio" value="0.9" "bipedal_speed_multiplier" value="6.2"​	"bipedal_sprint_speed_ratio" value="0.10" "bipedal_sprint_time_ratio" value="0.7" "bipedal_speed_multiplier" value="5.0"​

Here you have a comparative (Default-Modified) in video form:

Modified starts 00:13​

On the other hand, and as a result of this comment by @Signalize, we know that a change in the values of the armour is going to take place or is currently being worked on. It is plausible to think that these modifications will bring about a significant change by giving the armour the nature of "durability" and protection thant the game needs.

That said, another parameter I have been messing around with is "hit points" and "absorbed damage ratio".

Default	Modified
hit_points="100" absorbed_damage_ratio="1.0"​	hit_points="150" absorbed_damage_ratio="0.75"​

As a result of their modification along with the alteration of the speed parameter mentioned above, we see a significant increase in battle timing. In a 500vs500 equal to equal scenario, we can measure duration times that are around 50 seconds - 1 minute for the default values and between 2 minutes to 2.5 minutes for the modified values.

Modified starts 01:25​

Therefore, if all formations have a structural integrity where each agent has its living space, the speed of each agent is slightly reduced and the health and damage absorption values are slightly modified; could this improve the playable experience?

I have previously quoted Signalize and if you guys (@Dejan @Duh_TaleWorlds @Daegoth ) have the time and energy to shed some light on this issue, we would all appreciate the information.

Great post! The results look promising and I think would improve the player experience. +1

 

[Count Delinard]

Awesome post, Terco

 

[Rycon Caldestan]

I have to agree @Terco_Viejo , and to parrot others, great post as usual! I have mostly refrained from modified games within bannerlord but will likely be adjusting that plan with the release of mod tools. If changes you've laid out aren't implemented in Vanilla, it will absolutely be something I seek out from the modding community.

 

[Terco_Viejo]

Thanks guys, I'm doing the best I can; I think that with the videos + images I can show this kind of "problems" much better than I would by describing them in a paragraph.

Certainly right now we are all excited about the recent release of the modding tools but I have mixed feelings.
On the one hand, I'm sure we'll see complex mods that alleviate the actual shortcomings of the base game. However, on the other hand, I fear that Taleworlds will relax and be satisfied with what the base game currently offers; in my opinion much less than what could be required of this new release.

 

[Terco_Viejo]

This whole thread focuses on the AI's behaviour when operating as infantry; however the cavalry would need a review as well.

One of the first "eyebrow raises" I did, was because of this gif belonging to the Dev Blog 14/02/18.

Here the charge seemed much more powerful than it is currently manifested in the Native
​

In the discussion thread, I wrote a comment indicating that there was a confluence of the cavalry charge at the central point of the group being charged. Obviously, at that time we didn't have access to the game and therefore it was all conjecture.

Well, just as the infantry is prone to charge at the central point of the group being charged; the cavalry in its absence follows the same pattern.

Imo a cavalry charge should charge offensively parallel to the defensive line of an enemy formation as indicated by the colour green. This mid-point charge attempt should only be achieved by a wedge charge (maintaining the shape of the formation and a certain parallel). As with infantry, when a charge is ordered, the formation loses its integrity (what use are the formations in such a case?).

On the other hand, what is a cavalry charge?

Speed and momentum of a real cavalry charge demonstration
​

In theory it is a manoeuvre consisting of a massive, impetuous and resolute attack by a troop against the enemy troop. Well, as the charges are defined in the Native, we do not find "the massive" or "the impetuous" but rather "a walking in the park". In my opinion the cavalry charges are extremely slow and without high impact.

Once again I have been fiddling with the parameters and have altered a few resulting in the following:

Native	Modified

An increase in HP, speed and horse charge damage (for the test I have taken the Imperial charger breed) (* increased rein rotation too). In contrast, an increase in damage multiplier for the animal, an increase in fall damage for the rider and a more aggressive setting for the rear.This results in a considerable casualty impact for the object of charge, but also in casualties for the horses when they charge forward into a wall of spears. However, charging from the rear or the flanks to an enemy formation with these modifications would be lethal.

Native	Modified
hit_points="100" absorbed_damage_ratio="1.0" body_capsule radius="0.37" bipedal_delta_rotation_threshold_for_keeping_local_velocity" value="2.5" QuadrupedalRadius" value="0.8" MakesRearAttackDamageThreshold" value="10.0" FallSpeedReductionMultiplierForRiderDamage" value="0.77" speed="44" charge_damage="12"	hit_points="400" absorbed_damage_ratio="5.0" body_capsule radius="0.40" bipedal_delta_rotation_threshold_for_keeping_local_velocity" value="2.5" QuadrupedalRadius" value="1.0" MakesRearAttackDamageThreshold" value="0.1" FallSpeedReductionMultiplierForRiderDamage" value="0.95" speed="80" charge_damage="90"

As is also the case with infantry, the cavalry is slightly crowded; therefore the spacing in the formations has been slightly increased. As @Apocal comments in his thread, Taleworlds should consider the spawns grouped together in a formation with a large volume of agents when they are carried out in narrow environments. Otherwise, the typical "cliff suicide" accident occurs.

These changes for the cavalry would make it a devastating weapon against the enemy with a strategic approach, but with "carefree" employment would lead to a resounding decimation of allied troops. Making horses a valuable war asset.

 

[Blood Gryphon]

This whole thread focuses on the AI's behaviour when operating as infantry; however the cavalry would need a review as well.

One of the first "eyebrow raises" I did, was because of this gif belonging to the Dev Blog 14/02/18.

Here the charge seemed much more powerful than it is currently manifested in the Native
​

In the discussion thread, I wrote a comment indicating that there was a confluence of the cavalry charge at the central point of the group being charged. Obviously, at that time we didn't have access to the game and therefore it was all conjecture.

Well, just as the infantry is prone to charge at the central point of the group being charged; the cavalry in its absence follows the same pattern.

Imo a cavalry charge should charge offensively parallel to the defensive line of an enemy formation as indicated by the colour green. This mid-point charge attempt should only be achieved by a wedge charge (maintaining the shape of the formation and a certain parallel). As with infantry, when a charge is ordered, the formation loses its integrity (what use are the formations in such a case?).

On the other hand, what is a cavalry charge?

Speed and momentum of a real cavalry charge demonstration
​

In theory it is a manoeuvre consisting of a massive, impetuous and resolute attack by a troop against the enemy troop. Well, as the charges are defined in the Native, we do not find "the massive" or "the impetuous" but rather "a walking in the park". In my opinion the cavalry charges are extremely slow and without high impact.

Once again I have been fiddling with the parameters and have altered a few resulting in the following:

Native	Modified

An increase in HP, speed and horse charge damage (for the test I have taken the Imperial charger breed) (* increased rein rotation too). In contrast, an increase in damage multiplier for the animal, an increase in fall damage for the rider and a more aggressive setting for the rear.This results in a considerable casualty impact for the object of charge, but also in casualties for the horses when they charge forward into a wall of spears. However, charging from the rear or the flanks to an enemy formation with these modifications would be lethal.

Native	Modified
hit_points="100" absorbed_damage_ratio="1.0" body_capsule radius="0.37" bipedal_delta_rotation_threshold_for_keeping_local_velocity" value="2.5" QuadrupedalRadius" value="0.8" MakesRearAttackDamageThreshold" value="10.0" FallSpeedReductionMultiplierForRiderDamage" value="0.77" speed="44" charge_damage="12"	hit_points="400" absorbed_damage_ratio="5.0" body_capsule radius="0.40" bipedal_delta_rotation_threshold_for_keeping_local_velocity" value="2.5" QuadrupedalRadius" value="1.0" MakesRearAttackDamageThreshold" value="0.1" FallSpeedReductionMultiplierForRiderDamage" value="0.95" speed="80" charge_damage="90"

As is also the case with infantry, the cavalry is slightly crowded; therefore the spacing in the formations has been slightly increased. As @Apocal comments in his thread, Taleworlds should consider the spawns grouped together in a formation with a large volume of agents when they are carried out in narrow environments. Otherwise, the typical "cliff suicide" accident occurs.

These changes for the cavalry would make it a devastating weapon against the enemy with a strategic approach, but with "carefree" employment would lead to a resounding decimation of allied troops. Making horses a valuable war asset.

Your modified version looks amazing once again. Hopefully the devs will try all of these parameters out and at least consider it.

@Terco_Viejo your work is

 

[Terco_Viejo]

Your modified version looks amazing once again. Hopefully the devs will try all of these parameters out and at least consider it.

@Terco_Viejo your work is

Thanks my boi, these are just experiments that I put at the disposal of the general interest.

 

[Shun]

I love the version in the second gif. For the game, I think the charging cavalry needs to take more dmg from the charge itself, from damage to the horses, riders getting thrown off, etc especially if the enemy formation is more han a few lines deep.