Cannon Fire v Musket Fire

Odd event yesterday. An Austrian regiment was marching in column towards a French regiment drawn up in line. The line of march was along a stream and on the other side of the stream was a French 8 lb battery at about 300 yards just tearing the hell out of the Austrians with canister. None the less, they kept marching, but as soon as they got into musket range the French started firing, inflicted 3 casualties and THEN the Austrians routed immediately.

Is there some inherent difference in the morale impact of musket fire as opposed to artillery? As I think on it I have seen a lot of units stand under withering artillery fire when the more reasonable course of action would have been to get out of Dodge. With infantry where line faces line they seem to bang away for quite a while but when a column is advancing on a line they don't seem to have the same "staying power."
Is it just me seeing things that aren't there?

That attack column was probably on the verge of retreating due to low morale from all the arty casualties it took. Once in a fire fight with the infantry its morale dropped a little more and failed its morale check.

In general, assault columns will almost always lose a firefight with a battalion in line formation as it can't bring enough firepower to bear. The assault column's main purpose is to close with the enemy and threaten him with the bayonet. If it is stopped by enemy fire, it is usually doomed.

As regards long range artillery fire hitting units that are standing in reserve, infantry will simply stand and take this for literally hours (as they historically did) while cavalry will always manoeuvre to safer ground (that is further away), again this is historically what cavalry commanders tended to do. At Waterloo when the British cavalry was in reserve and cannon balls kept hitting the men the officers ordered the troopers to dismount and this saved a great number of casualties. We can't represent this in the game exactly but having the cavalry withdraw a little so the incoming fire is less effective is really the same result.

If you TC cavalry they don't do this, its a new AI response. The AI response with infantry is to stand and take it so if the mounting losses bother you (as a Nap commander they really shouldn't!) then you can order your brigades elsewhere.

Doubtless that's what happened with the column but it really looked strange because of the trail of bodies behind them but they kept going until someone fired a musket at them! There is the advantage of using skirmishers ahead of a column but the AI never seems to do that. Then again, I have never noted the AI being terribly adept at combined arms anyway. The infantry attacks, the artillery sets up at maximum range and the cavalry moves up behind the infantry but they never seem to move forward until the infantry starts to break.

When I attack I like to have an infantry column following a skirmisher screen with light cavalry on one flank and an art. a cheval btty on the other and the rest of the brigade following up. The AI usually pulls out.

This is not a complaint. Just an observation. I have played some games with really shitty AIs and SOW is far better than most.

We are working towards releasing a new version of the mod and one of the biggest changes is AI behaviour. Their combined arms tactics (at least cav working with inf) is now merciless. We played a game yesterday as 4 players vs the AI and in the division where I commanded a brigade two players were completely humiliated by a huge well organised AI attack. There is still a weakness in how the AI uses artillery; it never comes forward with its attacking troops, something we still have to work on.

It's true the AI is not very adept, but the latest version of the dll we are working on greatly improves its play.  For example, infantry and cavalry will work together to break squares and then trample the fleeing enemy.

With regards to the artillery, I find that much of the time the AI does a better job placing the guns than do humans.  It tends to find locations with good fields of view where they make life very uncomfortable for the enemy.  When the AI supports them with infantry or cavalry, they can become nearly impregnable positions. It greatest weakness is that it doesn't move forward very often when its infantry gains ground and the battle moves to places that are out of range.

That sounds really impressive, guys. Right now it is just a bit too easy to pick a position astride a road, set up a battery with infantry protecting it and blast them as they come up the road virtually saying "kill me, kill me."

The AI also works better in defence than in attack. Attacking, as you've noticed, the AI can stay on a road in column much too close to an enemy and the resulting massacre of the road column is not much of a challenge.

We haven't yet tackled this aspect but we've taken a preliminary look at it.

There was a game some years ago called Napoleon in Italy. When it switched from the strategic map to the battle map the AI promptly ran off the board. What should have been a battle was transformed into a track meet. And you got no credit for chasing them off the board.

Now THAT was a shitty AI.

Imho it would be nice if you could upload the mod's AI code to github or sim. and open its development to more in-depth public discussions and eventually small contributions.
For what it's worth, I've often got some interest and motivations from some of your debats here on specific aspects of the game but knowning I'd have to start from scratch and the vanilla dll module, that always sounded somewhat intimidating and useless to me, for example.
I don't know if the the SDK licence allows it but that'd be secondary for sure compared to how much would that solution hurt Kevin's paranoia on complains for the sake of complaining.
You've got Dr Ezra Sidran around the forum too and some his knowledges might eventually come in handy, who knows...

Hi

If you join us for a game you'll get the current WIP dll which is constantly being adjusted. You'll need some knowledge of C++ though and to be honest Kevin is our coder on that, we are mostly a test and feedback team, though we also suggest inputs to alter things.

There's two main areas where the KS team is making other improvements. First a complete overhaul of the 1806 Prussians with new correct sprites to finally replace the Fontenoy 1745 stand-ins we've been using for 2 years! Secondly a major overhaul of as many maps as we can get done during the Prussian sprite project using the SDK that NSD recently released. We are using maps from Gettysburg as well as community-made maps for that game which we've used in the mod for years but which have always had 1860s American/Colonial style buildings. We are working through as many maps as we can converting the architecture, fields, field boundaries, road surfaces and terrain textures to a more European looking setting. We've got Portuguese, Spanish, Italian, French, Dutch, German and Polish maps in the pipeline. Other modders and friends of the team have made Italian, Russian, Polish and German maps as well. A friend of the team is making us a beautiful new set of German buildings.

The issue is we'll release a new mod version when the 1806 Prussians are done and we've got as many maps finished in a reasonable time frame. The new dll will then be released with that at whatever stage it'll have reached by then. Beyond that wer will almost certainly continue working on the dll, it'll most likely never be finished!

Future "wish lists" include 1805 Russians and Austrians and for me personally the very early Russians, Austrians, French and Piedmontese to allow scenarios in N Italy from 1796 to 1804.

We don't see any value in releasing the dll to a wider modding crowd because we're a small team and resources are limited, so discussion will always be held here on the KS forum. Logistics of monitoring other fora would just make it impractical. But by all means join us for an MP game, get the dll and if you know the code, have a peek inside. Alternatively you can just play the game, observe how the AI behaves in its current state and offer ideas for improvement.

Imho it would be nice if you could upload the mod's AI code to github or sim. and open its development to more in-depth public discussions and eventually small contributions.
For what it's worth, I've often got some interest and motivations from some of your debats here on specific aspects of the game but knowning I'd have to start from scratch and the vanilla dll module, that always sounded somewhat intimidating and useless to me, for example.
I don't know if the the SDK licence allows it but that'd be secondary for sure compared to how much would that solution hurt Kevin's paranoia on complains for the sake of complaining.

You'd certainly be welcome to contribute coding ideas and observations. As you probably know, much of the original code was poorly organized. The lack of forethought led to some parts contradicting subsequent parts of the same function or other sections being unreachable. I've rewritten a large portion of it, trying to organize the code in a cleaner and much more understandable form. Subsequently, many of those contradictions then became visible. We've spent the past month finding and eradicating these as well as improving the AI behavior.

Future "wish lists" include 1805 Russians and Austrians and for me personally the very early Russians, Austrians, French and Piedmontese to allow scenarios in N Italy from 1796 to 1804. wrote:

Ooh. I hope I live long enough to see that! I always thought Napoleon's early campaigns in Italy were spectacular.

You mentioned there was an upgrade to the SDK editor. Did they make any significant changes? I always thought it was a bit clunky.

They released the map editing tools about 3 months ago so a couple of us have been attacking the ACW maps converting them to Europe as fast as we can.

http://www.norbsoftdev.net/forum/waterloo-modifications/75449-waterloo-sdk-v2

I'm sorry I don't know if their scenario editor is included as I never use it.

I don't use it either so I guess it wouldn't hurt to DL it and see if it is at all improved. Can't hurt.

I found the tools - I guess. But I'm way out of my pay grade here. I guess I'll just wait for you guys who know what you are doing to do it.

Just use the KS scenario generator.

Thanks both for the replies, first of all.
God knows how much would I've loved joining your (or NTW3 in the past) MP activities since game release or before (GB), but that's damned hard for me.
MP at nights never matched decently with my schedules and resistence to relaxation and sleep.
My most recent attempt dates around 10 days back... but I think I was late and had no luck nor reply in your TS lobby.

I've played quite a lot in SP mode, though, your mod and others too.
And looked into the exe, dll and SDK as well.
Performed some graphical profling too and I know C++ of course.

The vanilla AI module is quite a mess imho.
I must say the gameplay results are surprising enjoyable compared to the poor quality of its code and the rough/basic techniques it uses.
I'm happy and hopeful to read Kevin is trying to put some order and I wish him best of luck in all cases.
My "general" suggestion is to avoid seeking for the "winningest" AI, nevertheless.
Let us ill-treated Napoleonic Warfare's enthusiasts to have a lot of fun with it, not just being challenged from a the strict gaming side.
Tradeoff that with realism if possible.
The central problem from my perspective could be summed in the AI showing...
1) too much dynamism where units and general just had no way to achieve it or was materially forbidden;
2) too much hurry when it should assess the situation and make its critical decision (QDCM, CDMP, DDMP, you know).
The actual AI methods shouldn't even have be labeled "Think" perhaps.
It cannot process it's intelligence, assess the situation, the threats, the enemy possible COAs, etc. consistently.
It cannot even formulate its own COA and follow it, unluckily.
It all seems just a matter of reacting to countless conditions (and many situations will escape them, unavoidably).
There're no influence maps either, which would come in handy for sure for your latest skirmishers work, for example.
Defensive situations are slightly better handled by AI probably also because of the relative gain in reactivity compared to humans...
But now you wouldn't have enough informations either to break this path.
I mean, Enemy/Friend/Targ+Beg/Next/Close/etc. wouldn't be not enough to build an acceptable commander's situation understaning before decision-making.
If you want to change that scheme, you need to take whatever data you need and make good use of it from the DLL, even if the APIs don't expose that functionality.
Period, imho. It's nothing insulting or illegal.
Speaking half in jest, I showed Davide once how to obtain the bounding coordinates of unit's sprites (graphical-representation, not just flag) to improve the collision-avoidance code just by casting and offsetting the CUnit pointer.
Well, he sounded so intimidated and against it that I'm still really wondering what kind of blasphemy was I committing.
It looked more or less like: (const CXVec *)(((const std::uint8_t *) unit) + offset).
Understandable given his official role and position... but still an approach out of this world, said with love.
I stop here my initial slice of AI discussion as I was maybe a little too harsh and grumpy.
But I think you'll get the idea.

And thanks for your enduring dedication to the mod.

"I showed how to obtain the bounding coordinates of unit's sprites (graphical-representation, not just flag) to improve the collision-avoidance code just by casting and offsetting the CUnit pointer."

This is something I wondered for a long time if the game could handle. If you can communicate to us how this could be made to work that could be something that would open up all kinds of opportunities for deployments, manoeuvring, LoS and targetting.

Sure I can. Found this matlab plot hidden into my folders...

The offset is unit + 0x328.
I'm going to write a small code snippet and post it in a moment.

Here you go.
https://pastebin.com/TrsRFYWX
Needs to be tested of course but I hope it'll help.

The game already knows about the area covered by a battalion in any given formation and keeps track of it. The problem is that the game engine doesn't really care about overlap. That was a deliberate NSD decision. They did not want to have units wait for a clear spot when trying to access a road during a march. Rather than suspend an overlap rule for road marching only, they suspended it universally. Typical.

The KS AI does invoke overlap restrictions as much as it can. However, since we don't have the game engine code it is only a band-aid. It results in units falling back when they try to swarm into the front line during a battle.

That's exactly my point, Kevin.
Too much reaction was historically unfeasible.
Also rule-based reactions won't catch many situations, are hard to mantain and behaviours tend to degenerate more often than not.
What is needed, from my understanding, are situational understanding and planning processes driven by goals.
If the resulting collision was unexpected and unavoidable, that's absolutely fine for me (I wouldn't mind a morale/cohesion malus, btw).
But thinking about it a (possibly long) time before it's going to happen, well, that was perhaps one of the most relevant component of any commander's frame of reasoning at the time.
Visualizing COAs and potential/likely/worst-case failures...
The game, in spite of his enjoyability, isn't able to catch that, so far. Sadly.

About NSD deliberate decisions: most of time I consider them as just constrained tradeoffs with no real alternatives, mate.
And concerning the engine code not being entirely available, I really can't see that as an huge obstacle.
It's not the same with or without it of course, but for AI at least Norb was "sensible" enough to give us an "entry point" through the DLL module.
If you need more data, normally we can take it through pointers offseting as showed above.
If you need to redirect the flow of a specific function to the DLL, you may also be able to do it.
I see nothing repugnant or illegal in doing it, and as they showed no plans or intentions to change that path since the game initial release, neither conflicting at this point.
Just my humble opinion, obviously.

Have a nice day.

The game already knows about the area covered by a battalion in any given formation and keeps track of it.

Yes, that's sure.
In my snippet I only tried to provide an interface from game data to the moddable AI dll code...
That wasn't allowed, right? At least I could not find any evidence it was possible (APIs doc, code, etc.).
I hope I'm not wrong because... hell, I lost a good quarter of hour to compile it!

Any thought about my overall view of things (rules vs plans)?
My game AI experience / pratical skills are rather basic atm.
I'd just know where to look at (books, papers, army manuals ...).
So don't get intimidated at all and speak frankly.

The strategy portion of the AI is in the game engine, not the dll. Even the stance of a brigade or division is set there with no way to change it via the dll.