[IMPORTANT] Can Union AI hold Cemetery Hill on first day?

[long5hot]

Latest version - Confed v Cunning on the Cemetery phase, and no it still can't hold the Cemetery with all objectives lost.

 

http://imgur.com/a/2uZO4#0

 

Points of note:-

• The AI still advances units in isolated fashion, allowing them to be routed piecemeal.  Sometimes it creates two lines, holding a thin front line that's then routed easily after which the second line can be defeated in turn.  There's no point keeping reserves back if they're not actually committed somewhere, when needed, and in a decisive & sensible manner.
• The line was not shortened in the center despite flanking both east and south, so once more the Union was short of troops.
• The alignment of brigades was sometimes headscratching.  One point saw 3 brigades side-on to a powerful flanking advance, with another brigade at right angles to them & its side exposed to attack from elsewhere, and therefore exposing yet another brigade further up the line to crossfire (you'll see the pic for it).
• The AI still contested the town, just not as aggressively, and still moved the Iron Brigade forward.  When its outnumbered in a defensive battle the AI needed to concentrate on falling back, keeping the line straight & short with units maximising cover, sending reserves to hold advances on the flanks.  It still does not do so.
• Once again there was no movement of artillery, or of HQs.

Notably an early end to the battle was declared in the strange "Battle Delayed" period, which I still don't understand, and it would have made me furious had I been playing a campaign as it allowed 2 completely surrounded Union Corps to escape certain and imminent destruction.

[Harnis]

Just a note for Olee O'Hara: historically artillery would limber and run away if they wanted yo, Mich faster than infantry could keep up with. To be historically accurate artillery should actually move faster

[long5hot]

Just a note for Olee O'Hara: historically artillery would limber and run away if they wanted yo, Mich faster than infantry could keep up with. To be historically accurate artillery should actually move faster

 

Yes, but they'd find it hard if not impossible to do this if under fire from infantry.  Horses tend not to hang around being harnessed while being hit by minie balls.

[OLee O'Hara]

Just a note for Olee O'Hara: historically artillery would limber and run away if they wanted yo, Mich faster than infantry could keep up with. To be historically accurate artillery should actually move faster

 

Yes, but they'd find it hard if not impossible to do this if under fire from infantry.  Horses tend not to hang around being harnessed while being hit by minie balls.

 

Yes exactly, I think you're both right... I get the point that limbered artillery is faster than running inf... But I rather thought of situations like the artillery is in musket range of my infantry, my infantry shoots and charges the artillery,  the artillery shoots back and runs away afterwards. They literally run away on top speed by pulling their cannons backwards after the shot by using the blowback. At the moment artillery is too fast when unlimbered shooting and too slow when they're limbered moving.

I could imagine something like a second value: each artillery unit could get a certain number of horses. The horses would get shot more easily than men and the remaining horses would determine the top speed of an artillery unit. A limber/unlimber button could be added or the limbering could just be simulated by a slower acceleration for artillery units before they get on top speed. Could be affected like the marching formation of brigades. If you move artillery to a farway position they could 'change formation' till they reach limbered top speed. Ofc the same the other way round: Always when cannons arrive at a new position after having been on top speed (when limbered) they could slow down by changing formation that would need some extra time before it's possible to do the first shot.

 

 

Edit: Sorry, it seems my suggestions always end up in the wrong thread :/

[RichardMcCor]

Latest version - Confed v Cunning on the Cemetery phase, and no it still can't hold the Cemetery with all objectives lost.

 

http://imgur.com/a/2uZO4#0

 

Points of note:-

• The AI still advances units in isolated fashion, allowing them to be routed piecemeal.  Sometimes it creates two lines, holding a thin front line that's then routed easily after which the second line can be defeated in turn.  There's no point keeping reserves back if they're not actually committed somewhere, when needed, and in a decisive & sensible manner.
• The line was not shortened in the center despite flanking both east and south, so once more the Union was short of troops.
• The alignment of brigades was sometimes headscratching.  One point saw 3 brigades side-on to a powerful flanking advance, with another brigade at right angles to them & its side exposed to attack from elsewhere, and therefore exposing yet another brigade further up the line to crossfire (you'll see the pic for it).
• The AI still contested the town, just not as aggressively, and still moved the Iron Brigade forward.  When its outnumbered in a defensive battle the AI needed to concentrate on falling back, keeping the line straight & short with units maximising cover, sending reserves to hold advances on the flanks.  It still does not do so.
• Once again there was no movement of artillery, or of HQs.

Notably an early end to the battle was declared in the strange "Battle Delayed" period, which I still don't understand, and it would have made me furious had I been playing a campaign as it allowed 2 completely surrounded Union Corps to escape certain and imminent destruction.

Longshot... can you play the battle again but this time an union and demonstrate what the AI should have done (assuming that the CSA AI attempts the same strategy you did that is... which it probably will not). Try CSA AI set to Offensive.

 

Love your comments and pictures by the way.

[long5hot]

Longshot... can you play the battle again but this time an union and demonstrate what the AI should have done (assuming that the CSA AI attempts the same strategy you did that is... which it probably will not). Try CSA AI set to Offensive.

 

Love your comments and pictures by the way.

 

Good idea, I'll get onto it - although the last time I played this mission as Union I was able to hold all objectives by just letting the battle run & not clicking on anything.  The AI will need to have significantly improved if its going to force me to have to defend properly.

 

Edit: Ok, tried a mostly-no-click battle against Offensive.  "Mostly" because I gave the reinforcement brigades simple movement orders (basically to march to the Cemetery) then left them alone.  Here's the result:-  http://i.imgur.com/N4Yr0Ok.jpg

 

The AI had one strategy and one strategy only - frontal assault (I think this guy was in charge).  There's no attempt to flank or maneuver as I do in this battle, it just makes repeated charges ... except for when it doesn't, and keeps troops motionless for some reason.  At one point it had pushed my troops right off the Cemetery, had Pettigrew and Scales advancing in force straight towards the open objective point, then they stopped short and dithered for the rest of the battle. 

 

So playing against the AI I cannot replicate what could be done to stop the flanking attacks that I make when playing Confed.  Perhaps I should find a top-down map of the battlefield and draw what the AI does versus what I would do (if that'd be helpful at all).

[long5hot]

Ok, a top-down view of how a good AI would handle the flanking strategy I throw at it:-

 

http://imgur.com/a/WCvHF#0

 

Could an AI do this? 

 

It'd have to know that its outnumbered and therefore rule out impulsive attacks across the board, no matter what.  No more pushing Iron Brigade off the hill to attack whatever weak-looking enemy brigade that appears over Benners, no matter that the Iron Brigade's morale is so high.  No more contesting the town at all.

 

The AI would have to then look to where cover is around the objective points, and station troops there & where possible move artillery to vantage points that assist the troops, even if that places the artillery in the front line. 

 

Next, the AI needs to prioritize keeping a straight front line across the battlefield with no gaps or open ends that can be rolled up.  Each brigade linked to one beside it without too great an angle differential in their facings.  Where it cannot fill a gap, or where enemy troops are marching around the end of the line, it needs to either send reinforcements or fall back to shorten the line, keeping the objective points defended.

 

The AI as its stands now is not close to accomplishing this.  It holds a proper line in segments, then leaves the end hanging with big gaps until the next line segment.  It creates double lines leaving the outer one weak and exposed.  It doesn't respond to flanking attacks rolling up lines or to brigades being enfiladed - it leaves them in bad positions until they rout.  It often seems to let each brigade decide its own facing with "interesting" results.  And it marches units off objective points leaving them wide open, while being aggressive in areas that simply don't matter.  It has little concept of cover & using it to advantage. 

 

There's a lot to do to improve this AI.  Its pretty good, and I know how difficult AI programming is, and I don't expect it to be as brilliant as Napoleon or Wellington, but it does need to improve if its going to provide the game with decent replay value & longevity.

[David Fair]

Long5hot,

 

Great post and excellent points to improve the AI performance.

[RichardMcCor]

Excellent post. I think the way Darth set-up the AI with it's different personalities allows for some to be able to do the flanking while others just barrel on down. Lets see what he says in reply. I'm sure it will be positive.

[Nick Thomadis]

Ok, a top-down view of how a good AI would handle the flanking strategy I throw at it:-

 

http://imgur.com/a/WCvHF#0

 

Could an AI do this? 

 

It'd have to know that its outnumbered and therefore rule out impulsive attacks across the board, no matter what.  No more pushing Iron Brigade off the hill to attack whatever weak-looking enemy brigade that appears over Benners, no matter that the Iron Brigade's morale is so high.  No more contesting the town at all.

 

The AI would have to then look to where cover is around the objective points, and station troops there & where possible move artillery to vantage points that assist the troops, even if that places the artillery in the front line. 

 

Next, the AI needs to prioritize keeping a straight front line across the battlefield with no gaps or open ends that can be rolled up.  Each brigade linked to one beside it without too great an angle differential in their facings.  Where it cannot fill a gap, or where enemy troops are marching around the end of the line, it needs to either send reinforcements or fall back to shorten the line, keeping the objective points defended.

 

The AI as its stands now is not close to accomplishing this.  It holds a proper line in segments, then leaves the end hanging with big gaps until the next line segment.  It creates double lines leaving the outer one weak and exposed.  It doesn't respond to flanking attacks rolling up lines or to brigades being enfiladed - it leaves them in bad positions until they rout.  It often seems to let each brigade decide its own facing with "interesting" results.  And it marches units off objective points leaving them wide open, while being aggressive in areas that simply don't matter.  It has little concept of cover & using it to advantage. 

 

There's a lot to do to improve this AI.  Its pretty good, and I know how difficult AI programming is, and I don't expect it to be as brilliant as Napoleon or Wellington, but it does need to improve if its going to provide the game with decent replay value & longevity.

Excellent.

Please play this battle in latest 0.97 revision against Cunning but it is important to play in campaign, to follow the casualties and exhaustion that realistically affect the outcome.

To replicate this you need to hold in the first battle as Confederates the following:

Herr Ridge AND Seminary OR (Herr Ridge AND Oak Hill AND Oak Ridge).

Alternatively you can hold only Herr Ridge and hope that in next battle the Union AI places the defense on the lower ridges and on Cemetery Hill.

[long5hot]

Please play this battle in latest 0.97 revision against Cunning but it is important to play in campaign, to follow the casualties and exhaustion that realistically affect the outcome.
To replicate this you need to hold in the first battle as Confederates the following:
Herr Ridge AND Seminary OR (Herr Ridge AND Oak Hill AND Oak Ridge).
Alternatively you can hold only Herr Ridge and hope that in next battle the Union AI places the defense on the lower ridges and on Cemetery Hill.

 

Ok, I held Herr & the two Oaks, but the next map I see isn't the one I've been testing.  Rodes has not yet arrived, there's an objective point on McDonald's property south of the Seminary that's in Union hands, and indeed I start the phase with only 3 brigades and a sprinkling of artillery.  Heth's Skirmishers and Archer, who finished the previous phase alive & well, are not present.

 

Am happy to test this, but it isn't the same battle.

[long5hot]

And here's the result of the 2nd phase:-

http://imgur.com/a/Ot6Bf#0

Very disorganized battle from Cunning. It sent no reinforcements to defend Seminary Ridge so that line was quickly smashed once my new brigades arrived. Indeed the objective point was defended at the outset by brigades I'd already routed in the 1st phase so it didn't take long to do so again.

Then Cunning sent nearly all of its 5 brigade strong defensive force around the Cemetery, which could have assisted in the Seminary Ridge battle but didn't, north into the town on a wild goose chase after Doles and (I think) Iverson. It pushed those brigades out of the town, but that left the Cemetery completely exposed other than the rallying brigades I'd already broken.

Culps was strongly held with 2 brigades, until Cunning sent one of them (Krzyz) over to Benners to attack O'Neal. My reinforcements arrived on that wing, and Krzyz did not withdraw but duked it out over on Benners until it routed. Then I was able to do the same with Smith who was now the sole defender on Culps.

End result, Cemetery falls in the second phase. Even though the game tried to end the battle early and deny me the victory. I so hate that feature.

[Harnis]

The AI has come a long way. My last test of that Confeds advance to attack cemetery against cunning union AI resulted in a solid victory and I took all the objectives, but....

 

It only resulted in around 2000 casualties for my confeds and around 4k for the union. Why? Because instead of sitting there on cemetery hill while they get encircled and demolished, once some union brigades started to rout and I was pushing hard, the entire union army fell back to where the reinforcements were coming in. So while I won a massive victory, if that battle had been in a campaign, the union AI had preserved almost it's entire army in the face of a huge defeat. Instead of taking the desperate approach of holding the line at all costs and shortening its lines, sending reserves in, etc, it instead chose to fall back and fight another day.

[long5hot]

Ok, last playthrough of Cemetery on the final pre-release version. Comments on the images:-

http://imgur.com/a/vsNQy#0

Edit: I said in my end summary that Baxter had been destroyed, when it wasn't (though near enough for all intents and purposes).

[RichardMcCor]

Building on from what longshot observed about Iron Brigade breaking cover I noticed this last night as well as what the AI does in a volley vs volley slug fest.

 

COVER

The AI needs to place a heavier weighting on staying in cover. Its much more critical to be in good cover in the most recent updates and the AI does not seem to understand that.

 

SLUGFEST

In a shooting match, after a few volleys if you are losing more men than your opponent is each volley  you need to take that into account. A human can easily see that your poor brigade in no cover losing 50 men per volley vs the opponent in cover losing 5 men per volley will rout. The AI needs to also keep a count after a few volleys and make a change. Even a simple fallback will sometimes tempt the opponent out of cover.