34 replies to this topic

[Smokeskin]

Here's my list of suggestions for improvements. Some of them might have been fixed for beta 3 or 4 already, I haven't checked all of them.

I would appreciated if no one replied here before Arkhan, so he can get post #2 instead of somewhere down the page

I think the code team could mark them with
? clarification requested
I implemented
A agreed but not yet implemented
E examination hasn't been done yet
R rejected (it would be great if you noted why, for example because the AI can't do it, because you just disagree, or because coding it is outside of the scope of the project, or whatever the reason)



General
Squads should decap when they destroy an LP, then continue fighting
Effective AV? Is there a way for the AI to detect many opponent vehicles and build specific squads to counter it?
Tight teching Bos – go t3 with just 1 building is beneficial for several races (ie SM doesn’t go both Machine Cult and Sacred Artifact in t2)

SoB
Upgrade infantry with leaders ASAP, they are essential from tier 2 and on
Upgrade Celestians ASAP - they are rather useless without their weapon upgrades
Build early Celestians for AV
More upgrading of battle sister squads to heavy bolters (minor issue)
Seraphims don’t use Angelic Visage

Tau
Tau Commander should upgrade to missile launchers ASAP
Tau Commander should upgrade to plasma rifle ASAP
Stealth suits should built shas’vre, it adds double health to squad and EMP ability
TC places too many snare traps and does it in strange places
Kroots cannibalize under fire

SM
Should upgrade with missile launchers more often

Chaos
Needs to switch to mostly berserkers and some horrors instead of marines t2
Needs Obliterators in t3

Guard
Earlier Tactica Control needed, basic guardsmen are cannon fodder
More than 1 grenade launcher per squad in t1 needed
More heavy weapon teams in t2

Ork
Doesn't mass units, especially t1, plays passively

Edited by Smokeskin, 30 March 2008 - 08:25 PM.

[ArkhanTheBlack]

I addressed most stuff. I'll therefore mention only the stuff I didn't or couldn't solve:

Squads should decap when they destroy an LP, then continue fighting

We already discussed this in DC and decided that it's more efficient if they only cap if there's no threat.

Effective AV? Is there a way for the AI to detect many opponent vehicles and build specific squads to counter it?

Detecting isn't very reliable, but usually it's enough to give specific squads high unit ratings to see more of them on the battle field.

Tight teching Bos – go t3 with just 1 building is beneficial for several races (ie SM doesn’t go both Machine Cult and Sacred Artifact in t2)

Both are important in T2.

TC places too many snare traps and does it in strange places

I'm not sure if I really want to rewrite that code. We'll see...

More than 1 grenade launcher per squad in t1 needed

I can't force that so easily, but since the Tactica Control is now built earlier, there's a good chance that they built more now. They've already free upgrade rights in the harassing phase.

Ork
Doesn't mass units, especially t1, plays passively

That's pretty much the same as adding a point "Race X doesn't play good enough!". All races use the same core code and will stay passive if they think they are not strong enough compared to the enemy. In this case they will wait until they have gathered enough troops for an attack.
Orks gather exactly the same army strength in T1 like all other races (1000). I also had a test game with them and they had no problem to crush SM. Don't justify everything from one test game. It's map dependent and complex harassing fights were and will never be the strength of the AI.


General comments:
- I increased the unit rating of some units to increase the chance that the AI builds them. I also added Oblits to the build programs.
- The forced leader was a good idea. I've now included it for all races.
- 50% of all upgrades will now be anti-vehicle

Edited by ArkhanTheBlack, 30 March 2008 - 10:21 PM.

[LarkinVB]

I'll add my 2cents.

Since I love delving into the code I'd like to see some improvements to MoveToDisengage().

Once it was designed to make squads just dance. Now it is used for broken-dancing-retreating squads.

This is a problem cause you don't want a dancing squad to retreat to the next building some 150 away. (300 is the map from a-z). Therefore I'll ask Arkhan to improve this code or to let me do it. Some month ago squads were dancing really good. Just away small distance, then shooting again. Now they retreat instead, a long long way.

This is new (AI > 2.0). Squad micro can be improved with some minor changes to the code.

[ArkhanTheBlack]

1.) I can confirm now that Orcs weren't massing. It's a very odd problem that occurs because the Big Mek attaches and the build controller can't find him anymore and start to build a new one.

Since I love delving into the code I'd like to see some improvements to MoveToDisengage().

2.) I think the only thing that should be done is to avoid that a unit runs back the WHOLE way back to base. But I most certainly don't want the code back which forced the units to dance straight into the enemy base and get slaughtered miserably.

Edited by ArkhanTheBlack, 30 March 2008 - 11:07 PM.

[LarkinVB]

Ok.Ill try do some MoveToDisengage() which will be better. Already working on it.

[Smokeskin]

Sounds great with the changes

Decapping in combat:
In fights you're going to win (ie force the enemy to retreat), decapping under fire is inefficient, it is better to just wait until the enemy is forced off.
In fights you're going to retreat from, you're much better off decapping. Since you're in combat, your units are getting shot at anyway, and most squads will hurt the enemy more by decapping than their damage for the seconds it takes to decap.
It is extremely annoying to get decapped, it hurts your income and it forces you to assign a unit to cap it, so you can't move around and fight at full strength.
There are cases (mostly when you take out an LP with long range heavy weaponry) where trying to decap is suicide because you run into a lot of fire that you otherwise wouldn't be exposed to.
I understand that the AI can't well distinguish between the situations mentioned above, but overall, decapping in a winning fight is just a bit less inefficient, while decapping in a losing fight is a very strong move. I think it would work better to have the AI decap, but I'm open to the fact I can't predict how changes to the AI will actually work out in gameplay.

Wide t2 builds (ie both Machine Cult and Sacred Artifefact) vs tight t2 builds
The problem I have with getting both is that you mostly end up with few vehicles and few heroes with few researches. The vehicles quickly get taken out by AV, and you don't have all heroes out and they're not strong enough yet.
At the same cost, you could be going full vehicles and no heroes (saves both on SA building, commander cost, and research cost, allowing you more vehicles). Unlike before, were there are so few vehicles that they quickly get taken out, you now have enough vehicles to overwhelm his AV. You know how this works, 1 dreadnought will get beaten up badly on its way in, then get danced and taken out without having done much. 3 dreadnoughts on the other hand, 1 gets beaten up badly on the way in, but the 3 of them are now in among the enemy troops and wrecking some major havoc, and they're too many to get outdanced.
Or at the same cost, you could have no vehicles but all 3 heroes, strongly researched, that really deal out a lot of hurt and are VERY cost efficient because of their researches.
You get a more powerful battlefield presence in t2, and you can go t3 faster, by doing a tight BO imo.

TC snares:
That's cool, if it takes major coding (and I can imagine stuff like that needs some heavy testing to see the effects of the code too, so quite time consuming) to get better, by all means the issue isn't that big. It is also a very powerful ability, not seeing it used would be a big shame.

I'll look at updating my list with your changes later, work calls

[troubadour]

Wide t2 builds (ie both Machine Cult and Sacred Artifefact) vs tight t2 builds
The problem I have with getting both is that you mostly end up with few vehicles and few heroes with few researches. The vehicles quickly get taken out by AV, and you don't have all heroes out and they're not strong enough yet.
At the same cost, you could be going full vehicles and no heroes (saves both on SA building, commander cost, and research cost, allowing you more vehicles). Unlike before, were there are so few vehicles that they quickly get taken out, you now have enough vehicles to overwhelm his AV. You know how this works, 1 dreadnought will get beaten up badly on its way in, then get danced and taken out without having done much. 3 dreadnoughts on the other hand, 1 gets beaten up badly on the way in, but the 3 of them are now in among the enemy troops and wrecking some major havoc, and they're too many to get outdanced.
Or at the same cost, you could have no vehicles but all 3 heroes, strongly researched, that really deal out a lot of hurt and are VERY cost efficient because of their researches.
You get a more powerful battlefield presence in t2, and you can go t3 faster, by doing a tight BO imo.

200% agreed with that, We may discuss which building should be build first thought (Sacred Artifact or Machine Cult),
Usually if i think i got a slighly advantage and that my ennemy dont go for vehicule i go for 3 commanders BO and keep on building SM squads (RL + plasma or HB), if the game is balanced i go for vehicules and build Machine Cult and pump LS + Dreads, i build no commander (no Chappy, no Lib) and build SacredArtifact in ordre to research T3
It also depends if you are playing 1vs1, in team game if my ally go to vehicules (no matter which race he is playing) i go for 3 commanders BO

[ArkhanTheBlack]

If we delay one of the SM buildings then we delay the sacred artifact. The dreadnought is pretty much the best T2 vehicle in the whole game (from AI perspective), I don't think SM can use their heroes good enough to compete with the dreadnought way.

[troubadour]

If we delay one of the SM buildings then we delay the sacred artifact. The dreadnought is pretty much the best T2 vehicle in the whole game (from AI perspective), I don't think SM can use their heroes good enough to compete with the dreadnought way.

Yes i think the machine cult is the most efficient and all around-reliable way to go for the SM AI

[Smokeskin]

Space Marine Infantry power showcase

These AI files for SM don't build any vehicles at all (even dynamic building is disabled).

It always upgrades infantry sqads with heavy weapons, other than that the tactics files are untouched iirc, I only modified the strategy files to get the effect.

For some reason I can't get it to build chaplains and librarians in t2, so I tech to t3 quicker than I would've liked. It would be stronger in t2 if it would build commanders.

I've tried around 10-12 AI vs AI battles in total vs eldar, chaos and necron with it, where it has 100% win rate. 100%, not a single loss.

I've also tried it vs Tau, but AI vs AI tau completely decimate SM in tier 1, so it doesn't really matter what sort of build you do in t2.

Your AI is more than capable of handling all infantry armies and tri-hero setups and getting excellent results. Perhaps an only machine cult BO beats what I have here, but what I have here certainly beats the "wide t2" BOs currently in your AIs.

Try downloading it and see how it plays (remember to save your old SM files first).

I don't think we should always build the sacred artifact, I think we should have BOs that sometimes build machine cult, sometimes SA in t2. It would be more fun and varied to play the AI, and both builds are efficient imo. And of course dynamic building shouldn't be disabled like I've done here, that was just to show that all infantry can work (the first versions dynamic building was on and it kicked in perfectly late game and started spamming vehicles btw).

Attached Files

•  space_marine_race.rar   20.93KB   1 downloads

Edited by Smokeskin, 31 March 2008 - 03:53 PM.

[LarkinVB]

2.) I think the only thing that should be done is to avoid that a unit runs back the WHOLE way back to base. But I most certainly don't want the code back which forced the units to dance straight into the enemy base and get slaughtered miserably.

I liked the old code better. Better you fix it for yourself, seems we disagree on how to do it the right way. Now it is not good as they won't dance at all but always retreat to nearest building.

[dreddnott]

I commented about the dancing and retreating issues in another thread - however you decide to do it, it needs to be fixed.

Just having a CC unit next to an enemy shooty squad causes the enemy to completely abandon the fight!

[troubadour]

I don't think we should always build the sacred artifact, I think we should have BOs that sometimes build machine cult, sometimes SA in t2. It would be more fun and varied to play the AI, and both builds are efficient imo. And of course dynamic building shouldn't be disabled like I've done here, that was just to show that all infantry can work (the first versions dynamic building was on and it kicked in perfectly late game and started spamming vehicles btw).

Yes you are right the 2 BOs are efficient but if only one of them would make it in final release which one ?
I would go for vehicule BO because it seems to me (i may be wrong) that is the most all-around compare with 3 Commanders BO that is more suitable vs ennemy mass troops

Edited by troubadour, 01 April 2008 - 08:32 AM.

[Smokeskin]

I don't think we should always build the sacred artifact, I think we should have BOs that sometimes build machine cult, sometimes SA in t2. It would be more fun and varied to play the AI, and both builds are efficient imo. And of course dynamic building shouldn't be disabled like I've done here, that was just to show that all infantry can work (the first versions dynamic building was on and it kicked in perfectly late game and started spamming vehicles btw).

Yes you are right the 2 BOs are efficient but if only one of them would make it in final release which one ?
I would go for vehicule BO because it seems to me (i may be wrong) that is the most all-around compare with 3 Commanders BO that is more suitable vs ennemy mass troops

Why choose? There are already 4 BOs in the code. Have 1 or 2 of them do an SA build.

I don't know if you tried my tweaked SM BO, but seeing the AI hit t3 relatively early and drop pod terminators in to support the 3 commanders, grey knights and a mass of space marines - I just thought it was so cool I think it would be much more refreshing if it did sometimes one, sometimes the other.

[Smokeskin]

I commented about the dancing and retreating issues in another thread - however you decide to do it, it needs to be fixed.

Just having a CC unit next to an enemy shooty squad causes the enemy to completely abandon the fight!

If that's the problem, then it seems that a simple check for distance to nearest enemy could solve it, if it is above a certain value, stop dancing.

[ArkhanTheBlack]

If that's the problem, then it seems that a simple check for distance to nearest enemy could solve it, if it is above a certain value, stop dancing.

Done!

Why choose? There are already 4 BOs in the code. Have 1 or 2 of them do an SA build.

I don't know if you tried my tweaked SM BO, but seeing the AI hit t3 relatively early and drop pod terminators in to support the 3 commanders, grey knights and a mass of space marines - I just thought it was so cool biggrin1.gif I think it would be much more refreshing if it did sometimes one, sometimes the other.

The build is quite funny! It seems to hit the vehicle techers hard because they take a bit until their tank production is running. I combined it with the marine rush strategy. It really 'fits', and if things are running bad all alternative build strategies fall back to the standard build.

Does such a thing work for Chaos too?

[dreddnott]

Chaos early T3 with Sacrificial Circle and no Machine Pit is great because you get evil nasty Obliterators and Possessed Space Marines.

Here's how it works, broadly:

HQ builds 1 Heretic, 2 Cultist squads
Heretics build Barracks
Barracks makes 1 Raptor, 1 CSM for harrass/defense (no armory)
Heretics build plasma gens as needed
HQ techs to Tier 2, makes Chaos Sorcerer
Heretics build Sacrificial Circle, which researches Chains of Torment
Spam 2x or 3x Khorne Berserkers w/ Sorcerer supporting (Horror squad for AV)
HQ techs to Tier 3
Sacrificial Circle makes Possessed Marines, researches Daemonic Fire upgrade
Heretics build Daemon Pit, which makes Obliterators

This was killer in Dark Crusade due to better fast-tech. Soulstorm is slower but it still works well in team games.

[Smokeskin]

I don't know if you tried my tweaked SM BO, but seeing the AI hit t3 relatively early and drop pod terminators in to support the 3 commanders, grey knights and a mass of space marines - I just thought it was so cool biggrin1.gif I think it would be much more refreshing if it did sometimes one, sometimes the other.

The build is quite funny! It seems to hit the vehicle techers hard because they take a bit until their tank production is running. I combined it with the marine rush strategy. It really 'fits', and if things are running bad all alternative build strategies fall back to the standard build.

Does such a thing work for Chaos too?

Glad you like it
A few important things about it:
- I don't think it will work unless you make the tac squads force some heavy weapon upgrades through. Without the dreadnoughts as heavy hitters, they really need the extra punch from those heavy weapons. The current Upgrade() function in spacemarineinfantry.ai doesn't let them upgrade often enough.
- I think the heavy weapons research probably need to be included in the BO. In my "version" I just made a quick fix by adding it as the first line in the dynamic research, but that would affect all the other BOs too, so not a good idea.
- I can't figure out why I couldn't get it to build Chaplain and Librarian in t2? It would work much better if you could make it build them in t2 (and if it worked, I would delay t3 a bit compared to what I wrote in my BO, but I had to quicktech to get the 2 commanders out).

Btw, I think you should do the reverse in at least some of the other builds, remove sacred artifact from t2 and just go machine cult, pump some vehicles, t3.

I think it could work well for chaos too. Go t2, get a sorc and some berzerkers and horrors, then t3 and possessed space marines and obliterators, it's a common strategy for humans. I think compared to SM for chaos it works better as a fast tech build - don't mass a great deal of units in t1, but get some turret upgrades for good eco and defense, quick tech to t2, get some units, and quick tech to t3. I could try tweaking a BO and see what works. My main concern is that horrors are going to be their main AV, and they're a bit tricky to use I think.
I'm hoping I'll be able to get the sorcerer out in t2 though, I'm still mystified why I couldn't get libbys and chaps going for the SM.

EDIT: I hadn't seen your post dreddnott, but obviously I fully agree

Edited by Smokeskin, 02 April 2008 - 08:01 AM.

[LarkinVB]

You can force build units in the static BOs.

[ArkhanTheBlack]

- I can't figure out why I couldn't get it to build Chaplain and Librarian in t2? It would work much better if you could make it build them in t2

Your unit entries didn't work because you added an army strength requirement to them and army strength works different for units. The unit will only be built if they have LESS than the demanded army strength. Since your AI probably had 1000 army strength the most time, they did never build it.

- I think the heavy weapons research probably need to be included in the BO. In my "version" I just made a quick fix by adding it as the first line in the dynamic research, but that would affect all the other BOs too, so not a good idea.

I've done that already.

I also had to do some other tweaks. You didn't modify the force tech behaviour. Therefore the force tech doesn't find a vehicle building and techs straight to T3. This can be quite deadly for the opponent if the AI isn't in trouble, but if it had a bad start, they won't build troops in T2 anymore. I have fixed that now, so they check for the sacred artifact instead.
I also have to build two sacred artifacts or the researches block the grey knights and the librarian. It's pretty cheap for a building so I don't think it hurts the economy too much.

I also modified the infantry weapon upgrade behaviour for all races. They will now always upgrade if a squad is in good state. I think the advantage of having a good squad equipped with heavy weapons overrules the economic needs of the build controller.

I'm think about delaying the power research until the T3 was started. At the moment the dynamic generator builds jump in and they horde much more power as necessary in T2. I think that's the first time I've seeing the dynamic generator code working against our builds.

Edited by ArkhanTheBlack, 02 April 2008 - 10:19 AM.