16 replies to this topic

[ThetaOrion]

I looked a couple of hours ago, and Beta 9 is there.

[LarkinVB]

No gameplay changes according to the readme.

[Corsix]

1.65b9
- Updated Raceloader logic (thanks Corsix) which allows for greater dynamic faction integration

Good

Edited by Corsix, 15 February 2006 - 08:14 AM.

[thudo]

Yep and working GREAT! Thanks Corsix! Thats a big deal with the enhanced Raceloader! A major complement to the work Arkhan has done.

[Excedrin]

Is this the right place to post beta 9 replays? I guess that since they should be the same as beta 8 replays gameplay wise maybe I should use the other thread, but anyway, here's a game I just played.

http://dow.lerp.com/...0216.113908.rec

SM vs SM Insane on Blood River. It makes tons of generators and then starts the dread spam. I stopped the dread spam and it had nothing left...

The major issues in this replay are:
1. FC runs off on a suicide mission with no support
2. Pointless ASM jumps (jump in, watch all asm die as they run away sort of thing, without them trying to get into melee)
3. Resource management.. it tries to get tons of power income but not enough req income.
4. The spam issue
5. No contest for early game map control
6. Losing infantry, it only wants to run when morale is broken, not when the squad is taking heavy damage and about to die.

I suppose 1 and 2 are the same, basically, it's happy to fight unwinnable battles. 3 and 4 are related too, since it tries very hard to spam vehicles.

[Zenoth]

Just curious ...

What that with the -dev switch or not ?

[Excedrin]

It was without -dev, but I still haven't noticed a difference between -dev or not.

[thudo]

Excedrin.. There are 4 buildprograms. Some build generators fast and ONLY one spams Dreads (which I might add are overpowering but thats enuff said). I think it also has to be notes (although its known) that yer the ONLY player I know who can win against our AI on Insane which is itself God-like, thus, you are an uber-player. About 90% of people can't even win against our AI on HARD let alone anything above that.

That being said.. your wisdom and advice is important as you represent the summit of what we are trying to achieve (although it always appears too daunting to satisfy even the most elite DoW/WA players most of times). However, there are points you make that don't make sense sometimes - example: enemy AI troops only run away when morale is broken and NOT when almost dead. I would beg to differ.. this may happen sometimes but NOT always. Still.. worth investigating.

Still.. why not comment out the last 3 buildprograms and then just play the All-Arounder where there is no spam. Different result?

[Malkor]

I win against insane all the time... At least if it's IG. If it's the other guys, it depends if they go vehicles or not.

Edited by Malkor, 16 February 2006 - 11:19 PM.

[Finaldeath]

thudo, its not difficult to win against the AI on hard, even in 1v1, or FFA, nevermind team games, when it gets easier, as long as you know how to tech efficiently and don't waste units

I'll report on new betas later, need some time to sort stuff out however before.

[thudo]

I would tend to believe if it were not for the many followers who find it VERY challenging on HARD. Most of the people who can defeat the AI even on HARDER+ usually are seasoned RTSers and are "rapid clickers".

Re: Vehicle Rushes.. Part of me wants to remove them.. teching to vehicles is just too risky unless the AI has a LARGE map to play with. It has to be perfect and the AI just works better when it can rush in Tier1. Tier2 is too difficult for the AI to pull off a rush in most conditions UNLESS its an ally and not currently being attacked.

Anyway I have some good news.. Beta10 is coming..

[Malkor]

Actually I'm not a click spammer. Unlike silly twits who think APM has some remote relevence to skill, I tend to lean towards "make each click count." I'm actually pretty bad with macro and rely more on micro and sheer tactical supremecy, which is one of the reasons I tend to vehicle rush.

Edited by Malkor, 17 February 2006 - 06:54 AM.

[sath]

ai on anything on standard or any above usually like to create a handful of e.g storm boyz and a commander and send them into battle which is bad for me since i seem to be plagued with bad luck on harder difficulties no matter how many different methods i use to steer clear of harm and the only one that works is not building any barracks but defences and the ai might leave me alone because sometimes i think that if i'm on a large map with lots of players they secretly sign truces with eachother and it isn't until i'm dead do they start going after each other and according to my mates the rate which i click and direct troops and builders on an rts like C&C zero hour is kinda scary but in my eyes there are most probably people who make me look slow when playing online

8 fire prisms BAM! *dead* unless there is a bloodthirster in da house and a swarm of cultists.

i don't bother with chaos predators very much unless i'm bored with defilers who for chunks of artillery can really take care of themselves so 6 defilers can mop everything under the carpet then stamp it to death

and these are just a couple more reasons to vehicle rush!

ah but 6 obliterator squads with their additional troops can wreck anything since it's 30 obliterators and they're virtually walking tanks so in their own way they operate like vehicles only slow and are actually infantry

[Excedrin]

Ok, I edited marinestrategyinfo_hard.ai so that all builds except the "Allround strategy" are commented out. With that in place, it played exactly the same.

http://dow.lerp.com/...0217.010900.rec

Here's a rep, this one is even better than the other one since I won via T&H. There was a distinct point where I felt like saying, "silly AI, dreads can't uncap". As a note, in Rebirth, most units can't cap or uncap (some can uncap and not cap, like raptors), so I added some info to unitstats.ai to reflect this. Then changed resourceplan or whatever so that it would not select units that can't cap to go on capture missions.

Here's an example unitstats.ai entry from Rebirth, note that customstats.ai doesn't exist and I haven't added the squad size info. This is generated by a script. The troopercost key/value is used to switch squads with weapon upgrades to ranged stance.

	{
	  name = "chaos_raptor",
	  sbp_name = "chaos_squad_raptor",
	  ebp_name = "chaos_raptor",
	  class = UnitStatsAI.UC_HeavyInfantryMed,
	  able_to_capture = false,
	  able_to_uncapture = true,
	  troopercost = 145,
	  rating = 3.27514792899408,
	  potential =
	  {
		{
		  name = "chaos_bolt_pistol_raptor",
		  effectiveness = GenerateUnitEffectiveness(0.01,0.01,0.01,0.01,0.01,0.01,0.01,0.0
1,0.01,0.01,0.01,0.01,0.01,0.01),
		  range = UnitStatsAI.RT_Ranged,
		},
		{
		  name = "chaos_flamer_raptor",
		  effectiveness = GenerateUnitEffectiveness(0.01,0.01,0.01,0.01,0.01,0.01,0.01,0.0
1,0.01,0.01,0.01,0.01,0.01,0.01),
		  range = UnitStatsAI.RT_ShortRanged,
		},
		{
		  name = "chaos_plasma_gun_raptor",
		  effectiveness = GenerateUnitEffectiveness(3.89,3.89,3.89,3.89,3.89,3.89,3.89,3.8
9,3.89,3.89,3.89,3.89,3.89,3.89),
		  range = UnitStatsAI.RT_Ranged,
		},
		{
		  name = "chaos_chainsword_raptor",
		  effectiveness = GenerateUnitEffectiveness(0.05,0.05,0.05,0.05,0.05,0.05,0.05,0.0
5,0.05,0.05,0.05,0.05,0.05,0.05),
		  range = UnitStatsAI.RT_Melee,
		},
	  },
	  melee = true,
	},

Since I want to be able to look things up by key instead of having to iterate thru unitstats, I have an additional file called custom.ai that contains the following:

class 'Custom'

Stats = {}

function Custom:__init(UnitStats)
  local i, i2, race, squadstats
  for i, race in ipairs(UnitStats) do
	for i2, squadstats in ipairs(race) do
	  Stats[squadstats.sbp_name] = squadstats
	end
  end
end

Anyway, back to the skirmish AI...

		{ 2, 0, 0, 0, 2,   "Restrict",  "space_marine_squad_scout" },
		{ 2, 200, 100, 250, 1,   "Building",	"space_marine_vehicle_building" },
		{ 2, 150, 0, 500, 1,   "Research",	"support_cap_research" },
		{ 2, 170, 305, 0, 1,   "Unit",  "space_marine_squad_dreadnought" },

I think some tuning of these might help... If I understand correctly, it won't make a vehicle building unless it has 250 army, but 250 army is like 1 marine squad or a FC alone. Obviously, 0 army cost in the dreadnought entry ensures that it will spam dreads even when it has no infantry.

[ArkhanTheBlack]

Excedrin:
The captureplan never sends non - capturers on capture missions. What you saw is the attackplan initiating an attack to CONQUER the critical. Normally there's at least one squad in the attack force that can capture a point.

Considering the build programs, if you remove the vehicle, the AI will be worse than before...much worse! The problem of the AI is how it uses his forces. An average player will simply make a fast tech and beat the crap out of the AI. I've lots of replays from other people which prove this fact. Try it out, but it won't change anything. You mentioned in one post that an insane AI could beat (or almost beat) you when you let it reach the vehicle spam. And that's exactly the point. Vehicles are the only real chance for an AI to get strong. Tier 1 squads are vulnerable and too weak. You have to know exactly how to handle them and what you are able to attack. Without explicit building targeting to destroy the major military buildings first, this is hopeless.

One SM problem in the replay was that the AI didn't use the scouts for attack and defense because they start kind of weak with two troopers. I fixed this too in beta 10, so they'll use them at least in tier 1. In tier 2, 2- trooper squads with low price are sorted out for recovering and capture missions.

Thudo:
A few betas ago I restricted fast vehicle to 1:10 in 4 player and 1:5 in 6 player+. They don't occur in lesser games. You should also be aware that humans normally don't have the all seeing eye to figure out that exactly comp 2 is making a vehicle tech and making a rush with 2+ factions. I personally don't rush in 6+ player battles. If the allies support (and they do now very well) you can sacrifice lots of troops for nothing. The AI is also very tough in rebuilding itself at distant locations now. If we would remove all build programs with issues, we would have to remove all melee programs, because of their turret attack weakness, all vehicle builds and all allround builds (they build melee too) and only leave the ranged build. The AI can handle them better with dancing and can keep some distance to dangerous locations. But that would be rather boring...

[Excedrin]

Yea, I tried restricting vehicles to 800 military or so. In the early game it did better because it had more infantry, but then it lost it anyway and ended up with a single dread at a time. So I'll have to agree that removing the vehicle spam hurts it. But I suppose that it needs to simply play better in the early game, stop losing all infantry and handle melee troops better. Having vehicle spam as an option covers up the real issue, which is it's not so great at managing infantry (ASM jump in independently from the ranged stuff then immediately retreat, it makes too much heavy weapons, it'll frequently get caught with one squad vs 3 or FC vs an entire army, etc).

[thudo]

that humans normally don't have the all seeing eye to figure out that exactly comp 2 is making a vehicle tech and making a rush with 2+ factions.

Sometimes I wonder if Excedrin does.

Anyway.. all testers wait for Beta10 tonight. Its a doozey!