66 replies to this topic

[ThetaOrion]

I really liked Inquistor's and LarkinVB's reports this go around.

Thank you guys!

It's nice to see such competent people in the saddle giving it all a go!

[ArkhanTheBlack]

I have just had a look at the code in the necronbuildbasestrategy.ai file. In the function BuildFlexible() there are 4 possible research patterns that the NL upgrades will follow. All of them have at least one passively visible power (see my previous post), ergo none of the patterns were met. My guess is that m_iNecronLordResearchPattern would be null/0 (uninitialized), when the function is called since this would result in NO upgrades. I could not find any write to m_iNecronLordResearchPattern in all of the source files, but I only have a limited understanding of the workings, so that may be right.
I have two suggestions:
1. One of the strategies (or a new one) should be default (if m_iNecronLordResearchPattern == 0), so that it can never ever occur that the NL does not get upgraded.
2. Phylactery should be part of all or at least most of the cases, since it is always picked by players (I have seen a lot of pro replays).

Sometimes they research their abilities very late. I had several tests where the AI chose the phaseshifter build and it took a long time until the AI researched the next research and phase shifter was never activated since the NL never got below 30% health. It then looked indeed a bit fishy.
But I've seen now enough games where they used their abilities, so I know they work. I'll think about phylactery...

[Inquisitor]

Sometimes they research their abilities very late. I had several tests where the AI chose the phaseshifter build and it took a long time until the AI researched the next research and phase shifter was never activated since the NL never got below 30% health. It then looked indeed a bit fishy.
But I've seen now enough games where they used their abilities, so I know they work. I'll think about phylactery...

It must have built them extremely late (i.e. closer to never), because the game lasted over 40 minutes and the AI was never attacked in its base and none of its LP were ever threatened! I suggest that they be given a higher priority if possible, since they are picked by all good Necron players very early. Some human players even build the Forbidden Archive before the Summoning Core!! Besides, the vanilla 1.2 AI always equips them as early as possible.

I have a couple of questions about other stuff I saw in the code (sorry, but being an engineer and C#/C++ developer I just can't help it! ):

• Why is the Wraith the only detector Necrons have? (in necronbuildbasestrategy.ai)
  

  -- Add detector units (Best first, worst last)
  	self:AddDetectorUnit("necron_wraith_squad")
  

• Why do the function FindClosestFriendlyBaseOrStrategicPoint() return the the base when there is a closer strategic point in NecronInfantryTactic:MoveToDisengage()?
• Have you tested if CanJump() returns true for NW when the Summoning Core has been built?

Thanks!

[ArkhanTheBlack]

It must have built them extremely late (i.e. closer to never), because the game lasted over 40 minutes and the AI was never attacked in its base and none of its LP were ever threatened! I suggest that they be given a higher priority if possible, since they are picked by all good Necron players very early. Some human players even build the Forbidden Archive before the Summoning Core!! Besides, the vanilla 1.2 AI always equips them as early as possible.

I'm all ears for good combis. The AI picks them as soon as the archive is built but some abilities are tier 2 or 3 so they won't be picked so fast.

Why is the Wraith the only detector Necrons have? (in necronbuildbasestrategy.ai)

It isn't, but it's the only detector that needs special code to be built.

# Why do the function FindClosestFriendlyBaseOrStrategicPoint() return the the base when there is a closer strategic point in NecronInfantryTactic:MoveToDisengage()?

I fear I don't understand your question...

# Have you tested if CanJump() returns true for NW when the Summoning Core has been built?

You can trust me, the NW teleport does not work. You've no idea what we tried in 2.0 - 2.2 to get it working.

[Inquisitor]

It must have built them extremely late (i.e. closer to never), because the game lasted over 40 minutes and the AI was never attacked in its base and none of its LP were ever threatened! I suggest that they be given a higher priority if possible, since they are picked by all good Necron players very early. Some human players even build the Forbidden Archive before the Summoning Core!! Besides, the vanilla 1.2 AI always equips them as early as possible.

I'm all ears for good combis. The AI picks them as soon as the archive is built but some abilities are tier 2 or 3 so they won't be picked so fast.

Don't get me wrong about this. I just highly doubt that the AI picked any. I think they're essential to playing Necrons, so that's why I'm so persistent about it.

About good combis:

• The Phylactery is always picked as it makes the NL resistant to shooting and heals him faster.
• Solar Pulse is almost always picked by humans, since it is T1 and very useful for getting those slow NW in range without them being gunned down.
• The Lightning Field is always picked if the opposing race is strong in CC, because it protects the NL from CC damage. The damage it accumulates is not capped in 1.2, so it can deal insane amounts of damage if not used through the whole game (more here). I have seen it take out Defilers, preds, an undamaged Chapel Barracks... It only has a very short range (can't target large objects) and I don't pick it because of the damage it can deal, but for the protection it offers the NL. CC-oriented human players often refrain from attacking the NL with LF because 90% of the damge they give him is reflected back on themselves.

Typical combis by ladder players:

• SP, Phyl and LF (A good alround combi that gives maximum protection to the NL)
• SP, LF and Phase Shifter (because if PS is enabled when in CC, the NL heals himself while those attacking him get damaged by LF)
• Chronometron, SP, Phyl (feared by ranged armies like Tau, because the NL teleports in, casts SP and Chronometron and the slow FW cannot get away from the advancing NW)

Never picked by pro players in 1.2:

• Ressurrection Orb: Too expensive, raised army is only at 30% HP, only 24 cap, and it often maxes out you inf by NW and FO, which you probably don't want anyway.
• Veil of Darkness: Too expensive, and most good players know to bring spotters and thus offset it completely.
• Nightmare Shroud: Too expensive, and it is extremely seldom the the Necron player wants the army to run away from him. Most of the time goes by catching it anyway...

Why is the Wraith the only detector Necrons have? (in necronbuildbasestrategy.ai)

It isn't, but it's the only detector that needs special code to be built.

Oh, alright. I knew there was something I missed!

# Why do the function FindClosestFriendlyBaseOrStrategicPoint() return the the base when there is a closer strategic point in NecronInfantryTactic:MoveToDisengage()?

I fear I don't understand your question...

It is the behavior you see in the replay. There are much closer Strategic points with buildings, but the army disengages all the way back to the the base. There must be something wrong in the detection of wether there is a builing on the Strategic point or something. That's all.

# Have you tested if CanJump() returns true for NW when the Summoning Core has been built?

You can trust me, the NW teleport does not work. You've no idea what we tried in 2.0 - 2.2 to get it working.

Alright... Just checking!

[ArkhanTheBlack]

Okay, I'll try to change the NL abilities according to your suggestions. Since I don't want to skip abilities completely, I'll try a 2 good and 1 not so good approach with several patterns. We'll see if the problems still persist.

It is the behavior you see in the replay. There are much closer Strategic points with buildings, but the army disengages all the way back to the the base. There must be something wrong in the detection of wether there is a builing on the Strategic point or something. That's all.

This behaviour is intended. The closest point is not necessarily the best point. Usually you want something like a near save point which is also close to your base. I therefore use a system of attraction points to solve this problem. It's very obvious if you look at the input parameters of the function:

CpuManager:FindClosestFriendlyBaseOrStrategicPoint(vFromPos, vCloseTo, bNoThreat, bNoThreatOnPath, bMine)

[LarkinVB]

This behaviour is intended. The closest point is not necessarily the best point. Usually you want something like a near save point which is also close to your base. I therefore use a system of attraction points to solve this problem. It's very obvious if you look at the input parameters of the function:

Would be nice if it can be optimized so slow necrons will choose one near the attack point if they retreat to minimize their slow back/forth crawl distances.