AI improvements

[spaff]

I'll post a couple suggestions for AI improvement based on my single player gameplay. Feel free to add more suggestions. I know AI is difficult to program but I think some of the issues could be easily fixed.

Airforce:
-AI planes are very easy to destroy with grouping planes and destroying AI planes one by one. At the start of a scenario this isn't a huge problem but after AI only has a couple planes they basically become suicidal by sending single planes without any support.
- This could be fixed by having AI use groups of planes and use planes defensively when AI has lost air superiority. Also AI should use AA units for support when it doesn't have air superiority and bring fighters back when enemy planes are already damaged.

Ground forces:
-AI forces take huge losses while attacking bottleneck defenses (bridges, entrenched cities/forests etc.). Especially during river crossings AI takes huge losses while not getting anywhere.
-AI should use more artillery during these situations and not send forces attacking against superior defenses. Another option should be flanking while bombarding the defenses with planes/artillery.

[JB72]

Also - The AI shouldnt drive with truck loaded Infantry into the enemy - they do this every time - is like on the slaughter !

[Horst]

It doesn’t necessarily need some recoding of the AI, but rather a couple of different scenario design tweaks:

Air force: equalizing the expected fighter force for each scenario should already be enough balancing. For later reinforcements, two fighters instead of a single one could also work more challenging. The historical context can surely also play a role how strong an air force should be or not.

Ground forces: scenario designers often forget about the cheap and very common infantry guns used in every army. One infantry gun and one field gun for every three (heavy) infantry units is a fine, plausible distribution ratio. Heavy artillery is optionally added as extra support like it is usually historical done.
Artillery is indeed a decisive factor how strong an opponent is.

If AI units get a move order, they typically move at highest speed if there is no threat spotted. This can often lead to unfortunate situations like the very fast AA-vehicles racing into defensive positions first. As long as enemies are spotted nearby, the AI is at least smart enough to keep some distance with squishy unit types or moves carefully without transports.
What the AI needs is a good recon. With enough recon units on the ground and in the air, fewer accidents should happen.
I’m still glad there is hardly any cheating here by spying into the fog-of-war, so the player can cheer about every successful ambush they accomplish.

[GabeKnight]

Sorry to say, but again I have to agree with Horst here. The mentioned AI "issues" are rather (debatable) scenario design decisions.

Personally I'd be glad for an enemy airfield check, so the AI finally stops landing planes on airfields that are "contested"...
Or a better target-finding-algorithm for battered player's units. Also it would be nice if a fully entrenched AI unit stopped moving to the neighbouring hex before an attack.
But again, many AI "misbehaves" may be attributed to scen design.

[Horst]

No one needs to feel sorry for agreeing with me.

Yeah, this contested airfield issue. I have to test AI aircraft (mis)behavior next if this "set base" possibly affects this or not. Air Patrol can work without set routes too. As designer I'd always pick the airbase furthest away from the front lines. If this airfield falls then the game is usually over for the AI anyway.

[GabeKnight]

I have to test AI aircraft (mis)behavior next if this "set base" possibly affects this or not.

Didn't you or Bruce test the "set base" mechanic once already? There's a vague memory of reading about it - with the overall conclusion: "makes no sense / useless / will be overridden by the AI anyways".

As designer I'd always pick the airbase furthest away from the front lines. If this airfield falls then the game is usually over for the AI anyway.

Don't know about that. Don't forget that the vanilla planes are rather slow with very limited fuel. You have such an example in WinterWar's "Raate Road" scen, with its rather longish map and air-exit-hexes on the edges - no airfields. After the planes finally meet each other in the middle, they have to turn back already to refuel...
I'd rather have the AI change the base according to the battle.

[spaff]

Yeah, I wouldn't know if the points I mentioned are because of scenario design or AI programming. I'm just saying that I have played first half of endsieg and I have destroyed soviet airforce in every scenario with very simple plan of grouping my planes and destroying enemy planes. This makes gameplay boring because I can just fly all over the map because there are no soviet planes and very little AA.

[GabeKnight]

I'm just saying that I have played first half of endsieg and I have destroyed soviet airforce in every scenario with very simple plan of grouping my planes and destroying enemy planes.

But this comes with a cost. Usually your average core airforce consists of a "healthy mix" of dogfighters, tactical and/or even strategic bombers. To achieve air superiority in the Endsieg missions, you probably had to limit yourself to fighter planes only.

[spaff]

But this comes with a cost. Usually your average core airforce consists of a "healthy mix" of dogfighters, tactical and/or even strategic bombers. To achieve air superiority in the Endsieg missions, you probably had to limit yourself to fighter planes only.

I replayed first endsieg scenario and I destroyed soviet figthers after buying 1 fighter and 1 AA unit.