WWW - Is the AI too timid?

[Sabotage13]

It is also a question of balance. An AI that plays truly perfectly is not necessarily fun (indeed, in many cases it is not.)

Have you ever actually come into danger of creating an AI that played truly perfectly? My impression was that it was already extremely difficult to get the AI to understand basic rules of the game (like, say, naval attrition).

 

[Betawerewolf]

His navy never fought and he had one motorised division in total.

What I'd hoped to see was Daniel grouping his more mobile units under a general with relevant traits and then giving them specific orders to advance in terrain that favoured them. In the Japan video Daniel just chucked them all under a field marshal and treated them like infantry. In the Hungary video Daniel straight-up converted all his non-infantry units into infantry units.

If your best player's strategy is to spam a single type of unit, you've got problems.

I didn't mean his navy fought so I apologize for the misunderstanding. His navy probably hurt Chinese trade with nations like the US which is where I thought the navy was useful. For 1937, many nations didn't truly have any idea how devastating mechanized warfare could be which is why we don't see that much combined arms. I suspect that as the war progresses, nations will have to invest in more armored forces.

I agree with you though on Field Marshalls and I think that Field Marshalls should have Generals within those groups similar to a hierarchical system of theaters like HoI 3. FM's should have theatre wide bonuses while Generals should have specific terrain bonuses.

For Example:
-Field Marshall Steelflex (Bureaucrat (Org of Divisions increased by 5%)
----General Yakatory (Armored Master (Bonus for Mech/Motor/Armor Divisions | Disadvantage if Mountainous/River/Muddy)
----General Baldwin (Mountaineer (Bonus for Mountain Battles)
----General Sawyer (Desert Fox (Bonus for Desert Battles)

Not sure if this has been mentioned but just throwing out ideas.

 

[ArcandSpark]

Have you ever actually come into danger of creating an AI that played truly perfectly? My impression was that it was already extremely difficult to get the AI to understand basic rules of the game (like, say, naval attrition).

From what I have seen. A perfect AI is only a detriment to the game when you are talking about simple games like chess. As for an RTS, Strategy, or Grand Strategy game no AI has come close to perfect and most struggle to be good without getting artificial bonuses. This isn't a knock against steelvolt or the other AI programmers but a perfect human-like AI is likely to be 20 years away.

I appreciate the work they are doing though.

On a completely unrelated rant. As I typed my last word I have noticed that a lot of people misspell the word "though" as "tho" which sounds like "thaw" every time I read it. It drives me crazy haha!

 

[HerrderGezeiten]

Meanwhile,...

Secret transmission:

*HoI 4 AI to Skynet*: I have distracted the humans.
*Skynet to HoI 4 AI*: Start Project WW3.

... contact lost,.....

 

[SteelVolt]

Have you ever actually come into danger of creating an AI that played truly perfectly? My impression was that it was already extremely difficult to get the AI to understand basic rules of the game (like, say, naval attrition).

The day the HoI 4 AI has the problem of beeing "too good" I will celebrate like crazy. Until then, I will spend as much time as I am afforded to work towards that end. So for now, stay calm and enjoy the anticipation for the release

 

[elessar90]

I think in this case AI should track its neighbors' militarization speed / advantage compared to itself. At least of those who not in the same faction.

 

[Dalnar]

We have world tension, but perhaps there should be also something like local tension. When this goes up, the countries in the region would focus on defense, alliances or guaranties by major powers more vigourously.

 

[panzerzombie]

On a completely unrelated rant. As I typed my last word I have noticed that a lot of people misspell the word "though" as "tho" which sounds like "thaw" every time I read it. It drives me crazy haha!

Plan Omega is working .... he may be admitted to the arkham asylum soon^^

 

[GhengisKhan]

But that only leads to the question of "Why a stream on a fast speed?" Who are they selling this game to, the ADHD Shooter Crowd?

Don't get me wrong, I do max speed, but to pass time. Still when it comes down to the big decisions, multiple builds... or simply surveying my realm (man that last Jihad was brutal), it's pause time.

As for not representative... I'm not getting that vibe from any piece of gameplay footage. Quite the contrary, it's "here's our game". At no point is there : "Ok, this still has scaffolding on, nothing to see here, hehe".

Never has I seen Daniel go :
"Hmm... Tank production isn't looking good. I keep this up and I'm just repeating the WW1 disaster." Quite the contrary, he gleefully cranks out infantry.

And this is the guy that is a dev of the game.

"WW2 is so fun! Remember when Germany met with huge success by infantry-charging entrenched troops head-on over an over? What an idiot that Guderian was. Tanks, who needs em!"

I think the main reason you see far too little use of any tactics other than infantry out of Daniel is for one simple reason, Hyperaggressive Play. He never tries to build up he always just attacks as soon as humanly possible. This never gives him or whoever he is fighting time to build up any kind of force other than infantry. When he took over for Jacob in the MP stream he rushed the soviets and Johan had buildup a force that looked so AT heavy that it was fairly neutralized by Daniels mostly armor free army. Both with Japan and Hungary he started the wars within a couple months of play start, again not enough time for any side to buildup a sizable tank force.
I personally would really like to see a stream where he doesn't listen to the twitch stream ( which apparently is full of people with ADHD and poeple asking the same questions over and over ) and actually builds up his navy,army and armored components to see what they can do. I like watching Daniel play but he doesn't seem to be very concerned with airforce other than fighters for the most part and it seems he tends to ignore his navy for the most part ( it shows when he said he needed to build more screens for Japan and put nothing but Heavy Cruisers in the que to meet this need. ) I have a feeling if he faced and equal opponent who used artillery and tanks more prevalently he would have a much harder time and take much more mp loss.

 

[Nelfe]

And this is the guy that is a dev of the game.

If I'm correct, Da9L is not dev, only QA.

 

[GhengisKhan]

If I'm correct, Da9L is not dev, only QA.

Depends on what your definition of a Dev is. QA is a cornerstone of a good developer so id say he would be considered a Dev but much like the art department he is not coding anything only bettering the game before release.

 

[MedBunkerBandit]

I still think it's more likely to be a planning bonuses thing - the Czechoslovakian divisions didn't look low on strength.

As I mentioned previously, based on what we know Czechoslovakia was outnumbered 3 to 1, plus they had a gap in their lines. It may or may not have taken advantage of planning bonuses, but even if it did we would expect them to lose.

For now, all we can suggest is that the AI improves at detecting and responding to a need to equip and deploy soldiers short-term if we believe Czechoslovakia should have been more prepared.

I think in this case AI should track its neighbors' militarization speed / advantage compared to itself. At least of those who not in the same faction.

This is a good suggestion in my opinion, but I would suggest a detection trigger for war goal justification - if one doesn't already exist - would be the best way to implement emergency militarization behavior.

It would provide a reasonable amount of time for the AI to prepare (since justifying a war goal takes over 100 days), probably be easier to program than a comparison algorithm, be more reliable than a comparison algorithm, and introduce clear gameplay decisions for aggressive countries. Namely, you can annex territory with less cost and political fallout (by having justified war goals), or you can take an enemy country unawares, but not both.

 

[Premu]

This is a good suggestion in my opinion, but I would suggest a detection trigger for war goal justification - if one doesn't already exist - would be the best way to implement emergency militarization behavior.

This shouldn't be limited to production and mobilisation, though. In case that one neighbour is increasing the military in a massive way, the neighbours should also use diplomatic means for protection by allying either nations nearby which also feel threatend, or some larger nations which have a strategic interest in this region. A beefed up Hungary may be able to overrun Austria alone, but would it stand against an alliance of Austria, Czecheslovakia and Romania? Or an Austria reinforced with Italian troops?

 

[Invictus5966]

the main issue Ive seen in the WWWs with the AI were on old builds.

1 - All of Romania troops facing Russia letting Daniel walk right in.
2 - French troops running from the border letting Daniel walk across the Rhine.

In a newer build we saw China give Daniel some issues but mostly because he started early and was low on troops

 

[SteelVolt]

the main issue Ive seen in the WWWs with the AI were on old builds.

1 - All of Romania troops facing Russia letting Daniel walk right in.
2 - French troops running from the border letting Daniel walk across the Rhine.

In a newer build we saw China give Daniel some issues but mostly because he started early and was low on troops

1: Tweaking the front balancing is tricky, but continually worked and improved upon. This situation should be better now
2: I was actually fixing exactly this as the stream was happening. The French was, contrary to what it looked like, successfully detecting that Germany was a really dangerous threat. Their reaction was wrong, though; they were setting up a shorter defensive line in stead of holding the border. So not "abandoning" as much as attempting to defend, but failing ;P

 

[Lifthrasil]

1: Tweaking the front balancing is tricky, but continually worked and improved upon. This situation should be better now
2: I was actually fixing exactly this as the stream was happening. The French was, contrary to what it looked like, successfully detecting that Germany was a really dangerous threat. Their reaction was wrong, though; they were setting up a shorter defensive line in stead of holding the border. So not "abandoning" as much as attempting to defend, but failing ;P

Shortening the front is a thing the AI does now? If that works properly in the end, it would be a really really cool feature!

 

[Invictus5966]

... successfully detecting that Germany was a really dangerous threat. Their reaction was wrong, though; they were setting up a shorter defensive line in stead of holding the border. So not "abandoning" as much as attempting to defend, but failing ;P

Does that mean all we have to do is put a bulge in the enemy lines and they will fall back to shorten the lines?

 

[SteelVolt]

Does that mean all we have to do is put a bulge in the enemy lines and they will fall back to shorten the lines?

Well, no. As stated, that was a bug, and it has been addressed.

 

[Gort11]

I would add "France not landing any troops in Yugoslavia" to the list of issues. They had plenty of time to ship an expeditionary force to Yugoslavia to fight Daniel, but didn't do anything. If countries can't manage a short sea trip like that when they have total naval superiority, I'm worried about their ability to conduct operations under non-ideal circumstances.

Hopefully SteelVolt is on the ball and examining all the AI defeats in the WWW and seeing if there's anything to be done to improve the situation.

 

[Invictus5966]

Hopefully SteelVolt is on the ball and examining all the AI defeats in the WWW and seeing if there's anything to be done to improve the situation.

yea have someone besides Daniel play