Matrix Games Forums

Hi all,

ORIGINAL: jwilkerson

Right, the main "deathspiral" I think we've been workin on - over the past 6 months - is preventing the AI from getting stuck and doing "nothing" from some point forward. Now if you send everything including the kitchen sink to Java (as I did in a recent test) you can still (well could then anyway) stall the AI out at Java, but it will keep going in Burma, China, Solomons, CENPAC, etc. A good human player will usually be able to beat the AI, but the AI did kill my carriers in two prior games, so it isn't a "walk over" by any means.

Great news Joe! [&o][&o][&o]

BTW, when WitP was first published my fruiend Oleg (he posted quite often here) made few AI vs. AI tests at work (after he went home during the night) for many many days...

The original WitP AI (both Allied and Japanese) almost perfectly followed the historic WWII in the Pacific!

How is that nowadays with WitP-AE?

Did you guys run such long AI vs. AI games to see what is happening (and with ocassional "break into" to see situation, positions and stocks)?


Leo "Apollo11"

This may get posted as part of an ongoing AAR, but it is a little rare and I thought it would make a nice Christmas present:
 
Sub attack near Port Moresby  at 98,130
 
Japanese Ships
      DD Mutsuki
      DD Yayoi
      E Asanagi
      E Oite
      CL Yubari
 
Allied Ships
      SS S-42, hits 19, and is sunk
 
 
 
SS S-42 launches 4 torpedoes at DD Mutsuki
S-42 bottoming out ....
DD Yayoi fails to find sub, continues to search...
DD Mutsuki fails to find sub and abandons search
E Asanagi fails to find sub and abandons search
E Oite fails to find sub and abandons search
DD Yayoi attacking submerged sub ....
DD Yayoi attacking submerged sub ....
SS S-42 forced to surface!
DD Mutsuki firing on surfaced sub ....
E Asanagi firing on surfaced sub ....
E Oite firing on surfaced sub ....
CL Yubari firing on surfaced sub ....
DD Mutsuki firing on surfaced sub ....
E Asanagi firing on surfaced sub ....
E Oite firing on surfaced sub ....
CL Yubari firing on surfaced sub ....
Sub slips beneath the waves
 
 
 

ORIGINAL: jwilkerson

ORIGINAL: Andy Mac
I think we have licked the old AI deathspirals

Right, the main "deathspiral" I think we've been workin on - over the past 6 months - is preventing the AI from getting stuck and doing "nothing" from some point forward. Now if you send everything including the kitchen sink to Java (as I did in a recent test) you can still (well could then anyway) stall the AI out at Java, but it will keep going in Burma, China, Solomons, CENPAC, etc. A good human player will usually be able to beat the AI, but the AI did kill my carriers in two prior games, so it isn't a "walk over" by any means.

Excellent and appreciated descriptions with an obvious knowledge of past problems guiding the cure of those problems.Thank you.

ORIGINAL: Barb

How many main battery salvoes could be fired by old battlewagon until magazines become empty IRL? Has someone any idea?

I think they expected about a hundred shots per gun.

The USS Idaho was ordered, as a test on 10 October 1942, to "fire to exhaustion" by Admiral King. The account is detailed on page 352 of U.S. Battleships, by Friedman, in five paragraphs. I don't feel like typing them all in their entirety now... sorry. heh

Highpoints: 20 6-gun, 47 5-gun, 22 4-gun, 31 3-gun, 25 2-gun, 11 1-gun salvoes were fired (only the two forward 3-gun turrets were used).

The average salvo time was 67 seconds, but by the last 20 or so salvos, the rate of fire slowed down to once every two minutes, due to the slow arrival of powder or shell from the magazines (the crews of which were manually moving the bags/shell to the hoists).

2400 bags of powder were handled without incident, and each shell in the magazine was "walked" to wherever they needed to go to get to the hoists by rocking and turning the shells so that their points moved in a horizontal circle about 6 inches in diameter. [X(]

ORIGINAL: m10bob

ORIGINAL: jwilkerson

ORIGINAL: Andy Mac
I think we have licked the old AI deathspirals

Right, the main "deathspiral" I think we've been workin on - over the past 6 months - is preventing the AI from getting stuck and doing "nothing" from some point forward. Now if you send everything including the kitchen sink to Java (as I did in a recent test) you can still (well could then anyway) stall the AI out at Java, but it will keep going in Burma, China, Solomons, CENPAC, etc. A good human player will usually be able to beat the AI, but the AI did kill my carriers in two prior games, so it isn't a "walk over" by any means.

Excellent and appreciated descriptions with an obvious knowledge of past problems guiding the cure of those problems.Thank you.

To amplify just a touch on Joe, testing is pretty rigorous. You got the Coral Sea AAR going, there's other H2Hs going on; play against JP AI, play against Allied AI, all multi day turns, and an ongoing AI v AI, all designed to get to 1944 so as to slow down a bit and wring out anything that crops up in those "later years" that not many people get to.

Also, there's a ton of sandboxes for this & that. We got LCI(R)s that actually work! We've diddled with multiple matrix invasions from a Bn on a Div, 1, 2, 3 Rgts hitting 3, 2, 1 Rgts, to a Div on a Bn, with forts from 0 to lots. Woof !

My personal timer is on how long it takes the JP economy to come to a grinding halt if merchies aren't replaced. I'm assuming lots of high value merchies were converted to auxiliaries (in the historical course), making the remainder more interesting as sub targets.

Taking bets.

ORIGINAL: jwilkerson

ORIGINAL: Andy Mac
I think we have licked the old AI deathspirals

Right, the main "deathspiral" I think we've been workin on - over the past 6 months - is preventing the AI from getting stuck and doing "nothing" from some point forward. Now if you send everything including the kitchen sink to Java (as I did in a recent test) you can still (well could then anyway) stall the AI out at Java, but it will keep going in Burma, China, Solomons, CENPAC, etc. A good human player will usually be able to beat the AI, but the AI did kill my carriers in two prior games, so it isn't a "walk over" by any means.

Given that you're updating a game rather than redesigning it, I suspect death spirals are unavoidable. My experience in chess is that strategy is like tactics, just more general. That suggests the basic approach to AI in WiTP II should be to work on a number of levels of abstraction, such as:

1. The first question is the strategic direction. Top-level HQs are assigned to strategic directions.
2. The second question what objectives in the chosen direction.
3. The third question is the order of operations (including bypasses, alternatives, and switches) in the chosen direction. Operations in secondary directions need to be identified as well. This can probably be programmed.
4. The fourth question is the required resources for the operations.
5. The fifth question is how to organise each operation.
6. The sixth question is the tactics to be employed for each operation.
7. The seventh question is the real-time management and co-ordination of each operation.

From time to time, each level of decision is revisited.

The Japanese error was to misjudge the answer to question 3 for a secondary direction where the real goal was defensive. So make the surprise attack on Pearl Harbor a high-probability alternative operation with an unexpected side-effect on the length of the war.

Have the effects of torpedoes been adjusted to more accurately reflect real life results?  There are many, many RL examples of merchant ships sinking with only one torpedo hit, but there have been very few instances of that in my game vs the AI.  Similarly, DD's weren't all that resistant to single torpedo hits either, while cruisers could absorb one Long Lance torpedo without sinking (although two tended to sink them).

ORIGINAL: John Lansford

Have the effects of torpedoes been adjusted to more accurately reflect real life results?  There are many, many RL examples of merchant ships sinking with only one torpedo hit, but there have been very few instances of that in my game vs the AI.  Similarly, DD's weren't all that resistant to single torpedo hits either, while cruisers could absorb one Long Lance torpedo without sinking (although two tended to sink them).

The damage model in the game is quit conservative in the damage level per weapon hit.

IRL 1-2 bomb hits would damage a merchant beond repair, In WITP most merchants shrug of a single bomb hit. To compensate for this low level of damage per bomb, the amound of hits that are scored in game are also higher than IRL (until a certain exp level is gained, then a miss has a lower probability than that of a hit. Luckely this is fixed in AE).

I asked this question some time ago, and from the answers I understood that this high hit/damage ratio was done deliberatly to "spice" up the game.

IIRC, one of the reasons for merchant ships ability to shrug off damage is that a ship's durability (sort of like their "hit points" to borrow an RPG term) is based in large part on displacement, but the stock game does not factor in their basic construction.  Warships have much better compartmentalization versus a merchant ship, so they should have more "hit points" per unit displacement than merchant ships, but in stock at least, that is not the case.
 
I think that some of the mods played around with durability to try and make merchants more fragile, but I think there were unintended consequences.  I can't remember seeing if anything was done with this for AE or not.
 
Warship survivability depended a lot on where the hit was, at least for torpedoes.  Hits in the bow were typically less of a problem, although US Cruisers had a tendency to loss the whole bow from single hits.  I have read of some DDs that were so thin at the bow that a torpedo would pass right through and explode on the other side. Hits in the rear were bad for cruisers, IIRC, since the penetrations for the propeller shafts tended to weaken the overall structure of the hull, but relatively, a little better for DDs since it restricted the damage to one end.  Hits in the middle could break a DD in half, but in a cruiser were more likely to only flood several compartments (very possibly in the engineering spaces).  They might need a drydock, but at least they could still operate.  Of course, a TT hit near a cruiser's turrets had a chance of a magazine explosion, which would be almost certainly fatal (or severely damaging at least).

[Deleted by Admins]

being brutally honest for once it varies - if I have thought of it and programmed a counter then you will be met by opposition but the further out the game goes the harder it is to forsee every eventuality or more importantly the strength of the option

The AI does not think and pick an appropriate response it uses the forces I tell it to to counter an invasion.

So if I misjudge the likely time or force then t could be a sledgehammer response or possibly insufficient force

All the likely regions have counterattack forces specified but will they be enough only time will tell

p.s. I cannot and will not over promise on the AI I hope its better than stock but as good as is all we said it would be so I really don't want to build expectations to high

[Deleted by Admins]

HEY ANDY MAC Please email me - thomaslukemills@gmail.com

I tried your email link but it bounced back and your PM is full

ORIGINAL: Grognerd

That's all anyone can ask for! Thanks for the response. Can't wait to try it! PS I have two PBEM friends waiting too.

With all the improvements being made there is no way the AE AI could be worse than stock! And asking for a better AI than stock is definitely not asking too much!

How sonar will be factored in AE?

No change.

Thanks.

ORIGINAL: Dili

How sonar will be factored in AE?

yeah, my bathothermometers got axed too!

[:'(]

Well sonar was a crucial device in ASW. So it should have been there and could be just a bonus tick in ship anti-submarine detection capacity. Since i don't know maybe it required to much resources to do it. Not perfect but one option is to edit anti-submarine weapons making w/ sonar and without sonar versions.