Admirals Edition Naval Thread II

[Alikchi2]

Thanks for the insight - really, it's fascinating to hear how the guts of the game work, and very encouraging that you've done your best to tame the beast [&o]

[Andy Mac]

Of the two jobs I had on AE - Scripting AI has taken most of the last year doing the land OOB's took about 3 months.....

If we had started with the tools and AI kit we have now I estimate a full AI for the whole game would take about 2,500 man hours and then it would be ready for initial bug checking and testing.

The scripting engine has taken a LOT of time - I have started from scratch on the AI at least three times and the number of times we have to test to find a workaround for something AE just will not do is a crazy number

There is more I would want to do but most of those enhancements will need to wait for a patch

[bradfordkay]

ORIGINAL: Don Bowen

ORIGINAL: Dixie

ORIGINAL: Don Bowen


I have a full cup of coffee left and the heater needs to run the temp up a couple of degrees before I take my shower and go out. Someone ask a question...

Why do you never see baby pigeons?

42

After crisscrossing the world in search of the answer to life's mysteries, I found it in the middle of a highland marsh in northern Idaho:

[CV Zuikaku]

Does it even exist any possibility, maybe for some future patch, that surface TFs woul'd have an option similar to air groups? Do not use torpedoes. Sometimes it is really frustrating to see 10 DDs launching 40 torpedo spread from long range on the single AK. Or to be even able to choose if torpedoes are going to be used from long, medium or short range? Sorry if that was already answered on locked naval thread, and thanks in advance!

[jwilkerson]

ORIGINAL: Andy Mac

... the land OOB's took about 3 months.....

But understand this is because many people (including Andy) had been working on said "Land OOBs" for many years already. We did NOT sling these land OOBs (or Naval or Air) together in anything like "3 months" [:D][:D][:D]

[Grotius]

Strategic is now scripted via the editor (you'll love it!).

I can't wait to play with this! I don't recall many games in which AI scripts are exposed to the editor like this. This might actually induce me to try my hand at an AE mod (someday).

Everything from moving a unit from Miri to Brunei all the way up to KB showing up. But down under this layer, there are plenty of places where the code tells ships what to do, and those places are the realm of our Naval Team coder (a.k.a. Ex-missle-man-Don).

Fascinating. Have you found it easier to code the "strategic" layer or the "operational" layer? My guess is that the strategic layer is harder to test, if only because its consequences unfold more slowly. With the operational layer, presumably you can run a few turns to see if the AI creates and executes a Bombardment mission competently. With the strategic AI, you might have to play a few weeks or months to see if it's working well.

I have started from scratch on the AI at least three times

I enjoy PBEM as much as the next person, but I've probably spent the majority of my WITP time against the AI, so I for one appreciate all the work going into it.

[Andy Mac]

Oh gawd no I was talking the input and reinput and then the reinput yet again as we figured out what we wanted to do  - research took a lot longer

I really don't even want to think about how much I spent on books for this project - it makes me shudder [:D][:D][:D][:D] - I ended up going to the Imp War Museum reading room to get hold of some of the more obscure Indian Army Books

[Grotius]

Gotta love the Imperial War Museum. One of my favorite museums anywhere.

[Don Bowen]

ORIGINAL: CV Zuikaku

Does it even exist any possibility, maybe for some future patch, that surface TFs woul'd have an option similar to air groups? Do not use torpedoes. Sometimes it is really frustrating to see 10 DDs launching 40 torpedo spread from long range on the single AK. Or to be even able to choose if torpedoes are going to be used from long, medium or short range? Sorry if that was already answered on locked naval thread, and thanks in advance!

No such option exists, but the value of a target is now considered more closely before torpedoes are launched. Probably not suffcient for what I think you want, but all we got.

[Don Bowen]

ORIGINAL: Grotius

... With the strategic AI, you might have to play a few weeks or months to see if it's working well.

"Strategic", along with all the work on scripts, belongs to the AI team. I'll leave it for them to answer that part, but I can guess...

As to the "operational" level, it is a mark 1, left handed, bitch. So many variables and controls - player settings, AI overrides, hex type, air balance, mission type, new waypoint, patrol, and new follow/meet, computer vs human player, etc, etc, etc.

[Andy Mac]

Yup lets just say I have 3 PC's running at the moment 2 desktops and a laptop

All three are running AI v AI games overnight. I think my record was 5 AI build in one day all submitted as the final one for that week - then I found another way for the AI to do better

At various points I or others step in and spot things that have or have not worked.

On occasion I will or one of the testers will run 30 - 60 turn from an AI generated position - actually thats a lot of fun - if ever you get jaded run the AI for a year in an AI v AI game then take over whichever side you like.....now thats a lot of fun.

After I tweak and play with with the strategic AI files for a few days muttering to myself about why isnt the Ai attacking java or why is the Ai not bugging out or or or or or

I then run AI v AI to check what I want to happen is happening then I hand it over to the Beta who typically break it - they know all my nasty little tricks by now - I need to come up with some new ones just to throw them off - I could tell the AI to invade Sydney straight away on next build that would confuse em - my motto confusion to the Betas (obviously that would be just to confuse betas and not for final build) !!!!

I then either

a. go back to drawing board about the latest way the testers have figured out to defeat the AI and fix it
or
b. whinge at coders to give me another tool so that I can do a.

and then we go round and round again

occasionally I bypass a. and b. and either I or another team member comes up with a cunning plan for the AI - I typically then go to the coders who craft my little suprise into the AI if possible although we need to be carefull not to hardcode stuff - Don did the last one of those for me I managed to shock a few testers with that little suprise - which I wont spoil for you lot....

Typically though its round and round a. and b.

There is a skeleton of what historically happened and then I take over to 'try' and make the AI do it in such a way that it will at least partially respond when a human player goes off piste

Sometimes it works sometimes it doesnt.

The very best v AI games are the ones where you minimise the players advantages so no PDU AI will never get the usage a player will out of that tool and I would recommend 2 day turms, maybe 3

When I play the AI for real I play 2 day turns, hard difficulty and no PDU. Just now I play 3 day turns as its mostly testing.

I don't want to over promise all I can say is the AI is as good as stock - I personally think its a lot better but then I would say that its been pretty much all my spare time for the last year. I guess only you guys can give me an honest read on whether its better or not and that will be as soon as its ready - promise !!!

[Yamato hugger]

ORIGINAL: Andy Mac

I then run AI v AI to check what I want to happen is happening then I hand it over to the Beta who typically break it - they know all my nasty little tricks by now - I need to come up with some new ones just to throw them off - I could tell the AI to invade Sydney straight away on next build that would confuse em - my motto confusion to the Betas (obviously that would be just to confuse betas and not for final build) !!!!

Doesnt take much to confuse most of us. Andy just sits in his room and giggles to himself most of the time.

[Grotius]

Very interesting, Andy. So how far have you run one of those AI vs AI games? I hope you have found time to test into 1945 -- I've just been reading your AAR with PzB and was sobered by all the problems you guys had with Soviet activation, for example.

Apologies for all the questions. This is what happens when I get a week off work. [:'(]

[doc smith]

Well done, Slartibartfast!

[m10bob]

Most of us who play vs the AI find ourselves criticizing the opponent not for all the dumb things it does, but to a large extent because the more we know of history, we have countered it and pre-empted its' every move.
Any deviation in the programming of the AI will be an appreciated curve-ball.
Perhaps, with the AI playing a conservative, resilient game, maybe the reinforcements given the Allies will no longer be sufficient to turn this into a sure thing by 1943 ?

[Andy Mac]

The longest AI v AI game I have is 1944 - I have one running now with the latest build thats up to late 43 and I will probably try to take it to 45

It takes a loooong time to get to 45.

The issue we cannot with one AI file cater for all types of player some want a faithfull re creation of history, some want a few googlies but basically history and some want all out deviousness like a PBEM would give you.

The fact is we can never do the latter - I have some good PBEM's under my belt against good opponents and I am taking what I can from them and the testers e.g. a tester just pulled me up on being to aggressive with the Japanese opening not because it was ahistorical but because I was leaving a lot of TF's uncovered and allies if they were willing to lose DD's and CL's were causing mayhem so I adjust strengthen screening forces make th eopeing a little more conservative and we retest we go round and round !!!

As things stand we are somewhere in the spectrum of the middle ground with a few land mines for the unwary thrown in will they always work not sure - will it be as good as an all out PBEM probably not but I am 'trying' to be devious with the Japanese AI.

The allied AI doesnt need to be devious it just needs to keep it simple and grind forward so in some ways its easier but it also has to overcome more formidable defences so sloppiness that the Jap AI can get away with because Allied capability is low the Allied AI cannot really cope with so its different problems

I think we have licked the old AI deathspirals but that does not mean that someone out there will come up with a way to stall the AI by doing something crazy and creating a new one.

I am trying to think up every possible way an Allied player could make an unexpected stand and making sure the AI will react but the further out you go the weaker my crystal ball gets

[Barb]

Is there any chance there will be two kinds of bombardment missons for ships?
1. Hit and run (like japs henderson field runs)
2. Stay and hit repeatedly (like US doing multiple day bombardments against islands)

I know the second style I am writing about is possible in WITP, but BB with 9 ammo will fire their 6 ammo points at night, one in a day, and the last two in other turn or so. I would like to see more balanced and ammo conservation missions.
Maybe this could be done with escorts bombard/no escorst bombard switch. With escorts bombard the capital ships will be less willing to spend 2/3 of their ammo in one phase.

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

[jmscho]

I have not checked, but I have a dim and distant recollection that BBs carried about 100 rounds per main gun, but that may be a WW1 figure.

[castor troy]

ORIGINAL: Barb

Is there any chance there will be two kinds of bombardment missons for ships?
1. Hit and run (like japs henderson field runs)
2. Stay and hit repeatedly (like US doing multiple day bombardments against islands)

I know the second style I am writing about is possible in WITP, but BB with 9 ammo will fire their 6 ammo points at night, one in a day, and the last two in other turn or so. I would like to see more balanced and ammo conservation missions.
Maybe this could be done with escorts bombard/no escorst bombard switch. With escorts bombard the capital ships will be less willing to spend 2/3 of their ammo in one phase.

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

as it is now, your BBs won´t ever fire more than the 6 "ammo points" in a bombardment. The remaining three aren´t used for another bombardment, only ammo of your secondary armament. The 3 main armament ammo points are held in reserve for surface engagements. You always only get one effective bombardment from your BB/CA.

[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.