Does anyone play vs. the AI anymore?

[Siberian HEAT]

If you had a choice to pour all your R&D into improving the TOAW engine, would you focus on AI enhancements for solo play, or would you focus more on improving the engine for PBEM play?

I just wonder if "perfect AI gameplay" is too elusive given the current state of technology? Could you envision a time in the near future where the PO is good enough to "think" in the way a real person would?

I don't play against the PO at all, but is that because it is so bad currently? Hard to say...although I personally think there is no substitute for PBEM.

Let me know what you think!

[General Staff]

PBEM. No. Yes. AI in TOAW is an oxymoron.

[Dan Neely]

I occasionally play the AS*, when I'm bored, out of pbems and looking for a few hours of brainless distraction, seeing how fast I can blow the AI away (cuba is great for this). IMO though, putting any effort into the AS is a waste of Norms valuable time, there're many more important things for him to work on.

*Artifical stupidity

[Mad Cow]

I play against the PO in certain scenarios.

It really depends on the scenario.

But I prefer PBEM, by far.

I think that effeortd on the current TOAW should focus on making it a better PBEM experience. If Norm were to ever work on a new operation game like TOAW, then I would hope the PO would be better.

:)

[Chuck?]

No, too much other stuff to do. A game against the PO always leaves me feeling empty and disappointed.

[laszlo.nemedi]

Before I start to PBEM I made a plan of a better AI for TOAW (or a new wargame like TOAW).
In the weekend I translate it to English. Actually I made some test program for it and it works somehow. But some improvement on dividing/regrouping of units very needed for it.

Of course it is far from the perfect general, but very good on maintaining solid frontline, and doing breakthrough assault (meanwhile maintaining that frontline).

[aussiesta]

I keep faith in AI/AS. I reckon Norm should include more different tracks for it, in order to make it less predictable, and maybe work on more specific formation orders for particular scenarios, depicting situations such as guerrilla operations. But I'd say that scenario designers could do a lot more, in terms of finetuning possible AI reactions and operational plans. Very often, low AI simply reflects a poor effort from the scenario designer. Also, I have the untested feeling that AI could do much better with smaller formations in many scenarios. That said, it looks hard to get more than medium AI the way TOAW works. But if you combine a serious in-built planning with attention to detail, such as tripwire reactions, several unpredictable tracks and maybe a small advantage for AI in a given game, well, the Machine doesn't look so dumb then.

[Eric Weider]

i don't sleep with blow up dolls and i don't play the P.O. !

[Martin Schenkel]

Originally posted by aussiesta
But I'd say that scenario designers could do a lot more, in terms of finetuning possible AI reactions and operational plans. Very often, low AI simply reflects a poor effort from the scenario designer. Also, I have the untested feeling that AI could do much better with smaller formations in many scenarios.

I agree. The sometimes depressing skill of the PO IMO often reflects little effort on the designer's part. I think an adequate PO can be programmed, provided the designer spends enough time, and thoroughly thinks it through. On the other hand, the PO just isn't suited to many scenarios, such those with large formations (although sometimes large formations can be used in instances to re-create certain effects), and I don't think even a thorough PO would do very well in large strategic scenarios. But for small-medium scenarios, it just takes more effort on the designer's part to make the PO a bit more realistic.

[LaPalice]

I play against the PO to learn a game, how it works. I had tried Korea 50-51 with TOAW. Otherwise I don’t play against the AI, not because of its weakness, but because it’s more fun to play against a real opponent.
At the same time I would say that developers should try to build a good AI opponent. It can help to learn the game for example.
Instead of playing against the AI, when I don’t have any PBEM turn to play, I play the two sides, as I did when I played non-computer wargames.
And now, with the PBEM possibilities, I must say that I don’t play “traditional” wargames any more, or very rarely.

La Palice.

[LanceRunolfsson]

Yeah More objective trakcs needed and more orders. AIs should in theory be a lot better at some things than people like projecting enemy positions in fog of war enviroments. And anything that has to do with crunching numbers. But there are a lot of basic things in the game that need adressing. There has got to be some way to get the PO to Garrison supply points and Objectives etc.

[aussiesta]

Originally posted by LanceRunolfsson
There has got to be some way to get the PO to Garrison supply points and Objectives etc.

That's the thing with some scenarios - there's a way to garrison supply points and objectives, by using units on garrison orders. Only many designers can't be bothered with that. And that also applies to objectives taken during the game, by simply using the events engine and setting units on garrison duty as soon as such and such location is taken. All the same, I don't mean to criticize designers - it's a hard work and my proposals don't make it any easier.;)

[LOK]

IMHO: Don't waste Norm's time with the AI/PO. There are a million other things he can do to make the game better.
If I can't find a PBEM opponent (has never been a problem) I'd rather play both sides (hotseat) than the PO.
Playing the PO vs. PBEM is akin to virtual vs. real sex :D

[aussiesta]

Originally posted by LOK
IMHO: Don't waste Norm's time with the AI/PO. There are a million other things he can do to make the game better.
If I can't find a PBEM opponent (has never been a problem) I'd rather play both sides (hotseat) than the PO.
Playing the PO vs. PBEM is akin to virtual vs. real sex :D

I agree. There's other priorities in terms of improvements for the TOAW engine. But raising the current AI standards in most scenarios should be a priority for all those who are not Norm. See, there's lots of people who enjoy virtual sex.:p

[LOK]

Originally posted by aussiesta


See, there's lots of people who enjoy virtual sex.:p
I knew I'd be in hot water after I wrote that sentence :rolleyes:

Never tried virtual sex but I can't imagine it can be as much fun as the real thing...

Just like I said playing against another human "general" in TOAW is much more fun than playing against the PO.

[Wolf]

I have never played a single game against the PO, I played about three turns of the tutorial, and dived straight into PBEM:laugh:

[tigersqn]

I haven't played the PO since I joined WHQ and got some PBEM opponents to play against.

[Southern Dandy]

There's certainly some fine-tuning that could be done for the PO, but as has been pointed out by others, most of the onus to do so rests with the respective scenario designer.

I know that I would like to see some enhancements and improvements made...with my work schedule the endless trainwreck that it usually is, the PO is the only opponent that I can count on to put up with my unpredictable time constraints to play this game.

Of course, if I spent less time checking in on the boards here, I could play more....but then I'd miss you guys! :cry:

That's it for now, though I'm sure I'll think of something to add on later....

John

[tklemme]

As a designer, my biggest gripe with the AI is the way it conducts the defense of objectives. Even when given defense orders, a formation will not move to defend an objective (unless it's already sitting on top of it) until the objective is taken by the enemy.

"Sarge, don't you think we should move over there before the enemy takes the objective?"

"Nah. I like to let the enemy take it first, then counterattack. More challenging that way."

It was the same in Tanks! and Age of Rifles. As a matter of fact, the behavior of the AI is so similar I suspect lots of the code is the same.

In any case, it's unlikely that Norm will spend much time on the AI.

Todd

[Fading Captain]

Has anyone ever played through Operation Mars or El Alamein? I've started and stopped both of them several times, because PBEM games keep getting a higher priority. I've gotten maybe six turns into El Alamein, and, if I recall right, it doesn't exactly look like a cakewalk for the human player.

Still I would like to play through both of them someday when I get a good chunk of time.

[GunnerC]

Just a thought - aren't we missing a key point in this discussion ?

Many of the new board members here and at TheBlitz start off by saying 'I've played loads of times against the PO and now want to try PBEM ...'.

If the PO is too awful won't they just get turned off the game before ever trying PBEM? If we want to keep increasing our active playing membership we shouldn't ignore the value of the PO as a recruitment factor.

[aussiesta]

Originally posted by tklemme
As a designer, my biggest gripe with the AI is the way it conducts the defense of objectives. Even when given defense orders, a formation will not move to defend an objective (unless it's already sitting on top of it) until the objective is taken by the enemy.


Maybe other orders should be tried in those cases - I believe the 'screen' order, for instance, could be useful sometimes. But I really don't have a lot of designing experience. In cases of extremely important objectives, reinforcements in garrison mode could be called by using 'activate event' sequences. Not the same but... My point - there's probably ways to make the AI a better opponent, for all those who don't have time for PBEM. The right combination of loss tolerance, orders and programmed reactions could do wonders in small/medium scenarios.

[Menschenfresser]

If we want to keep increasing our active playing membership we shouldn't ignore the value of the PO as a recruitment factor.

Amen!

I also think you can tell the quality of a game's fundamental structure in so far as people will play vs. the AI long after they've mastered it.

Personally, any time I pick up the game outside of PBEM, I play against myself. Sometimes in my more geeky moments, I'll devise two general strategies (like: main axis of attack northwest or main axis southwest.) for both sides and then flip a coin for either.

Speaking of, has anyone every thought of including news events into a scenario which would function as a supreme high command? For instance, when the 3rd ID arrives as reinforcement a series of events is triggered which would tell the player where to send the division. Not sure if you can manage an IF...ELSE clause with the engine.

Dunno, just ideas.

[aussiesta]

Originally posted by Menschenfresser
Speaking of, has anyone every thought of including news events into a scenario which would function as a supreme high command? For instance, when the 3rd ID arrives as reinforcement a series of events is triggered which would tell the player where to send the division. Not sure if you can manage an IF...ELSE clause with the engine. Dunno, just ideas.

It shouldn't be difficult - you mean, like 'OK go take city B instead of city A, since we changed our mind'? I guess all you need is to activate an event that will give you extra points if you take city B and will simulate a Supreme kick in the ass in case you insist on attacking city A. Right?

[Menschenfresser]

In case it wasn't clear, my suggestion was simply an idea toward handicapping human players vs the AI.


Or orders that run on timed events...

Capture city A before turn 3 or there's a 50% chance you will be replaced.
The same for city B before turn 6 or it's a 60% chance.
etc...

Within many scenarios the player does occupy a Stalin or Hitler position, but the smaller ones aren't so, giving the player a few armies or corps, meaning that an authority over the player would exist isn't beyond the bounds of reality.

[Weyoun]

You can't always find a person to play against. I feel that Norm should put a lot more time in fixing up the AI before anything else.. AI is very important and I would like to see it fixed up. :)

[jlbetin]

I will put my little stone in the wall,

Since I play toaw I allways use IA, I never play PBEM. I think there is a lot of player like me round the world which rely on IA only to get fun. So to get a better IA for me and them could be very important

Second point I fight hard to try to realize my 1st scenario, I became crazy as the GUI is awfull, something could be done there

Der wanderer:clown:

[Redwolf]

Slightly beside the topic, but I tried to find a wargame developer who would make an interface so that I could plug my own programmed opponent in.

Disappointment so far, wargame developers are of course protective of their unit database, visibility and cover rules etc, I would need all that. Plus a semi-hefty chunk of development work for the interface, although I can live with an extremly rough one.

Posting here in case anybody has suggestions whom I should ask next.

[aussiesta]

Originally posted by Redwolf
Slightly beside the topic, but I tried to find a wargame developer who would make an interface so that I could plug my own programmed opponent in.


Wouldn't be easier just to play with the events engine and redesign the AI?

[Dan Neely]

IF you're a hot enough programmer, you can do it for any game by processing screen grabs and by driving the mouse/keyboard. :D

[ALF]

I play the PO most of the time, so I hope Norm takes some look at things there. Still in its current state things could be better if designers used the events and features more efficient. Why isn't the 3-way objective path implemented more often? Why not "freeze" formations due to events more often? Why is everything that could ever happen in the scenario foretold in the description or documentation instead of surprising the player by events? There's more to be done here for sure.

[ALF]

Originally posted by Fading Captain
Has anyone ever played through Operation Mars or El Alamein? I've started and stopped both of them several times, because PBEM games keep getting a higher priority. I've gotten maybe six turns into El Alamein, and, if I recall right, it doesn't exactly look like a cakewalk for the human player.

Still I would like to play through both of them someday when I get a good chunk of time.

Those two are the best ones made for TOAW. Scenario designers: play and inspire yourselves!

[Redwolf]

Originally posted by aussiesta


Wouldn't be easier just to play with the events engine and redesign the AI?

I don't understand what exactly you mean here.

I am fine with whatever interface a developer may find suitable: exchanging bunches of ASCII or XML files, a loadable DLL, a CORBA interface, a TCP client like it is being used in multiplayer.

I don't care, except... [see next posting]

[Redwolf]

Originally posted by Dan Neely
IF you're a hot enough programmer, you can do it for any game by processing screen grabs and by driving the mouse/keyboard. :D

This is the only interface I would refuse to use.

I did some projects of this kind (non-game related) in the past and in fact I have a reasonable toolset for this stuff.

However, it is just not practical. You need a lot of extra mouseclick and a lot pf extra scrolls to always keep the game "in sync" with what your engine thinks. The moment you have to scroll the map things become a complete mess.

In all the applications I remote controlled so far I always had to insert code to completely restart the application because the extended number of actions I need bombards them with so many fast-paced mouseclicks that most applications just die within a few hours (which would be weeks with a human clicker). In fact, in my last project I inserted code to restart Windows(!) at regular intervals (my toolkit works over the network, so the remote control stays up when the controlled machine reboots).

The applications are just not reliable and predictable enough. They will go out of sync with your package and you have to restart them. If they are connected to your game in a way that requires that both player sides reload the game this becomes outright annoying.

It is also a substancial problem that the raw speed of reading the screen is not sufficient. I have a fast OCR package for bitwise-exact fonts. However, when the game displays some random text somewhere the human gets the location of the next text for free: the whole screen is already before his eyes without further computation (that's the point about a graphics card) and the visual capabilities allow a human to locate changed (movement in nature terms) regions easily. For my remote control I would have to recapture the screen back into the memory. This is not a fast operation on most modern graphics PCI and AGP cards (some projects which abuse 3D cards for vision projects or general computation also suffer from this problem). In a word: no matter how good and fast my own scanning code is, I have few hope to reliably read bunches of random text appearing on the screen for just long enough that a human can read it. And if I attempted to do so I would trash the game's performance by loading the machine with high-frequency screen polls.

I would have to come up with guessed values for all that movement speed and visiblity stuff, which is obviously dependent on a lot of factors starting from weather. An programmed opponent not knowing about this stuff would not be competitive.

Last but not least I would consider it as an unfriendly act towards the game developer. I would use his/her game in a way that wasn't intended. Since I am snarfing the whole unit database on the way you can imagine that some may not feel comfortable about the idea.

[aussiesta]

Originally posted by Redwolf


This is the only interface I would refuse to use.

I did some projects of this kind (non-game related) in the past and in fact I have a reasonable toolset for this stuff...



From that point on, I didn't understand a thing.
All the programming I've done is in DIV2, a not very sophisticated or famous system, so it's no surprise. It looks to me like you do use really sophisticated stuff for playing, though. Correct me if I'm wrong. I'm really a rookie when it comes to TOAW.
What I meant was that maybe it's just easier, especially for people who don't know much about writing code, just to tinker with the internal TOAW engine to enhance the AI effectiveness. You know - just tell formation A to go one way in case the enemy does something, instead of letting the AI to figure it all out by itself. Do as designers do. Nothing fancy, you see. No strings attached.
By the way, I also think scenario designers should warn people in advance of whatever they may find along the road. Unlike Alf, I hate it when something comes up that was really inexpected - for everyone who hasn't played the scenario before or read the events list.

[Wolfe Tone]

I used to play against the AI all the time right back to when I first got TOAW I. The first PBEM game I played was against Brian “Siberian Heat”. It was Europe Aflame with me as the Axis. We started that back in August 2001 and finished it in IIRC around March last year. BTW he won! Since then the vast majority of games I have played have been PBEM.
I would certainly play more against the AI if it were better. In the Korean War scenario the Reds can be quite a challenge if you set the ‘intelligence’ rating to 3.
At the moment I am playing INSIDE THE REICH: AXIS, which is WWII in Europe. Like EA but not as deep. You basically play as the Axis and see how well you can do. I have reached turn 95 and it’s June 1941 so I am about to invade the SU. I have set the AI rating to 3 to provide a challenge. So far the War has gone more or less along historical lines. Poland fell in 3 weeks. France was invaded in March 1940 and fell by late June. The Low Countries, Norway and Yugoslavia are under Axis occupation.
In North Africa the British and Free French hold Morocco and Algeria and until recently were pushing hard against Tunisia. The city of Tunis nearly fell on a number of occasions. However I reinforced the Vichy French with Italian units and the German Mountain Korps. That front has now stabilized. To the east the Italians got off to a great start by destroying a British invasion force and then advancing to the gates of Alexandria before the British rallied and drove them back across Cyrenicia. Just in time Rommel and the AK arrived to stop the British. The front has now stabilized there after very hard fighting around Brega at the western end of Cyrenicia.
The invasion of the SU is about to begin. I have two Panzer Armies and two infantry ones ready to cross the Border and head straight for Moscow. There is also one strong SS Panzer Corps available in support here too. This is the ‘Schwerpunkt’ of the whole campaign. One other Army will also be used to shield the ‘Drang Nach Moscow’ from Russian attacks from the north. Further south one Panzer Army and two infantry ones with the assistance of the Rumanians and Hungarians are ready to overrun the Ukraine. Hopefully Kiev will be taken early on and I will develop the advance from there depending upon circumstances. On the extreme southern edge of the front one German Army is available to support the Rumanians in their attempt to take Odessa. If that is successful then this Army will seal off the Crimea and support operations to the north and east.
BTW I am using the term ‘Army’ here but they are designated ‘Corps’ which does not fit a formation with 10 to 12 divisions.
Some things could be improved in this scenario. Like there is an option to invade Britain right after the Fall of France but you only get a sealift of two divisions worth for one game turn only and that’s it for the rest of the game!
Also while there is a ‘sea road’ between Sicily and North Africa there is none between Norway and Denmark, which means all the units based in Norway are stuck there for the duration of the war. The Axis partners look a bit too strong I think but that might be proved otherwise once they tangle with the Russians! Conversely the Italians withdraw completely in late 1943, which IMO needs to be changed. You are free to move the Bulgarian Army around too which I don’t think reflects their actual historical performance very well. I think most of them stayed at home.
But maybe the next 20 to 30 turns will show if this game against the AI turns out to be a winner or a washout! So far I have enjoyed it anyway.

[laszlo.nemedi]

Some thoughts about the question (not the original intent of me).

I like playing against PO because much quicker turn rate than any PBEM. It makes the games feel different, more exciting. In PBEM, in some turn I cannot follow my original plan only because I simply forgot it.

The only really problem with PO: if you do a breakthrough and envelop a few PO unit and destroy them, the PO never recover. It can defend objectives only, but leaves a lot of big gaps in the frontline so you can pour a lot of units to its back. I personally like touching the enemy spearhead just to see his movements and slowing him enough to get enough time for a counter attack (sometimes no need to panic just let his forces being overextended).

That was the reason I planned a PO (let's call it PPO now) who is very good in maintaining frontline (keep in touch with the enemy). If PPO units are enveloped, he tries to go in a hedgehog position and looks for spare unit to build up a reserve for breakthrough to the main forces. Or if PPO start to envelop an enemy unit (if the envelopment succesful-the attack order should be changed too), he makes a solid frontline not just in the outer envelopment, but in the inner circle.

That (looking for reserve) can be adjusted (or can be random) to make him more unpredictible (e.g.: very aggressive PPO could make a very-very thin frontline just to save a lot of spare units to attack, or it can maintain a very-very hard frontline and no units remain to attack).

These subprograms are ready (I tested them), or can be easily implemented (originally I am a programmer-mathematician). These algorithms are known, nothing really about neuralnetworks or other egzotic theory...

It is new thought for me (originally I planned a new PO for a new game, because Norm finished patching the game in that time): it can be combined with the existing formation orders.


Laszlo

[Dog 1]

I'd love to see a better AI. I play vs it all the time.....

[MikeJ]

If I do play against the AI, I usually play side 2. AI, with +2 handicap is reasonably entertaining on the attack, though it's regard (or lack there of) for losses puts Soviet commanders to shame.

On the defence... well the AI is usually pretty bad so I don't bother... one exception is Operation Bagration. It's just fun having that much power at your fingertips :).

[aussiesta]

Originally posted by laszlo.nemedi
That was the reason I planned a PO (let's call it PPO now) who is very good in maintaining frontline (keep in touch with the enemy). If PPO units are enveloped, he tries to go in a hedgehog position and looks for spare unit to build up a reserve for breakthrough to the main forces. Or if PPO start to envelop an enemy unit (if the envelopment succesful-the attack order should be changed too), he makes a solid frontline not just in the outer envelopment, but in the inner circle.

It would be great if something can be worked up for offensive encirclements. In modern scenarios, especially those with choppers available for deploying troops behind enemy lines, or lots of enemy strongholds than can be cut off (say, any Vietnam scenario), I've noticed that one of the PO most irritating deficiencies is that it simply does not attempt to completely surround the enemy - it always lets a way out open, which very often avoids the evaporation of enemy units.