Mini AI for PBEM

[juntoalmar]

I’m not sure if I already posted this in some thread in the past (my apologies in advance, if I’m repeating myself).

I was thinking of the possibilities of playing this game more than solitaire game. Given the fact that:

a) the AI will be quite complicated to implement, some people even think it will never be accomplished or will never be “competent”, given the complexity of the game. So it will take a while… if it can be done at all.

b) because of the rules design, any form of multiplayer game involves a lot of “Non phasing player” interaction (interceptions, assault/blitzkrieg table, units lost chosen…). This makes that multiplayer, either NetPlay or workarounds (like NetPlay) are basically a “hot seat” game, where both players are at their homes connected by the Internet. Which I don’t like. I like to take a lot of time thinking my moves, and having the non-phasing player waiting for just a little interaction seems ineffective (I know, I know, just like a boardgame). I don’t want to wait for people and I would like to read the manual before trying that invasion or decisive naval move (without others rolling their eyes in despair).

I was thinking… what about a first step of “mini AI” that will take decisions for the non phasing player? This is the way I played PBEM of Avalon Hill Third Reich on the PC. On the Allied turn, he switched on Axis AI, and switched off before sending the turn to the opponent. Axis player will switch Allied AI on before playing his turn and turn of on finish.

This should be far easier than a full AI and would allow a PBEM game (thus multiplayer without the need of NetPlay to be finished). I know the caveats, that the AI will not be perfect and will take some funny decisions some times. But on some scenarios like Barbarossa, I don’t think it would be that serious or threatening for the game result, and would allow us to play multiplayer games easily, and without all the technical issues related to simultaneous playing through NetPlay.

Even more important (at least, to me), the possibility to play a multiplayer game off-line (not being all the players synchronously connected to play). Receive the turn, play when you can (a maybe on a week’s time) and send it to the other(s) players. Global War with many players won’t be harder to implement (I’m not sure if Netplay is intended to work with more than 2 players).

What do you think?

[brian brian]

I like it a lot. It has been suggested before, a long time ago now of course. I also like thinking of it as a "first step" - leading to improvements/progress for both people hoping to play other people at remote distances AND people who want to just battle against an AI.

A few other things that could improve the play-flow of the game would be tweaks to the play system itself, to cut out a few of the interaction steps where you have to wait for the opponent to decide something. (These could possibly improve the 'skill' of the AI as well). My suggestions would be:

1 No "Do you intercept?" request when a naval force enters a sea zone and then stops there. Very occasionally this is important, when the non-phasing player is low on units to dis-organize to initiate a search. A large portion of the time in ftf play, no one bothers asking this question - there are more than enough questions to ask already; and also because a successful search doesn't even do anything - you still have to find the enemy again in normal naval combat anyway, if the intercepted force just stops moving. And, many of such moves are totally obviously going to stop in the zone in question. Perhaps this could be turned on and off for the purists. "Solitaire" mode in use currently already has an advantage here - the moving player can just decline this for the non-phasing player most of the time.

(All of these should be looked at in terms of how people are handling things as the game is currently played by people emailing the game file back and forth, I bet there are a lot of gentlemen agreements on how to save on the interaction steps)

2 A single "CAP" phase for the impulse, and yes, then a FTR counter could end up being available on CAP for differing opposing air missions, such as both Port Strike and Ground Support. These are 2 month turns and counters representing hundreds of aircraft after all. And really, would a FTR wing just beyond it's 1/2 intercept range really be only able to cover one hex? Aren't air missions supposed to represent more than a single bombing run (in 2 months!?). This just shouldn't in any way be handled like a tactical air combat game, but parts of air combat system is. I mean, aircraft are given enormous flexibility in rebasing here there and everywhere (as well as flying out to distant parts of sea zones) in the interest of simplicity, but denied this flexibility when trying to defend against enemy bombers.

3 Some FTRs on automatic intercept status. So the Ju-88 night-FTR3 sitting just behind the hexes of the Ruhr flies up automatically when the Stirlings come over at night. Everyone knows why it is sitting in that hex, just designate it. This would be gamed by opponents with four bombers sitting in England - send over a sacrificial old model first, let it be intercepted, then completely smash Dusseldorf with the other 3 on the next impulse while Colonel Hogan somehow talks Colonel Klink into seeing nothing, again. So the designation would need a few filters and auto-intercept would have to be secret. This idea starts to run into the structure of the design - a design meant to keep a game moving along with only the visual information of the counters after all, rather than the power of computer software to manage information for the players.


Those are just a few I thought of quickly and have less to do with an AI than just streamlining control of the counters. Most of these aren't that consequential in face-to-face play. But I suspect remote play will turn out to be a fair bit slower than face-to-face play as remote players will have many more distractions around them than ftf players. And mismatched play speeds is already an inherent problem of any type of mental (not video) gaming. Anything that can be done to reduce that problem of differing play styles is worth considering, imo.

Of course, before we have a 21st Century game, we have to perfect a 20th Century game, and that's just the way it is.

[pzgndr]

Primarily this is accomplished by the creation of Standing Orders (SOs) by both players. An SO is more or less a previously written set of instructions to MWiF from the non-phasing player, for what to do when the phasing player does XYZ.

This is what 4.1.5 says in the player's manual for when PBEM is implemented in a future expansion. A default set of (editable) SOs could probably be developed to do this and implemented as an interim mini-AI at some point? I think there are other aspects of the AI that could be phased in as they are developed for solitaire games, where players could allow the AI to complete particular phases. This might be challenging to implement, but could provide Steve with helpful player feedback along the long iterative road of AI development so we're not waiting forever for a fully functioning AI.

a) the AI will be quite complicated to implement, some people even think it will never be accomplished or will never be “competent”, given the complexity of the game. So it will take a while… if it can be done at all.

I remain optimistic that it can be done, but I agree that it will take a while. Probably easier for the half-map scenarios first, and then for the global war scenarios which would be more challenging to script.

[Hairog]

I like it too. When the AI screws up you could think of it as a rogue general or admiral that didn't follow orders. Happens all the time in military operations. You would have to work around it and adapt

[brian brian]

ORIGINAL: Hairog

I like it too. When the AI screws up you could think of it as a rogue general or admiral that didn't follow orders. Happens all the time in military operations. You would have to work around it and adapt

Exactly. Just like the action limits, and even the reason for rolling combat dice. One's subordinates aren't as perfect ... as you are.

[Centuur]

ORIGINAL: brian brian

1 No "Do you intercept?" request when a naval force enters a sea zone and then stops there. Very occasionally this is important, when the non-phasing player is low on units to dis-organize to initiate a search. A large portion of the time in ftf play, no one bothers asking this question - there are more than enough questions to ask already; and also because a successful search doesn't even do anything - you still have to find the enemy again in normal naval combat anyway, if the intercepted force just stops moving. And, many of such moves are totally obviously going to stop in the zone in question. Perhaps this could be turned on and off for the purists. "Solitaire" mode in use currently already has an advantage here - the moving player can just decline this for the non-phasing player most of the time.

Strange, this one. I've come to the conclusion that it is always important to have the other side decide to intercept or not, especially when that player is the Japanese or Italian. In ftf games, I have always kept my mouth shut and simply moved my naval units in a sea area and asked if there would be an interception or not (even if I intended to stay in that sea area). My goal was to make sure a ship or plane gets disorganised and 1/10 (or 1/5) oil had to be used (or at least make my opponent ask himself the question if he will reorganise the ship at sea or not during the use oil phase).
I always think that every small chance an oil point gets wasted this way on the Axis side should be exploited...

In general, I agree. Standard Orders are needed for PBEM...

[Jagdtiger14]

Centuur is very correct. This happens a lot in face to face games all the time...trying to get your opponent to flip something naval, but it also happens during CAP on land as well...not so much for oil reasons, but more importantly trying to get your opponent to commit something valuable early, or get any aircraft committed away from an area you actually intend to attack into, etc...

[brian brian]

sorry to borrow the thread for the below. perhaps the first place a bit of AI code could be extremely useful would be to let it handle air combats for a player. the AI will have to eventually do that anyway. maybe it could use a sliding scale of outcome preferences that a player could set in advance as I would imagine the AI would have to make a similar decision internally. when it is the most critical hex on the front, slide the scale to order your bomber pilots to push on no matter what. when you are just making a diversion, slide the scale back to force protection before valor.

re: "Do you fly CAP for Port Strike?" - ask yourself how many times a game this actually happens. do you want to answer that question at the very beginning of every single opponent impulse, for each active Major Power? sometimes there is an obvious possibility of needing this - then ask the question. the same is true for most opportunities for CAP. good opponents know when to ask this question, and when not to. perhaps you can sucker an opponent into wasting a CAP mission. how many times will that happen after the first?

re: offering an intercept chance for a naval force that is going to stop - again, good players know when to do this and when to keep the game moving. in the Med if there are 5 corps / side battling it out in Algeria and nothing happening in Egypt, you know the big naval force entering the West Med isn't on the way to the East Med. conversely, a major Kriegsmarine task force is unlikely to be stopping to fight in the North Sea in the middle of winter with no CW units in France and no CP in the zone to start with. and yes a patrolling RN squadron probably is going to want to intercept. I'm sure most reading this can think of many such examples.

if you want to play this game by always moving one ship at a time and asking for an intercept chance in every zone I have units just to stop the ship on a successful search roll, which achieves nothing except making the ship stop where it is going to stop anyway, I am probably going to find a different opponent to play. there are times when an inferior naval force needs to use that tactic, sure. but not every time.

if the game never reduces the scores of interaction points, players will move on to more streamlined game systems that are quicker to play. human players are able to filter those requests as they play and a bit of coding to facilitate that while playing electronically would be a great thing. I truly believe remote play could well be slower than face-to-face, and every bit as difficult to schedule.

[Jagdtiger14]

I remember a game where I as the phasing player did not ask my opponent specifically if he wanted to fly CAP (up until then we were not in the habit since it was very rare)...although he was watching everything I was doing. I went on to do some critical ground strikes (flipped Zhukov). The next game day he was all upset at me for not having asked to fly CAP. From that point on we make it automatic that the phasing player calls out everything in the sequence...all that really happens every impulse is a quick call: "CAP?"...which does not slow the game at all.

[brian brian]

True. But if your opponent 500 miles away has gone into the other room, or fallen into a YouTube video on another browser window, or taken a phone call (a thing of the past perhaps)...4x per impulse, for each non-phasing Major Power...that means before you can touch an air counter and designate your bombing missions, you have to wait for a response. not too big a deal as immediately after you announce your air missions, your opponent can intercept, unless they have gone into the other room, or fallen into a YouTube video on another browser window, or taken a phone call (a thing of the past perhaps), and then you can do ground combat (with another CAP phase), but afterwards is the Air Resupply phase...I don't think all this never-ending call-and-response is what people are looking for in PBEM style play.

a single CAP phase would eliminate 4 of those interactive steps, and avoid this: "Where are our damn fighters? These Stukas are killing us!" - "They appear to be circling over the harbor, sir." - "But that's only 5 miles away!"

and all of which is why juntoalmar's idea is so good.

[juntoalmar]

Yes, I calculate that making such mini AI would be much faster than NetPlay and full AI (or half scenario AI), thus having players playing Multiplayer games easily and sooner, and enjoying (more) the game while the development continues.

If such implementation of PBEM game was accomplish, I wouldn't mind waiting 2 years for AI.

[Jagdtiger14]

A single CAP phase would be a very good idea!