Matrix Games Forums

ORIGINAL: Andy Mac
I know. If the script-writing tools could be made easier to wield I think more scripts would be done by the community.

I was so hoping for players to write scipts so I could play v AI but no one has

AI editor is there to be used but no one does [:@]

ORIGINAL: Fallschirmjager

I want War in the Atlantic. 1939-1945

The map stretches from Tierra del Fuego to Nova Scotia to Murmansk to Capetown

The U-Boat war is covered in full. Land combat is upgraded but still a side show. The war in the Med would be incredibly interesting.

ORIGINAL: JeffK

Given that AI is a major hurdle, one question which could be asked by the "Matrix Poll" could be would players be happy with no AI for a Grand Campaign but well developed AI for scenarios?

I've played so many games head to head that it used to be the only way.

Plus developing AI for smaller campaigns might see breakthroughs in making it work in the GC.

Keep asking questions Guys, it will only see the product developed further.

ORIGINAL: Bullwinkle58

[The question is can Matrix still do this kind of thing? Does it want to? Has the Slitherine merger and announced forays into tablet-based games inalterably changed the company to an extent they can't do a major mission anymore unless GG's name is attached? Time will tell. I hope they can and will.

ORIGINAL: Commander Cody

Some excellent thoughts, Moose. I agree that AI, GUI and variable start conditions are good focus points. To expand on the starting conditions, perhaps some variability in terms of CV vs. BB vs. LBA vs. LCU pre-war building priority could be allowed. I'd also like to see Allied production as an option.

ORIGINAL: janh

ORIGINAL: Andy Mac
I know. If the script-writing tools could be made easier to wield I think more scripts would be done by the community.

I was so hoping for players to write scipts so I could play v AI but no one has

AI editor is there to be used but no one does [:@]

I played around with the scripting of the Scen1 variants for a while, initially just to remove the CV raids and bug up the group size needs to make AI go for safer runs. Your devious raids and such are a fun things -- nothing teaches you more than to be caught off-guard. But after that happens twice, you'll even against the AI start to provide better rear area security. After that, it would be best if AI could "see thru fow" before starting such a script, else...

If the editor and scripting structuring were a bit more user friendly, i.e. if one could for instance easier catch with prior or subsequent script depended on a particular one, or what other scripts used a certain base as trigger, this would make things surely more comfortable. Right now one best has to dig through one script set and try to understand all you implemented first before changing anything (or take the risk and screw them up). And that takes a lot of time.

Beyond that, improving of creating scripts that would make AI really smarter and prevent it from suicidal raids, piecemeal invasions, amphib or naval ops against player stronghold just isn't possible with the functions there are. You cannot teach it to add more LCU's to a script if say the first AI landing wave hitting Fidji's beaches is whipped by much superior player forces than usually there. You cannot teach it to postpone the date for a (counter-)invasion until the IJ/Allied CV are spotted elsewhere, far off. You cannot teach it to postpone a landing to wait for more assets, more CV, or until the 4EB flattened Guadalcanal's airfield. The assets it waits for need to be predetermined, and are not determined based on how the AAR plays out and what enemy opposition that is, has been spotted or can expected to be there based on previous moves in that region. One a particular script runs, and no trigger condition or date cancels it, it will keep feeding stuff into its ops more than often in small TFs coming one by one or in small groups, rather than doing what a player would do: retreat, regroup and come again with a concerted effort and enough to get the job done.

Without a more powerful scripting toolkit, I couldn't see how to systematically improve on your scripts. One still needs to pick dates, base triggers, LCU, LBA and naval group size requirements. One still needs to "guess-timate" the typical course of a game, maybe close to historical, and hope the player also sticks to a historical course for your guess-timates and scripts to make sense. One the player deviates too much from that course, we all now how AI winds up.
I came to the simple conclusion that one could come up with more surprises and raids (which is of course less surprising to play oneself...), or for sake of safety request AI always to create largest possible TFs -- which in the end would work best with the Ironman Scens -- but other than that Andy already has squeezed pretty much all out the scripting engine there is.

But just as a word of caution -- even if there was a scripting engine with some 1000 different commands like ARMA2, which would let you test/check or modify/adjust anything from the parameters of any LCU, LBA or ship, via any production/research facility up to the force distribution in a theater via macros, functions, or triggered functions ("event-handlers"), it would still be a long way of creating an AI that could maybe even properly handle an invasion of OZ or India, or adapt to the players habit of using his CVs (as one group versus splitting them).

Yet there would be the possibility to systematically do so, which would hopefully interest more people for modding the AI. There are so many knowledgeable and very skilled people here, it would probably make big leaps forward. Just imagine if Andy had had the ability with his scripting to include if-tests for his raids that selected the raid target from all bases in range from the present position of the raiding TF, testing (a) that the player CV are distant or weaker than the participating AI ones, (b) LBA is weaker within in a circle around the target, and importantly: (c) some juicy targets are there, be it LBA overstacking, merchants and tankers or damaged ships in habor or whatever he could have devised for that particular raid.

One other benefit I want to mention towards those would ask for a better AI on a smaller operational scale would be that those scripts/functions in the GC governing that particular area/time interval could then be extracted to use as AI core for small scenarios -- basically try to create an AI code that is as much as abstracted as possible, very object-oriented, and can be tied to together sort of like lego building. All that would surely be doable, but to create such a scripting engine might be an enormous task in itself. And as someone pointed out above, such an approach would not seem typical of G&G games.

ORIGINAL: vettim89

I 100% agree that a more robust system could be portable to other theatres and eras. I system with built in flexibility could be used to do a game around the Cuban Missle crisis or Gallipoli. Add in a full boonus features and you could do some "What If" Cold Scenarios. How about a game that mirros the "Fleet Series" for the Cold War battles at Sea. Other than Harpoon (which is a completely different scale), no one has created a PC game that truly models that era.

I have bought numerous game titles since I first purchased WiTP. WitP/AE are the only games I still play. I think Matrix needs to realize that a robust system that intergrates Land/Air/Naval combat could not only have a lot of potential for WWII games but open up an entire new area where little work would need to be done to produce an entire series of games.

ORIGINAL: Nikademus

ORIGINAL: Bullwinkle58

[The question is can Matrix still do this kind of thing? Does it want to? Has the Slitherine merger and announced forays into tablet-based games inalterably changed the company to an extent they can't do a major mission anymore unless GG's name is attached? Time will tell. I hope they can and will.

Matrix is a publishing company. What it would take is a team of programmers, developers, testers and researchers. Will it happen? Not any time soon if it ever does. It'll also not be done by Henderson Field Designs. One go-around was enough for Gary Grigsby (who has stated he'll never attempt such a mega game again), and one go-around was enough for HFD.

ORIGINAL: Bullwinkle58

ORIGINAL: Nikademus

ORIGINAL: Bullwinkle58

[The question is can Matrix still do this kind of thing? Does it want to? Has the Slitherine merger and announced forays into tablet-based games inalterably changed the company to an extent they can't do a major mission anymore unless GG's name is attached? Time will tell. I hope they can and will.

Matrix is a publishing company. What it would take is a team of programmers, developers, testers and researchers. Will it happen? Not any time soon if it ever does. It'll also not be done by Henderson Field Designs. One go-around was enough for Gary Grigsby (who has stated he'll never attempt such a mega game again), and one go-around was enough for HFD.

I'm looking for reasons to do this, not reasons it can't be done. Realizing you were on the team are you perhaps being a bit defensive about HFD? I tried very hard to give appropriate praise for what HFD did. It was a remarkable achievement and is still delivering and still being supported. But I sometimes get the feeling that some current devs think that any call for movement forward denegrates the AE achievement. If that is the case I will once again say "not from my corner." However, that said, nothing is perfect and most creative products don't ever have to settle out and die.

I realize Matrix is a publishing company, but that doesn't have to mean just one thing. Many firms in many types of creative industries operate both as fishermen, sweeping up proposals which swim by, and hunters, going out and assembling teams to execute IP they control the rights to. Book publishers do it every day, especially in non-fiction. Few non-fiction authors begin work without a contract in hand. Movie studios do it too, hiring a director to do a screenplay they already own, said director to put together the rest of the team in concert with an executive producer. This is especially true in sequels where the risk is more known and there are financials from the last production in the series to use to set budgets and marketing expenses. TV networks do it to some extent, less than in films, but they do exercise hiring and firing control over show-runners and do sometimes go shopping for a "name" to do a show where the pilot is good but the producers need help.

In the game industry there have been multiple models. Some in-house developers who were also publishers (Microprose for example), some pure publishers (EA I thnk is this), and lots and lots of usually small pure development houses, most of which flamed up and died off quickly in the now 30+ year history of the industry. Matrix could put together a team if they wanted to. If they don't I don't expect to see a clean-sheet PTO game from someone else in my lifetime. They are in a unique strategic position for as long as interest in WWII lasts. That won't be forever, but it'll be for the rest of my life at least.

Last night I found and read an interesting interview with GG, done while WITP was nearing completion. It was interesting in that it reviewed his career in games as well as explored his feelings about Now vs. Then. He confessed to really missing the 80s when he could crank out an alpha--alone--in a month and make a six-figure income from a few such titles per year. WITP at that point was driving him bats, and he swore never again. The team was 2x3 and it was, and is, three guys. Old school, but not capable of spanning all the skill sets needed for a modern, A-class game. But yeah, WITP got finished, it was great, it set a new bar. GG rested, then did the ACW game (no slouch in terms of complexity) and is now involved in a decade's-worth of effort in the current WitE, WitW, etc. product stream. Those are mega games as well. He wasn't done.

However, my point is that GG took WITP as far as he could. He was mentally flat when it shipped by his own admission. And yet a different team, larger, with different experiences and skill sets came together as HFD and took it at least twice as far. HFD did yeoman's work, and I'm not suggesting they have anything left in the tank either. But that doesn't mean WITP needs to be over. It might mean that the next relay needs different skills once again and to work on areas HFD was not tasked to undertake.

If interested, the interview with GG is at

http://www.wargamer.com/article/1383/in ... ry-grigsby

ORIGINAL: Canoerebel

Interest in the American Civil War has lasted 150 years. Interest in the Spanish American War and World War I is nearly non existent in the USA. Interest in World War II is vast and will, IMO, rival the American Civil War. In other words, in 2090 you'll see WWII afficianaods on a scale comparable to what we see with the Civil War today.

There is no doubt somebody will one day do another massive Pacific Theater game. It will be somebody or some group with unique skill sets that allows them to take on the project and do it well. Obviously, Matrix Games and it's allies has a corner on this right now, but whether they have the will to do it in the near future seems doubtful. If they don't, somebody's going to come along and meet the need. Nature abhors a vacuum.

ORIGINAL: fcharton

Hi Bullwinkle,

A few more points I thought about today at work.

The economic part of the problem is relatively simple. There are two sides to it, revenues, and costs.

A limited rework, and user contributed specs and tests would reduce the costs, but one must keep in mind that improving on an already good game (and working on old code) cannot be a small endeavor. If I understand correctly, AE was made possible because a dedicated group of users absorbed most of the cost, by working for free, or very low cost. I don’t know whether this route is still possible, but it needs to be considered at some point.

I have to differ here on ethical grounds. I don't know Henderson's deal, and if they were fine with it I'm fine it. But I have a problem with a for-profit company assuming they'll use unpaid volunteer labor ongoing and pocket all the profit. On the one hand everyone is free to enter any legal contract they wish, but on the other it makes me feel a little dirty to send payment off to Matrix knowing they got the goods for nothing or close to it. I did it, but I'd rather not do it again. And I don't think Matrix should try to do it again. They have at least twice as much data now on the legs WITP as a franchise has and could have. If it makes financial sense to do another cut find a way to do it clean. I personally think the only give there is falls on the revenue side; development costs what it costs. The feature set should be controlled to make that side of the equation reasonable, but this game of all their games has a unique demand elasticity. If they want to explore the upper reaches of retail pricing as a condition to green-light this I'm up for that.

On the revenue side, we all agree that the regulars would probably fund a new version, and anyway, a new WITP will prolong the product shelf life, ie more revenues. Another possibility would be if some of the new concepts (AI/UI) could be adapted to other games, or serve as the base for a new series.

I'm not so sure as I said elsewhere, but it's possible. Depends a lot on their contracts with the various devs in the stable and the state of the IP.

Now for the technical aspects… I imagine WITP as a set of four parts.

At the core is the “model”, those are the game file, the OOB and various datafiles, plus the map, and whichever parameter the game uses.

Then you have the “turn engine”, a machine that runs, and “plays” a turn, ie changes the model for today into the model for tomorrow, producing reports and a combat replay on the way.

In between, you have the user interface, which serves as an editor for the model. It allows the player to visualize part of the model (limited by FOW) and to update it (before sending it to the opponent and feeding it into the engine to get the model for tomorrow). Note that Tracker acts like a “semi-editor”, in that it allows one to visualize a turn, but you have to turn to the original UI to input your orders;

What about the AI? Well, the AI is just an automated version of the editor. It takes a turn, visualizes it and inputs changes, just like a player does (but without the wide screen, the cursing, the beer drinking, and the unhappy spouse).

In my opinion, WITP2 is about redoing the two last parts, while keeping the two first unchanged. Note that this could be done on top of the current game, but having some sort of Tracker-like tool which could not only visualize, but also edit turns. You would just use the old AE for turn processing and combat replay. (I am not advocating that. I am giving this example as a proof of concept that this four part model works)

First, a great way to visualize the thing. Excellent breakdown. To the extent I have only a user's view I think you are correct in how you group the pieces and functions.

As an aside, after reading the interview with GG I linked to in another post, and having played a number of his 1980s games on my AppleIIe, I see the remnants of that era in WITP. He came out of a strong procedural programming language background. Game flow was on a master loop architecture. I'm pretty sure some of his 8-bit games were written in BASIC, and although I looked mightily and was unable to find confirmation, I could swear I read somewhere in the distant past that "Pacific War" was written in a late version of PASCAL. That might be untrue. But my memories of how it phased seem to me not to indicate a strong object oriented programming mindset. I believe UV and WITP/AE are in some form of C (+, ++, # etc. I have no idea.) But GG loved his grand game loops in olden times, and WITP shows it. (WitE still goes there but less visibly.)

In that architectural mindset, and given that WITP was his baby and not a committee's, I think you're correct with your breakdown and the reasons for the dividing lines. It's why I think (or hope) that the opponent AI is largely to one side of the core EXE loop, with defined entry and exit points to the UI and algorithm and DB-accessing core of the engine. If, as you say, the AI and the UI are mostly divorced, or at least not deeply embedded, in the core of the exec. then modifications might be possible on the periphery where the player experiences the game at less than break-the-bank cost.

Two points from my POV, however:

1) GG made an early, core decision to make the map the focus of the UI. For everything which flies, drives, dives, or marches this is great. It allows the normal human visualization routines developed by our hunter-gatherer ancestors to kick in. "It's about THAT far from Haiphong to Saigon, so it shoud take about THIS long to get there." Much easier to play by eye than have to decide from a distance spreadshet. However, GG in making the map the interface made production, economic, and logistic tasks like undergoing root canal. Particularly for the Japanese the stock, as-shipped routines were really hard to use. Everything was drill-down from a hex. The addition of the industry tab and some sorting capability helped a lot, but Tracker is needed by many players to get a handle on things. Making a drill-down map the core management interface was not the work of a pro UI designer. As you say, it could be made even more seamless by incorporating some Tracker-type functionality into the core game. But, as in this whole discusison, that it might get too deeply into core game code.

2) When I say "AI" I mean the computer-opponent capabilities only. In the forum the term "AI" is sometimes used for this and sometimes used to mean all processes perfomed by the game which are not under player control, such as naval targetting. Some routine functions like pilot management we have discussed as being automatable, but that to me is not "AI". As well, those functions also are embedded in what you call "the model" rather than the semi-external "AI". Seeking to make them run without interface input might again require digging too deeply into core code. I'd rather advocate for improvements in opponent AI and scripting, what I have called "bolt on" aspects of the game. (Possibly naively.)

It is probably good (although not quite correct) to picture the UI as two different subsystems. You have a visualization tool, and an “input tool” to give your orders.

Right. And to some extent GG made them the same thing in ways a pure-Windows developer might not have.

At the heart of the visualization tool would be a better, zoomable, easier to search, map, and a series of user customizable reports, with data intensive reports like tracker, and more graphic visualization. I believe modern UI frameworks can handle this. One point I would really like to have is a report generator, something where you build your own dedicated screens. Something I don’t like with the current UI and Tracker is that you can’t choose what gets displayed and what doesn’t. Also, I’d love the UI to allow for “historical reports”. This would be practical for AAR, but also to keep tracks of past information.

I agree. This is in the "bolt-on" camp. If the Tracker guys, Damien and Floyd, can take a save turn file and peel it like an onion I'd think a Matrix team with full EXE access and specs could go one farther. Data management is the first step to making decisions, and GG's tables, while useful and familiar now, are not great for figuring out a strategy. I think incorporating a report generator could be done fairly cheaply and provide a lot of oomph marketing-wise. It's hard to screenshot the world's greatest AI, but you can show report screens and grogs begin to drool.[:)]

The input tool would be much linked with the “auto input tool”, aka the AI. Players would be allowed (under certain restrictions) to script and delegate to the AI some repetitive chores. Note that the AI already handles such tasks, like moving units, loading and unloading, scheduling bombardments and landings. I like to think of that scripted input as a batman. The actual AI would be made of two parts, a full fledged batman, and a scripted master plan.

This is where I get leeery. Not only on an ergonomic basis if you have some tactical decisions made here and others there, but on, again, an EXE digging basis where you might have to really latch onto things which barely work now. The kind of things you refer to would be great features if designed into the engine from the deck plates, but I'm afraid trying to graft into the current code blows up a budget pretty quickly.

One new thing the batman would make possible is multi-turn orders. This already exists, in the form of patrol orders for subs, auto-convoys, ship movement with waypoints, refuel and follow orders. You mentioned something similar for pilot training, but you could also delegate some tasks (besiege Yenan, sweep Manila, bomb Tokyo), and perhaps even have the quality depend on the ratings of your local commander.

I believe this could improve the game in several ways. First, multi turn orders would allow for better handling of three day, or even longer turns, therefore shortening the campaign. We all love our one day turn, I’m sure, but some of us would love to see what 1944 is like before Alzheimer sets in and we all forget how to spell Babeldoab (or is it daob?). Also, this might help reduce the control players have on their troops. You could very well decide that some orders can only be input every other day, or take a while to implement (à la Gamers/MMP). And then, initiative (the ability to countermand previous orders) could be decided by game parameters, leader ability and so on.

Again, here, I'm leery. I've played all three cycles, and find the 3-days are past the point I'm comfortable. A week-turn cycle in WitE works due to the scale POV of control (larger formations mostly) as well as speed of advance of the units against a contested, old-school front-line of battle. Going past 3-days in WITP and ships and planes can be in and out before the opponent ever sees them. The PTO is not a battle line enviro. At certain times in the war it's a 360 degree problem for each side. I do agree that longer cycles addresses issues like attracting different types of players, and it does somewhat disguise AI weaknesses by letting the AI always play 1-day turns even when the human is tied gagged to a stake for five days or a week, but I think you'd have to be very careful not to break a system designed around very low-levels of unit abstraction.

Finally, such multi turn orders could be a game balancing factor, and boon for the AI. Just imagine an AI like the current one (ie slightly dumb) but reacting on a much faster cycle than the player. For instance, the player would be on three day turns, but the AI would react on a daily, or lower, basis. To some extent, such a use of multi turn orders would serve as a “poor man improved AI” (and would be a perfect illustration of OODA and other cycle based warfare models).

And this is how, dear reader, I spoilt what looked this morning like a perfect day of work…

Well, it still was perfect, right? Maybe not profitable . . . [:)]
Hope you didn't get fired.

ORIGINAL: jwilkerson

I've spent some of my "spare time" in the past two years working on a new AI model. I want something that plays like I do .. that thinks like I do ... that can react to circumstances ... that looks at data across time (trends) and not just "what are we doing this turn". Each level of the AI essentially solves the same problem ... how to I accomplish my mission with the forces I have available, given what I know about the enemy and given what I know about terrain, weather, supply, etc.
I've learned a lot ... I've learned that experienced game players have a lot going on "under their hoods" ... in terms of matching up resources across a multiple dimensional data set. I think we do this mostly without realizing we do it! As I've tried to break down what I do and why I do it - I've realized how much I am doing that I do not even know I am doing!
So what would it take to finish this and make a working AI engine? I'd say at least four people - game players who are willing to analyze and write down how they play - and do this a lot for the first year - from all this we synthesize the AI "algorithms" ... the second year is spent testing. This is for the land AI ... for the Naval, Air and strategic aspects you need other teams.
I haven't seen the resource availability within our team to make this happen - so thus far - it has not.

I wouldn't rule out a WITP2 one day. But it might not have the Joe vision of a new AI [:)]

ORIGINAL: treespider
Some of what is proposed is great to dream about...but the current engine simply cannot handle it, a rewrite from the ground up is required and I don't think you're going to find anyone in the near future willing to take on the task.

This is where I get leeery. Not only on an ergonomic basis if you have some tactical decisions made here and others there, but on, again, an EXE digging basis where you might have to really latch onto things which barely work now. The kind of things you refer to would be great features if designed into the engine from the deck plates, but I'm afraid trying to graft into the current code blows up a budget pretty quickly.

Again, here, I'm leery. I've played all three cycles, and find the 3-days are past the point I'm comfortable. A week-turn cycle in WitE works due to the scale POV of control (larger formations mostly) as well as speed of advance of the units against a contested, old-school front-line of battle. Going past 3-days in WITP and ships and planes can be in and out before the opponent ever sees them. The PTO is not a battle line enviro. At certain times in the war it's a 360 degree problem for each side. I do agree that longer cycles addresses issues like attracting different types of players, and it does somewhat disguise AI weaknesses by letting the AI always play 1-day turns even when the human is tied gagged to a stake for five days or a week, but I think you'd have to be very careful not to break a system designed around very low-levels of unit abstraction.