Matrix Games Forums

As new research is attained it appears to affect the 14 attributes of units.

1. Is the upgrade retroactive to existing units already built?

2. If not retroactive, is there a way to somehow refit out-of-date/obsolete units? or maybe disband them for some kind of economic or manpower refund?

Another qestion while im here,

3. Do units somehow gain experience?

ORIGINAL: MButtazoni

As new research is attained it appears to affect the 14 attributes of units.

1. Is the upgrade retroactive to existing units already built?

2. If not retroactive, is there a way to somehow refit out-of-date/obsolete units? or maybe disband them for some kind of economic or manpower refund?

Another qestion while im here,

3. Do units somehow gain experience?

1. Yes, the refit cost is built into the research cost.
3. No, there is no experience gain for units.

ORIGINAL: Joel Billings
1. Yes, the refit cost is built into the research cost.

so does the quantity of built units affect the research costs?

ie. the more "tank units" i have built the more "tank research" will cost to attain the next "breakthrough"?

I also hope the research benefots aren't as dramatic as they were in Strategic Command.

ORIGINAL: neuromancer

I also hope the research benefots aren't as dramatic as they were in Strategic Command.

I'm not sure about that.

Here's just some of the things that were historically "researched" during WWII:
- Atom Bomb
- Jets
- Radar

Here's some of the technologies that increased 10 fold (for some countries):
- Torpedoes
- Logistics Capabilities
- Production Capabilities
- Strategic Bombing Range

Yes, but in SC the improvements were far greater than even 10 times. A level 5 plane would swat anything it came against, except another level 5 plane, even a level 4 was at a significant disadvantage. It was like the individual levels of tech were exponental, Level 1 gave a very small bonus while Level 5 gave a crushing bonus.

I don't disagree with technology development, it just shouldn't be - as I said - quite as dramatic as in SC.

ORIGINAL: neuromancer

I also hope the research benefots aren't as dramatic as they were in Strategic Command.

It's not so much breakthroughs but a steady progression of improvements. You only gain one on your rating that is being researched. In some cases (like ASW), 1 can make a big difference, but with others, 1 is not decisive. However, you can keep increasing the value.

On a related note, I for one hope that the focus of the game is not technology research and development, but rather military strategy. I would prefer to spend my time analyzing whether the U.S./Britain should attack Italy, or the Balkans, or Scandanavia rather than spend countless hours on quieter subs versus heavy bombers versus the a-bomb.

My first impression of WAW is that it provides a good mix of both strategy and reseach, but if most of my time is spent on the latter rather than the former, I'll quickly lose interest (see HOI).

HoI got research wrong though

It was a Strategic Game but the research was way to detailed for the scale of game.

HoI was a classic case of "Design for Process" instead of "Design for Effect".

ORIGINAL: Joel Billings
In some cases (like ASW), 1 can make a big difference, but with others, 1 is not decisive. However, you can keep increasing the value.

Need to make sure though Joel that one side doesn't suddenly completely dominate an aspect of the game that would alter play balance too much. I am sure I am preaching to the converted here.

Will the AI keep track of a Human Players Research Tendencies in order to select the best research strategy. For example - if the Human player has focused on researching XXX for the last 5 games will the AI be more likely to counter it with ZZZZZ or by adjusting his war strategy.

Example - Human player for the last 4 games has built large sub fleets, will the AI use this knowledge to select a counter strategy. I am assuming that the game saves a limited record of the human players strategy for previous games played.

ORIGINAL: EdwinP

Will the AI keep track of a Human Players Research Tendencies in order to select the best research strategy. For example - if the Human player has focused on researching XXX for the last 5 games will the AI be more likely to counter it with ZZZZZ or by adjusting his war strategy.

Example - Human player for the last 4 games has built large sub fleets, will the AI use this knowledge to select a counter strategy. I am assuming that the game saves a limited record of the human players strategy for previous games played.

No, our AI does not learn the tendencies of it's opponents. In over 150 games, I recall we only had one at SSI that did that (Fortress).

Call me cynical, but I've almost given up on AI entirely. I'm tired of being disappointed by games that promise 'the most challenging AI ever', but don't deliver. The latest example being Korsun Pocket, which is an superb, fun, outstanding game--probably the most accurate and playable of its type. But in my third game against the AI, with the difficulty set at 'computer++', I trounced it with ease; it wasn't even close.

I look forward to WAW, but am not holding my breath on the AI. Nothing beats a real person.

ORIGINAL: Rummy

Call me cynical, but I've almost given up on AI entirely. I'm tired of being disappointed by games that promise 'the most challenging AI ever', but don't deliver. The latest example being Korsun Pocket, which is an superb, fun, outstanding game--probably the most accurate and playable of its type. But in my third game against the AI, with the difficulty set at 'computer++', I trounced it with ease; it wasn't even close.

I look forward to WAW, but am not holding my breath on the AI. Nothing beats a real person.

I think a clear version of reality may be striking for you. There is no version of AI that can give humans a real match. Look what IBM has spent on Big Blue just for a decent chess game.

Hey Rummy:

Want to play a pbem game of Korsun? I hope I'm better than the computer [8D]

Yohan

ORIGINAL: Yohan
I think a clear version of reality may be striking for you. There is no version of AI that can give humans a real match. Look what IBM has spent on Big Blue just for a decent chess game.

And Chess is relatively simple... 64 positions on the board, 32 pieces with very well defined and limited movement.

The entire method of computer chess is to calculate EVERY possible move. The computer will have calculated every possible move and counter move. The key - and the necessity of a super computer - is to do it in a realistic time frame.

I remember the human opponent being really shaken by his defeat. I wasn't even remotely surprised. With Big Blue, it was completely impossible to 'out think' it, because it literally knew every possible move you could make before you did.

I suppose it was possible for you to get into a chain of moves that makes it impossible for the computer to counter, but it would be extremely unlikely.

But this was a multi-million dollar computer designed for the sole purpose of being able to defeat a world master chess player. But it wasn't 'smart', the entire process was brute force. Calculate all the possible moves, and then select the path with the highest probability of success (i.e. fewest possible results that end in defeat). Repeat after every move until the game ends.
And at no time will you be able to point at the computer's game and say "Here is it where it lost, if it had gone here instead of there, it wouldn't have lost", because it would have already known that, and unless A and B both had an equal chance of defeat, it wouldn't have picked the bad direction.

Meanwhile Big Blue probably couldn't run solitare for you.

AI is very hard to do. And if they can make a really good AI, or even - God forbid - MI (machine intelligence, real intelligence possessed by a machine), then they would be working for a bigger company than Matrix!

Check back with me in a couple months. Work keeps me pretty busy till the summer.

Good point, Neuromancer (what the heck does that mean, by the way?). I'm certainly not blaming game designers for shorting us on AI. I am just tired of having my expectations raised. My real problem is a lack of time to play with actual people (having two toddlers does restrict one's time).

Why is it though, that some of my favorite old games like Harpoon and the V For Victory series seemed more challenging? Even TOAW was tougher than Korsun's AI.

Or am I just getting better?

ORIGINAL: Rummy
Good point, Neuromancer (what the heck does that mean, by the way?).

Interestingly enough it was an AI (an MI actually) in a book of the same name. I've been using that as my handle since 1989. Altough I've late 'vce started shortenng it to just Neuron.

And a neuromancer would literally be a magician of the mind.

Why is it though, that some of my favorite old games like Harpoon and the V For Victory series seemed more challenging? Even TOAW was tougher than Korsun's AI.

Or am I just getting better?

I imagine that improved skill is definitely a part of it. I do remember TOAW having a pretty tough AI, but I think it tended to 'cheat'. That is the usual way to make an AI better, is to let it cheat - it 'knows' things that the human player could not, or does things that the human player is not allowed to.

Sometimes this is an acceptable way to get around shortcomings in the AI.

For example; the CORE mod for HoI gets around one piece of the ludicously bad HoI AI by 'cheating'. Basically the AI cannot figure out the concept of escorts for its bombers (obviously a key piece of code left out of the base AI scripts in the game). The CORE mod has the AI build a special bomber squadron type that is assumed to have the escorts built in. Not a perfect solution, but a decent resolution with the available tools.


I find that the Uncommon Valor AI is pretty good, at least in certain areas. It plays a good sea game, and a good air game. Where it sucks is on land. It doesn't invade like it should, in fact it does it very rarely. This is fine for a defensive game where the computer is trying to slow the human advance, but sucks the other way. the problem with UV is that the game starts with the Japanese having the advantage and they should be the ones invading, and then switches to an allied advantage where they should be invading.

Of course, I find that playing against a human is much more interesting. So much so that I am tempted to abandon my computer opponent games and just play PBEM games.
In my games against the computer I just let my troops pile up until I am ready to invade with them because I know the computer won't be invadinbg me. Against a human, that is a suicidal assumption!

Of course I make so many dumb mistakes (just check my current AAR for examples) that I am not sure if I'm a 'better' opponent than the AI. But I'm pretty sure I'm more interesting!

"Neuromancer" is the name of William Gibson's 1984 sci-fi novel that was embedded early on into the cyberpunk "culture." It was heavy on the cyberscience (you could get a computer motherboard interface stuck in your head and become sort of a walking CPU with plug-ins). I thought it rather silly.