Massive Assault

MAN 2 Discussion
Page 3 of 5

Author:  Quitch [ Mon Jan 09, 2006 11:41 am ]
Post subject: 

I don't have the MAN2 beta, so I'm speaking from previous experience. On my P4 3Ghz I never had to wait more than ten seconds for the AI to think.

If the AI is thinking for five seconds it is thinking X ply. Allow it to think for ten seconds and that's probably another ply. Five minutes and... well.

I was in touch with the AI programmer during the course of Domination and the biggest limiting factor is the depth of search the AI can do and the number of searches the AI can do. It's why, even in Domination, the AI doesn't utilise transports well, because they exponentially increase the number of calculations. Such a problem can be greatly helped by simply throwing more processor cycles at it. Sure, you can improve the AIs algorithms, but that takes time and money, processor cycles can be allocated according to how long the user is prepared to wait, and the longer you wait the better the AIs move (up to a point, there is obviously an optimal setting for any algorithm). IMO it's fast currently simply because the developers don't think people want to wait any longer than that for the AI to move. There's no pre-ordained reason the AI has to move within that timeframe.

Author:  Pitor [ Sun Jan 15, 2006 9:55 am ]
Post subject: 

Such an option is a good idea but I would be surprised that u can get a big improvement of the performance of the AI except if u accept to wait
some hours or if u possess some supercalculator.
The main reason is due to the exponential increase of time needed when u want to deepen the deapth of search in the moves tree (just think to chess : there are much less possible moves in chess than there are in MAN, but the complete chess problem still remains computationnally intractable). Of course, like for chess, it is possible to use a lots of 'tricks' and heuristics to lower the number of computations needed and the dev clearly used a lot as the AI performs quite impressively well for the small time it need to compute a solution :wink:
But even with such heuristics, im afraid the the AI will need either much more computing resources or more intensive developpement of 'smart algorithms' (and so costs for the dev) if u want to see it really improve in acceptable time.

Author:  Quitch [ Sun Jan 15, 2006 10:15 am ]
Post subject: 

Well, right now it moves in ten seconds (or did in earlier games) on my machine. How many ply was that? Maybe two (its move and my move)? So four ply would be about, say, a minute or two? Not more than five I suspect. With a short wait time you just doubled the ability of the AI to play. Sure, the cost increases fast, but you can make some improvements if the user is simply prepared to wait.

Add in the ability to minimise to a system tray icon while the AI thinks, and you're onto a winner :) No wasting processor cyles on tree animations and scrolling, etc.

The way the series is going each game is built on top of the last anyway, so I suspect the algorithms for the AI will get better, and not as an additional cost since it will be a new game. All the while you're teaching users to give the AI time to play, and thus increasing its smarts cost free.

Plus, processors WILL get faster. No point thinking so short term that you have a fixed time setting as two years from now that time will be a joke.

I wouldn't have thought turning the "think time" setting into a variable would take much work. No harm in adding it as an "is this worth it?" feature.

By todays standards, my P4 3Ghz processor is stone age anyway. The lowest end Athlon available on most hardware sites would equal it at the very least. No OEM ships processors slower than this any more.

Author:  Maelstrom [ Sun Jan 15, 2006 11:00 am ]
Post subject: 

I don't think just throwing more processing power at the AI is going to be enough to really change things. Certainly, it may help a little, but you're talking about a game incredibly deep in strategy with billions of possible ways the game may play out for any given setup of allies (of which there are probably thousands possible for each map).

This game compares very poorly to chess in this field. Chess has one board setup, 64 possible movement locations, moves only one piece every turn, exactly 16 pieces that are set up exactly the same way when you start, and a rulebook only a couple pages long.

Also consider how much time has been spent by scientists, AI engineers, etc, discussing the problems found in chess (I wonder how many masters thesis and doctral studies have been made about chess). What it comes down to is that chess is completely tractible. With a powerful enough computer it can select the best move to improve its position by looking at every possible way the game can play out. A few simple optimizatins and some known strategies programmed in and you have a really intelligent AI chess player that can think of a perfect move fairly quickly.

The same is vastly different in MA. The number of possible moves and countermoves are staggering. I'm with Pitor in thinking how amazing it is that the AI plays as well as it does in how short of time it plans out it moves.

I am willing to bet that if they let the AI have as much time as it wanted to process, it wouldn't play much better than it already does. The heuristics in MA are just too stinking complex to expect much better.

Author:  Quitch [ Sun Jan 15, 2006 11:10 am ]
Post subject: 

Which is why I'd like to see the beta testers get the feature, so that we know once and for all. Hell, 100 games of a 5 second thinking AI vs. a five minute thinking AI complete with results would be a good way of finding out whether the AI benefits or not (100 on a small map, 100 on a big map).

The algorithms aren't going to get better without developer time, and it seems obvious to me that no magical development is going to happen any time soon. Processor time is as free as someone wants it to be, so let's see what happens when a realistic amount of it (five minutes worth) is thrown at the problem.

The AIs sole biggest weakness is transports, and the problem with transports is the AI can't calculate deep enough to use them (effectively). You can help with smarter branching, but in the end the only thing that's going to solve that is processor time. Until the day the AI is driving everything around in transports the AI simply won't compete. Tactically it's already pretty good at using its firepower, but you can dance rings around it with transports no matter the odds.

Author:  RadRx [ Wed Jan 18, 2006 12:16 pm ]
Post subject:  Man 2 comments

First, is there any way I can join the beta-test for MAN 2? I've signed up, but haven't heard anything back. I love the game and would love to participate in the test.

A few have argued against a chess clock like format or timer. Maybe some don't like it, but it would be easy to implement an option for unlimited time for those. A chess-clock timer allows for games to be played at a speed mutually agreed upon by the players. This is true whether the game is intended as real-time, or over a matter of days.

Please do not exclude the current ability to "play by mail", preferably with more advanced timer capabilities as described above.

A better ladder/scoring & tournament features would be easy to implement and goes a long ways to adding & keeping interest (both by garnering the competitive nature in most of us, as well as helping us to play against players at similar skill levels). Just playing random games with no goal in mind gets somewhat mundane after a while, no matter how good the game.

One thing this game lacks is the ability to involve more than 2 players (team or free-for-all).

The lifetime membership will be a big draw. I imagine the current pricing structure turns many away.

All I've played is MAN, so I don't know the current balance. But it seems to me that there should be some other option to shoot down airplanes, like an option to purchase anti-air.

It seems like a component of "fog of war" could be added which would offer many new possibilities for strategy, such that we don't see activities of our opponents, except for what is within a certain range of friendly units/territory.

My most controversial recommendation, is to consider a model in which damage isn't calculated until after the combat phase is "submitted", and includes a random component. Perhaps instead of a tank causing 2 damage, it causes between 0 and 4 points of damage, with 2 being most common, along a bell-shaped curve. This would completely change the game, in a way that at the very least would be interesting, and probably richer (adding a component of risk to strategy).

Author:  Quitch [ Wed Jan 18, 2006 12:27 pm ]
Post subject: 

Once you add randomness and fog of war you've:

a) Removed two of the things that draw the current fan base.

b) Made it a 3D Battle for Wesnoth.

Author:  RadRx [ Wed Jan 18, 2006 12:35 pm ]
Post subject:  Fog of War

Maybe Fog of War is not the right term. I was not suggesting we provide anything less than full information about units that could be seen (within range or shooting at your units).

I was suggesting that units that are well behind enemy lines and not otherwise engaging your units could be hidden. This would strengthen some additional tactics commonly employed in war (reserves, surprise attacks, ambush, etc) and would in my opinion add depth to the game. Such a feature could easily be turned on or off upon starting a game for those who had a strong preference.

I recognize that providing a random component even as an option is likely a controversial suggestion.

Author:  Rextrent [ Wed Jan 18, 2006 12:41 pm ]
Post subject: 

Anti-aircraft strength points could be added to the capitals and the 6 surrounding hexes( up to a certain point limit?) This would be nice....As the little glowing anti-air unit from Dom. is cool, very pretty, but kind of useless in most situations....It costs too much....and needs a 1-point attack ability....maybe that is one solution....instead of a glowing "magnet" barrier....more like an AA battery.....
Fog of war or "dice-roll" hit factors would give too much chance and there's already plenty of that....Of course, it could be optional....but sounds complicated for devs. 8)

Author:  Quitch [ Wed Jan 18, 2006 12:46 pm ]
Post subject: 

Can't say as I've ever found planes overpowered on anything except the small maps, and even then if you've allowed a build up of them then surely your opponent should reap the reward of this expensive investment.

I too would like to see team play and FFA. The reasons for this not being in, we are told, are two-fold:

1. The games would take too long.

2. Who would control the guerillas?

To 1 I say that it's entirely up to the players. If it takes too long then people won't play it outside of LAN mode, but personally I can't see why it matters how long it takes. If it takes a long time then I can start up more games to keep me occupied so that I've always got a few games on the go my side.

To 2, I don't see they simply can't be given to the next player whose turn it is as this is the only way of maintaining their present effect, and is also entirely non-random and allows for the sort of planning that the rest of the game allows.

To me, that's the next logical step for the game.

Author:  Rextrent [ Wed Jan 18, 2006 1:09 pm ]
Post subject: 

Multiplayer could be like RISK....just like current gameplay, except adding one or two more opponents....sounds much simpler and even fun....except for third player who is incapable and gives you the shaft!....or plays favorite with one player...but that would just be the...risk.

Author:  Maelstrom [ Wed Jan 18, 2006 3:36 pm ]
Post subject: 

This subject is getting too big and random. Made some comments about multiplayer in a seperate thread: ... 2412#22412

Author:  Rextrent [ Wed Jan 18, 2006 8:43 pm ]
Post subject: 

Yes, herr fuhrer!...the best thing about threads(in my beastly opinion) is when they "get interesting"....I've learned more reading threads that sprung leaks than ...well, you get the point....
I will look for a multiplayer's all chitchat anyway. Sometimes very informative and educational...."I just came in here for an argument!....Oh, that's down the hall....It's getting hit on the head lessons in here!" :wink:

Author:  storm440 [ Thu Jan 19, 2006 10:54 pm ]
Post subject:  MAN2

What happens to games that are in progress when you install the new beta download?

Author:  Maelstrom [ Fri Jan 20, 2006 12:30 pm ]
Post subject: 


You have 6 months to complete any MAN games.

Author:  storm440 [ Fri Jan 20, 2006 3:34 pm ]
Post subject:  rankings

will all existing players that join man2 keep their current win/loss totals and ranking or will everyone start from ground zero?

Author:  Nick_WN [ Sat Jan 21, 2006 7:56 am ]
Post subject: 


We integrated new ranking system based on Gliko rating for MAN2, which we believe to be more accurate. This system is still working in test mode and we might re-calculate ratings shall our programmer add more factors to count scores.

In any case, we were planning to transfer w/l for MAN players in MAN2. Now we 're waiting for MAN players to move in MAN2 community and then total re-calculation will happen. (I guess in 2-3 weeks time)

Kind Regards,

Author:  Quitch [ Sat Jan 21, 2006 8:06 am ]
Post subject: 

I hope you're going to integrate the score into the main panel at the bottom of the lobby. In fact, it needs to be in a lot more places, like public challenges etc. (e.g. rather than setting conscript to marshall you set 1200 to 1400)

Author:  storm440 [ Sun Jan 22, 2006 10:35 am ]
Post subject:  first man2 game

A few thoughts on playing my first man2 game. first, there really needs to be somesort of offline play where new players can play against the computer at there leasure and learn the game. the default 10 min clock timer is the dumbest thing i have ever seen and will do nothing but creat frustration in new players. I've played over 160 ma games and could not complete a move on a med size planet twice in the time allowed, I can not imagine the frustration a newbee must feel trying to play the computer for the first time. Except for trying to satisify the players that what to shoot first and think second and who resent playing people that catch there mistakes, why is there a time clock in the first place? The min time should be changed to 20 min and an unlimited game time needs to be added. There are a lot of fast twitch games like doom out there for the players that want a really fast paced game, why try to turn ma into one of them. Lastly, the game crashed on turn 14 and now generates windows errors that bring me back to the desktop every time i try to finish it. How can i correct this? it was against the AI, which was loseing badly, so I don't want to get an undeserved L out of this.

Author:  Quitch [ Sun Jan 22, 2006 10:54 am ]
Post subject: 

While I don't agree with a minimum of 20 minutes (10 would be a definate max if you had any interest in getting a quick game online... the MA equivalent of speed chess I guess), I do agree that it seems ridiculous that:

a) There's no "infinite time" option, especially seeing as how the majority of games will take place in a PBEM format, complete with ANOTHER timelimit for how long you can hold a turn before losing!

b) The default time is 10 minutes, especially for the "training" AI.

Why are the AIs only online? What if my net connection is down?

Page 3 of 5 All times are UTC - 5 hours
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group