My impressions and thoughts

[thiemo]

After a some 10 games my impressions and thoughts. Not sure if to distil them into feature requests as they might alter the character of mg entirely. I would happily discuss the points, however. If it would be best to move the specific points into separate threads or point to already existing threads I shall do so or feel free to do it yourself.

I think it is already a great game. I am really amazed once again how good the quality and functionality of OSS can be. I do not want to imply that OSS developers (and - sorry for not mentioning all on by one - others having their share with bringing OSS to us) are less gifted than paid one. I just refer to the thing that not too many people can afford to spend a big deal of time in unpaid work - I imagine.

My overall impression on MG is that it is quite close to WarCraft. The general setup is acquire resources, put them to good use and defeat the opponent is the strategic aspect that compels me.

However, I feel that the weakness of WarCraft is present here. To me it very much feels that the AI ranges from dumb to particularly dumb - I mean no offence with that and will detail in a second. My feeling about stronger AI is basically - as in WarCraft - that the AI has more initial resources and perhaps some production/harvest/mining speed advantage. I did not test it out. For example I just finished a custom build where on Tropical Arena with a CPU Mega. Eventually I lined up a couple of sphinxes behind a ridge of rocks. AI kept sending its units just along the other border and they were grilled by the sphinxes when passing them. The first unit got through almost unscathed but could be finished off by my other units easily. All following where sent through hell dying or getting out almost dead.

I am not happy with some behaviour of the units. E. g. they do not try to avoid fire if they move to a certain point. Then, the units with distance attacks fire to the spot where the enemy is at the time of firing not taking into account the enemy's movement.

To me, there is a big imbalance between strategy and tactics. Maybe this is inherent in the setup. There is no way a human can control its units as fast as the computer. I mostly find myself deciding, I have not got the time to do this or that. E. g. with an ongoing battle I neglect my harvesting and unit production. While being in a game me against an AI I can pause the game and only blame myself if not doing so. But in network games with more than one human player I cannot. Sure, other players are much, much better in controlling their units, I just do not know how - maybe it’s just me being slow because of age and lack of practice.

Coop games are not so cooperative in my opinion. The ones I have played felt like, each one tried to build up as much momentum for oneself and targets the enemy wherever she sees fit. Sure, one cannot oblige the players to cooperate closely but it is too easy not to do so and still succeed.

I think it is good that one can control everything of the units and it is bad that one is obliged to.

To me the current setup boils down to a battle of material in the sense of “who is able to produce faster than the other” putting a strong emphasis on the strategic aspects of the game thereby neglecting the tactics do to lack of deputising.

Some possible improvements (I do not repeat point I already found in the forum)
a) I would like to be able to tell a building to keep training unit XYZ as long there is enough resources where enough also could be a threshold of the resources defined by me.

b) Make the units smarter such that the majority tries to avoid fire if moving to a point. Maybe its ok for Golems and the like not to. Make the units with range attack take into account the movement of their target.

c) Make unit groups concentrate attack. If I group units they only act as a group if I explicitly tell them what to do. E. g. I tell group one to battle enemy unit x they do, but when that unit perishes, the group members start to fire “randomly” at enemy units. I do not see why they should not pick a common target again. It could be implemented as follows. I make some presumptions. All units get investigated one after another. For the inspected unit it is decided if the unit continues its action or if a new action is set up. If the unit loses its attack target it picks a new one. So if the unit belongs to a group it gets the group target if there is one, if not, it just picks one the way it is done already and that target becomes group target. I see that in the heat of the battle it is not so realistic that one can identify (immediately) the aim of the others of the group. Therefore one could introduce a random element to make some individuals aim not at the group target. One could argue that those dumb rookies from factory are just incapable of coordination but I doubt that reflects sort of reality. Sure they would improve in finding the group target with raising experience but the attempt should be there from the beginning. One could also define a leader unit - be it a special unit type, be it an especially experienced unit - that has to be present in a group to have the advantage of coordinated actions. I guess it also would be a good idea if the factions should have varying effectiveness of coordination. The romans for instance are know to have had very extensive military drill such that their discipline was very high being precondition for extended coordinated actions.

d) Give the units an operation mode where they retreat fighting such that they preserve health by losing ground. Problematic is to define where to they ought to retreat.

e) I would like the units stop immediately when they have killed there immediate target and there is no alternative target in sight - literally. Why do they run to the spot of death of their "victim"? To see how it decomposes? I usually do not want it and would like to have a button to deselect this behaviour.

f) I would like to tell units with range attack to fire behind the nearest line of enemy. I mean, I can have so much collateral damage I am not sure whether to use a sphinx, archmage, dragon, ... or not. Is the damage of a splash attack a function of the distance of the affected spot to the centre of impact? It would make sense to me.

g) I would like to be able to build up a firewall ;-) . I. e. that range attack units keep firing on a spot no matter if there is a enemy unit or not. On the other hand, if the units themselfs got smarter, they ought to avoid such spots (running into my next ambush ;-) ).

h) I would like to be able to give my units standing orders. E. g. my healers at times run into battle even though they are not yet affected by them or even in danger to get affected. So I would like to tell them that they ought to hold position and e. g. after having healed another unit, switch back to hold position.

i) I like the idea of healing very much but in my opinion it is too tedious to be much of use. I have to tell which healer unit is to heal which unit. Totally impractical especially in combat. I would like to see some hospital areas. I imagine several variants of these areas. i. Some radius of a "healer" unit the healer automagically starts to heal if a wounded unit enters it. ii. Mark an area on the map healing ground and staff it with healers. Every unit in that area with healers gets healed eventually. iii. Special hospital buildings that need to be staffed with healer units and units to be healed need to enter it.

j) Having healing areas I would like to be able to tell my units to seek healing and they move and look for it autonomously. After being healed they either return to their previous position or stay just outside the healing zone. So those could be to separate actions or one action with a switch.

k) I would like to be able to tell my units to employ guerilla tactics when attacking. E. g. I do not see common sense in it that an archmage stays longer than necessary in the brunt of the battle especially if he has run out of energy.

l) I would like to be able to give general guidelines of behaviour ranging from "keep out of trouble" over "have good own damage to enemy damage ratio" to "hack and slay no matter what".

Maybe my ideas would lead to a boringly slow game. Sure it is supposed to be real time strategy game but why is it possible to pause or change the speed of the game then anyway?

Cheers

Thiemo

[-Archmage-]

So essentially less micro and more strategy, with more intelligent units. Totally on board, you've highlighted some of the primary reasons I don't play MG(or most strategy games). Auto repair(in a different fork of the game: GAE) made workers maintain buildings and healers maintain troops without any attention needed, but the devs don't seem to want any unit autonomy at all. There was even a discussion about making newly spawned troops fight back if enemies are in range, instead of walking(non-combat movement) to the meeting point and getting slaughtered. As far as I know, that feature request has been rejected, so units will continue to commit suicide if your base is under attack. 

[thiemo]

:-) know these discussions. I actually do not see why one should not implement both such that the player can choose the behaviour.

[andy_5995]

If it would be best to move the specific points into separate threads or point to already existing threads I shall do so or feel free to do it yourself.

The main reason it works better to have each request in a separate thread is to make them easier to track the status and discuss them. You may see threads where multiple issues are mentioned, 1 or 2 are discussed but then other good points are missed or forgotten about because everything gets all jumbled into one big thread.

I'll move this into the main MG board, since it's not an actual feature request thread, but contains several points for users to discuss.

[wciow]

Hi, Thiemo. It seems your points mainly revolve around MG's AI and its simplicity.
IMO the problems with MG's AI fall into two distinct categories:

1, The overall AI when playing against a CPU opponent in terms of both strategy and tactics.

My feeling about this is that expecting a smart challenging cpu mode is asking too much. Every year lots of big budget games with dedicated AI programmers get slated for the stupidity of their AI and lack of challenge. MG is a relatively small FOSS game and will probably never have an experienced AI coder who could re-write it in any big way. IMO we should accept that cpu mode is basically a training tool and being able to crank up its resource multiplier to provide a brute force difficulty increase is fine. If you want a tactical challenge play another human instead.

I would also like to point out that the MG team have included the Mega cpu mode which operates slightly differently from the standard AI and a set of switches in the faction XML which allow modders to tweak the AI quite a bit to improve its performance with each faction. Sadly these settings have gone largely untried and untested up till now.

Another way to feasibly better the AI would be through the LUA scripting capabilities. The problem is that whilst this may work in 1p vs 1cpu scenarios it would be too slow for large online multiplayer games. I would however like to see a scenario dedicated to providing a responsive, tactical challenge in a local 1v1 game.

2, Making smarter units that do tasks automatically (repair, heal, build etc) and avoid suicidal attacks and generally react to each other in ways the player would expect.

This is the area in which I feel MG could be usefully improved. Some of the issues with this are not strictly AI but built into the way the game works. For example, units cannot automatically use more than 1 type of attack for each field. Giving a unit two attacks (with better strengths against units or buildings for instance) is totally ignored by the AI and a micromanagement headache for players. Each unit has to be individually selected and told which attack to use against which target. In a big battle this is basically impossible without constant pausing of the game. As you mentioned, units will not automatically heal or repair, each command has o be manually given by the player. Also units will not attack units outside their range even if they are being attacked  This can lead to situations where a unit will simply sit around being fired upon until dead.

One suggestion I particularly agree with is your point 'g)'. Units should be able to fire on empty spaces with splash attacks for artillery or similar weapons. I understand this might require quite major changes to the way attacks work, but IMO this would make gameplay much more interesting.

Finally I would  like to say that as a long-time modder, I appreciate the hard work that the MG team put in. MG is a fun and playable game already, by pointing out some of the above issues I am simply trying to give ideas that I think would improve the player experience 

[titi]

I admit I must really laugh about this. Nearly every new player I met is shocked about the horrible strong AI and you state its dumb and too easy to win .

The AI seems to be dumb in some cases yes, but very strong! Imagine the AI would be smart then you would nearly never win. We already have different AI behaviours because every type of CPU has a different behaviour not just a higher multiplier! For example Ultra and Mega fight a bit smarter by attacking standing units in range first. Of course they do this just in some rare cases, because I added a low chance of doing this ( ultra does it less often than mega does it). Imagine I would raise this to 100%. All your archers would be killed within seconds and you have no chance to controll this as fast as a computer can do this!
So in several places it would be really easy to make the AI smarter.

You already complain about wanting less micro management, I bet you would even cry more if I make the AI smarter and stronger. In the end this would lead to a half automated AI fights were you can make some strategic decisions, so a completly different game! If you want this you must play another game.
So I give a big NO to too much automation. Micromanagement is an important part of this game! There may be some exceptions which are too annoying but I really don't want a general change regarding this.

Beside of this I myself am pretty sure if the the AI would be smarter, the game would be a lot less fun!
This would mean you need to lower the AIs general production rate to give a human player at least a tiny chance to win. So just some units on the field and not those big attacking ( and yes often a bit dumb ) armies that attack you now! But would that be fun ? The motivation is a bit like in the "serious sam" or tower defense games, you as a player feel really good if you can stop a huge army because you fight smarter than the attacking AI! And you feel even better if you can do this with a human team in a coop game!

MegaGlest has its own style. If I should sum its up its a classic RTS game with strategy, micromanaged fights and a bit of tower defense, with a special eye on multiplayer coop games vs AI.

[Coldfusionstorm]

A good human can make decisions just as well and better than a AI.

Automation is bad and should only in cases where it makes sense, ex repair units standing next to a building repeairs nearest building and if you want to make the most out of it you can just select the right building you _really_ want to be repaired.

However the Pathing and Response time of units makes me want to kill myself. waiting for units to turn around and then doing a a slow a** attack makes me want to kill myself, non-litterally. (but this is for another topic (the unit thing).

[-Archmage-]

So I give a big NO to too much automation. Micromanagement is an important part of this game!

Exactly why I don't play this game anymore, or most RTS games, they all center on micromanagement instead of strategy. The ability to micromanage is great and fine, but the fact that I have to constantly stop units one by one from killing themselves is really not my idea of fun. I want to be the general, not the cohort counselor. 

[titi]

@Archmage: If you want to play a pure strategy games MG is indeed not the right game for you and it looks like you already found out that RTS is not your genre. But why the heck do you feel the need to state this here and tell us it's a bad game because you don't like the whole genre in general?
BTW it's not shooter, racing game or management game too, just in case the next one complains about it 

@Coldfusionstorm: That's correct about strategy, but MG is not a strategy game in first place. It has some strategy aspects and as I stated Micromanagement too ( in fight too ) this micromanagement can be handled a lot more effective by a computer so a human player will always fall back here. And as I already stated, no MG will not turn to a pure strategy game.

About the slow reaction of units, that's true but technically I see no way to change this. But everyone including the AI has this command delay so its fair. This might be a bit ugly for people who play Moba games where it is core feature that you can react very fast. But this is no moba game and no shooter where you only have to deal with one or a very little number of units and a server controls everything. RTS games handle tons of units and work technically different. I don't think there is no realistic way to handle this. Because this is a fact MG uses this as a part of the unit behaviour too. There are some unit types which react slower or faster than other units, so this must be taken into account too by the player.

[thiemo]

My feeling about this is that expecting a smart challenging cpu mode is asking too much. Every year lots of big budget games with dedicated AI programmers get slated for the stupidity of their AI and lack of challenge. MG is a relatively small FOSS game and will probably never have an experienced AI coder who could re-write it in any big way.

I am far from being expert with respect to AI. I certainly do not think MG should get a shiny self-learning neuronal network AI thingy that would overcome the player anyway. I see the AI rather as expert system where the designer of the game defines some situation patterns that if matched lead to decisions.

... and a set of switches in the faction XML which allow modders to tweak the AI quite a bit to improve its performance with each faction. Sadly these settings have gone largely untried and untested up till now.

I fear most of us are just lazy slobs like me being reluctant to delve into XML files.

Another way to feasibly better the AI would be through the LUA scripting capabilities. The problem is that whilst this may work in 1p vs 1cpu scenarios it would be too slow for large online multiplayer games.

Maybe the overhead for scripting or for more situation patterns to compare with gets compensated by less need for many many units.

I would however like to see a scenario dedicated to providing a responsive, tactical challenge in a local 1v1 game.

I second that.

Finally I would  like to say that as a long-time modder, I appreciate the hard work that the MG team put in. MG is a fun and playable game already, by pointing out some of the above issues I am simply trying to give ideas that I think would improve the player experience 

Agreed

--

... horrible strong AI and you state its dumb and too easy to win .

To me, strength and wits are not the same. The difficulty to overcome the AI is not its cleverness but its sheer advantage in resources -  if you just keep the multiplier when selecting the MEGA. I wonder how a MEGA 4.9 can be dealt with but I dare say that after having been given the hint to put the resource input at prime I was able to defeat MEGA 2.5, I should repeat it anytime. Is your experience different?

...Imagine the AI would be smart then you would nearly never win.

I don't agree there. If the AI was smarter, you would not have to put the mutliplier so high to still get a decent fight. As far as I have understood, the multiplier exists to compensate the lack of wits in the AI. BTW, I very much have the feeling that AI does not suffer from the fog of war. The AI headed to my base straight away. I have never experienced reconnaissance missions by the AI.

...For example Ultra and Mega fight a bit smarter by attacking standing units in range first. Of course they do this just in some rare cases, because I added a low chance of doing this ( ultra does it less often than mega does it). Imagine I would raise this to 100%. All your archers would be killed within seconds and you have no chance to controll this as fast as a computer can do this!
So in several places it would be really easy to make the AI smarter.

I quite agree that one would need to get some automation to compensate the AIs vast advantage in control speed of the units if one wanted a smarter AI

You already complain about wanting less micro management, I bet you would even cry more if I make the AI smarter and stronger.

I am sorry if my suggestions came across as complaints. It was not my intention.

In the end this would lead to a half automated AI fights were you can make some strategic decisions, so a completly different game! ... Micromanagement is an important part of this game!

Sorry, my believe is that it IS already half automated AI fights. I cannot imagine that one can control 50, 30 or even only twenty units in a fight - not efficiently at least. Sure, you can group the units but they tend very much to get different amount of damage so one might want to retreat all but the healthy and only lightly wounded. This is not possible without loosing control. Either, you select the ones to retreat one by one and order them out of battle to discover them shortly after in the fight again because you ordered their group on the next target or either you cannot us the group any longer, or you get the whole group out of combat or you just sacrifice some  of them while the group stays in action. Another example: My experience is that it is useless to put more than two or three melee units onto one single field target (the majority of the foe's units, I am sure). If I put more they get in each others way or - being as likely as former said - the are blocked by other enemy units thus running around them and getting hit for nothing. As for range attack units it is slightly different. I guess you can target about 5 units on a still single field enemy unit. If you put  more, the surplus range attack units need to go around the front line 5, being hindered by other units of yours or the landscape artefacts. Having said that a fairly decent army of 20 units would end up in at least 3 groups. Maybe I am just not quick enough but I just cannot control three groups in a battle. I just cannot distinguish which unit belongs to which group thus rendering groups almost useless for me. And - what I have seen in the coop games suggests that much more experienced players operate battles - not skirmishes - like me. They throw in as many units they can grab, retreat the whole lot if there is too much damage to a certain gut feeling "percentage" of units and hope apart from that most of them survive. I do not call this micromanagement. That is, politely speaking, just very crude tactics. The single unit certainly controlled by the computer instruction set for the unit and only now and then a human command.

If you want this you must play another game.

As I stated in my initial post. I did not put my points in as feature requests - andy5995 move the post thankfully to the right forum - because they would alter the character of the game vastly.

Beside of this I myself am pretty sure if the the AI would be smarter, the game would be a lot less fun!

This is subject to the eye of the beholder, imho.

This would mean you need to lower the AIs general production rate to give a human player at least a tiny chance to win.

I do not consider this to necessarily be a bad thing. After all, if the AI would get smarter siblings one was free to choose which way to go.

So just some units on the field and not those big attacking ( and yes often a bit dumb ) armies that attack you now! But would that be fun ? The motivation is a bit like in the "serious sam" or tower defense games, you as a player feel really good if you can stop a huge army because you fight smarter than the attacking AI! And you feel even better if you can do this with a human team in a coop game!

It can be fun to watch a cillion of units bustling about knocking each others heads off of which I only get some glimpses because there happens so much in so short a time. Personally, I get rather bored after the fifth time or so. I play games to influence things and an epic slaughter is rather a film to me.

[titi]

Just a reply to one thing:

Sorry, my believe is that it IS already half automated AI fights.

If you don't control your units and groups in a fight you will not get better. Of course it's possible to win just by focussing on unit production, but when you set a stronger AI this gets more and more unlikely. 
You maybe cannot control each unit, but you have to find your way to make groups and handle them, else its really hard to win vs stronger AIs.

For example if there suddenly show up some air units you have to shoot them down very fast or they kill you. One thunderbird can take out a whole roman army if noone shoots at it. Or if suddenly some catapults show up and you are trying to defend a totem. In this case you must take them out quickly or your totem is gone. Or moving melee fighters in front of the archers to make them shoot at you without getting killed too fast gives time to your archers to kill the enemy ones which are busy shooting at your melee unit. Or making small groups of your front archers and controlling them to fire at the enemy archers instead of the melee units attacking you.
Beside many other tricks this is what you control and do in a fight. Just producing units and sending them to the battlefield is definitly not the way the game is meant.

[Omega]

I admit I must really laugh about this. Nearly every new player I met is shocked about the horrible strong AI and you state its dumb and too easy to win .

Hmm, I feel like that's mostly because:

• The AI doesn't need to waste time making decisions. It can make them almost instantly. Means little downtime. New players? They get overwhelmed easily and thus aren't maximizing production.
• New players don't realize the importance of maximizing resource production. The AI can be much more effective at that simply because it knows this.
• Those both really add up to allowing the CPU to be an effective brute forcer. They can easily do a good job at producing many units. Of course, humans *can* do better, but for a beginner, it's easy for the AI to simply have that upper edge.

[titi_son]

Even though i kind of like the dumb AI of megaglest, i think its time for a new AI which does play a little more strategic. And i dont mean in fights at this point but thinks like the fact that you can distract the ai with one scout and delay the first fight by this. Or that the AI always expands too late. (Often you just wait and defend your base, until the ai does not have any new resources). I do NOT want to change or remove the old ai because it leads to funny games, but want a new AI which is smarter. (Call it Cpu EXPERT or whatever )

For the other ideas mention in this thread:
Yeah they are great ideas. But they dont fit to the megapack. To have those things for other (more strategic) techtrees would be nice, but as titi (as far as i know) does not really want to develop an engine (which may also be a little bit sad i must admit) but a game.
Because you need to bother about micromanagement, is in my opinion also the reason why we dont have things like poison attacks or the megapack not using area effects and other features, which already exist.