• We have updated our Community Code of Conduct. Please read through the new rules for the forum that are an integral part of Paradox Interactive’s User Agreement.

SteelOnyx

Corporal
6 Badges
Dec 26, 2019
25
0
  • Pillars of Eternity
  • BATTLETECH
  • BATTLETECH: Flashpoint
  • BATTLETECH: Season pass
  • BATTLETECH: Heavy Metal
  • Naval War: Arctic Circle
Even on the hardest settings, the game is entirely too easy and one dimensional. The following could help to change that and make it much more immersive:

  1. Improved AI
  2. Dynamically adjusting the Mech tonnage on story line missions to the tonnage deployed
  3. Multiple lances on the same mission
  4. A reason to actually use light and medium mechs as part of a mission strategy once the heavies become available
  5. Change up of the mech load outs
  6. World Play

Improved AI
The AI is entirely too predictable and doesn't behave anywhere close to the way real people would in a similar situation. If a lance just had two of it's mechs completely blown away and were still facing double the tonnage they would not charge forward, they would run. Most people are not suicidal, the AI should not be either and it would make it an additional challenge in the game.

In the same vein give the AI pilots names and make it so they can eject or run away if possible. It would make the game more personal than some random faction head appearing in a pop up message to tell the player how they will make us pay. It also would contribute to the strategy there would be some incentive to keep the AI warriors from amassing more XP with each encounter.

In attacking, a lance would not send the faster mechs ahead to battle the opposing lance by themselves only to get blown away by greater fire power. Put some logic in place so it actually checks what the odds are (based on sensor data) so the AI doesn't march its mechs single file into a hail of gunfire to uselessly die. Maybe the quick and dirty is to only have the AI deploy mechs with relatively the same run and jump capabilities.

Provide the AI with some actual strategy, at times they just seem to randomly attack the nearest available target instead of doing any kind of threat analysis to see which mechs it should target first. Some people go for the lightest armored first, others go based on who can do the most damage.

Give the AI the ability to target critical areas on player mechs. If I see a mech coming at me that has an AC20 or multiple PPC's, the first thing I do is look for the area on the mech to target with the least amount of armor that is easy to hit with a called shot and will disable the weapons.

Stop with multiple mechs in a lance opting to sensor lock. In my own play style I found sensor locking useful early in the game, but once the range enhancers are available it is a tool I almost never use. However the AI seems to love using it later in the game, I have been in situations where 2 out of 4 AI 80 and 90 ton mechs opt to hide behind a boulder and sensor lock me instead of actually firing or at least maneuvering to a position that will give them a better firing solution next round.

Lastly it would be nice if the game adjusted AI to the play strategy of the player. For example in my play style I almost never close to a range of less than 400m. It is stupid for the AI to throw a couple of Crabs mounted with AC20 at me since I have usually blown off their arms, legs or torso before they can even get in range to use them. To counter the AI should either switch more often to specialty long range mechs or mechs capable of closing the gap faster. The point is it should be tracking what the player does and adjusting accordingly especially if we have fought against that faction multiple times.

Dynamically Adjust Story Line Mech Tonnage
I usually get access to heavier mechs fairly early on because I go for the harder missions. The result is that by time I get to the story line missions my lowest tonnage mech is significantly higher than anything the AI deploys.

Multiple lances on the same mission
The original PC game supported multiple lances and from a systems standpoint there are multiple missions in this version where there are 12 or more mechs on the screen at any given time.

Late Game Light and Medium Mechs
At a certain point in the game keeping light and medium mechs becomes a waste of bay space and cash because they serve no strategic purpose later in the game. If there is a mission that requires a speed element, I just grab a heavy or assault mech capable of hitting 120 and great jumping. It would be nice to have situations that actually require the use of those mechs which would be made much easier with being able to have multiple lances.

Change Up the Mech Load Outs
The game appears to keep the AI mechs loaded out according to the base variants which in many cases means more weapons than they will actually ever use in combat resulting in wasted tonnage that could be used for extra armor or cooling thereby making them more effective in combat. While a mech having a balance of long, middle and short ranged weapons is nice on paper, more often than not it just winds up in them doing mediocre damage under AI control as the player deploys mechs and tactics for the purpose of maximum damage.

World Play
The original BATTLETECH is focused around putting a drop ship on a planet as a base of operations and accomplishing the strategic objectives, i.e. topple the planetary government. The missions would be a lot more engaging if it was up to the player to determine where to set up base camp on a planet, decide which targets to hit first, ie bases, factories, power plants, etc. while at the same time having to defend against counter attacks. It would make for nice level progression as the first mission might be to capture just a single town on a planet and eventually working their way up to a full planet assault.
 
Upvote 0
MechCommander. I know different company etc. but it was a manage the lance game as opposed to a simulator like Mechwarrior. I originally thought it was called Battletech as well, but that is just my memory fading with age.
Mechcommander isn't the original PC game.
It is also Real time point and click rather than a turn based strategy game.
While it does have more than 4 available units playable, it's not a lot more either.
I also think it's fair to say that the Mechcommander AI being more action oriented didnt take into account as many variables as this games AI which is one of your key critiques.


However, Mitch from HBS was also involved with the Mechcommander series, fyi.
You might recognize some parallels in pilot chatter especially. Its certainly a fun game in its own right.
 
Mechcommander isn't the original PC game.
It is also Real time point and click rather than a turn based strategy game.
While it does have more than 4 available units playable, it's not a lot more either.
I also think it's fair to say that the Mechcommander AI being more action oriented didnt take into account as many variables as this games AI which is one of your key critiques.


However, Mitch from HBS was also involved with the Mechcommander series, fyi.
You might recognize some parallels in pilot chatter especially.

Yeah, but I have some fond memories of that game with many hours of game play so it is my go to reference, although I'm probably rating it on a curve. I was just glad to go from a 4x10 piece of cardboard with foam hexes for mountains and miniatures to a PC. A shame Star Fleet Command never got the same treatment, the one try was more like a sim than fleet battles.
 
Yeah, but I have some fond memories of that game with many hours of game play so it is my go to reference, although I'm probably rating it on a curve. I was just glad to go from a 4x10 piece of cardboard with foam hexes for mountains and miniatures to a PC. A shame Star Fleet Command never got the same treatment, the one try was more like a sim than fleet battles.
Star Trek: I had a lot of fun with this one (images courtesy of google search)
pic175656.jpg


pic56566.jpg


pic56562.jpg
 
Regarding developing better AI I guess I don't fully grasp the challenge.
Not many people do who haven't studied AI, or worked with it. It is quite a bit harder of a problem than it might intuitively seem.

Let's for example, take the aforementioned "move towards the enemy and fire" AI, that most would call "simple". It has a number of hard problems to solve:
  • Which enemy should it walk towards? The closest by line-of-sight? The closest in movement cost? One that isn't closest? Which one in that case?
  • Should it jump or move on the ground? Walk, run, or sprint?
  • If it can reach an enemy, should it still fire its weapons or should it try a melee attack?
  • If it can't reach any enemy, which should it move towards? Does it try to move to an optimal range for its weapon loadout, a maximum range so as to minimise return fire, or a minimum range to be able to fire close-range weaponry? Can it get too close?
  • Does the state of the enemy matter? Should it try to move towards and fire at the weakest enemy or the toughest? The most damaged or the least damaged?
  • Should it take into account whether it can conceivable one-shot any enemies?
  • Which weapons should it fire? Does its heat levels play into this? The enemies evasion or other defenses?
  • Does what terrain it ends up in matter? Should it prefer to keep to cover or is it fine to walk in the open? Should it ever break line-of-sight?
  • What if it doesn't have line-of-sight to an enemy?
  • What if there are no enemies on sensors?
These are just off the top of my head, there's plenty more problems that need to be solved even for a very rudimentary "move towards the enemy and fire" AI. All of these factors and more needs to be mapped into a decision tree and weighted correctly, so the AI behaves "naturally" in as many situations as possible. And of course, code has to be written to actually make each and every one of those decisions, with tie-breakers when two or more options are equally strong.

It's a lot of work, and it's not easy.

As a side note, @Timaeus posted above that the BATTLETECH AI takes over 50 factors into account for its decisions, what he didn't mention (or at least didn't emphasise) is that the AI evaluates all these factors for every movement point it can reach before deciding which one to move to and what action to take once there. So for a Locust, as an extreme example, that's quite the number of calculations that are made.

in many ways the we are still in the same relative space in terms of AI even though we have I7 and I8 processors compared to the Pentium 2's of that era.
AI is not necessarily a problem that gets easier with more computing power. Parts of it does, but not all of it. Also, the AI of MechCommander 2 wasn't even close to the AI we have in this game.
 
MechCommander. I know different company etc. but it was a manage the lance game as opposed to a simulator like Mechwarrior. I originally thought it was called Battletech as well, but that is just my memory fading with age.

As noted, Mechcommander wasn't the original PC game.

That title, I believe, belongs to The Crescent Hawks Inception.

As for AI.

Somehow I don't think the people clamoring for "better AI" will be happy until we're battling against Skynet or WOPR.
 
...Somehow I don't think the people clamoring for "better AI" will be happy until we're battling against Skynet or WOPR.
A “Better AI” would make a better BATTLETECH Difficulty Setting than replacement for the Vanilla AI. : )

While I would gladly fight HBS’s WOPR, I have no interest in seeing hordes of BATTLETECH Younglings slaughtered in their first attempts at playing the game. :bow:
 
@Prussian Havoc

Very true. That way, as @stjobe said, it can/could be controlled by the player so that those that may not be as adept at handling the AI could still enjoy the game while giving those that don't seem to have a problem beating the AI can have their challenge too.

On another note, I do hope we eventually hear something from HBS about the Multiplayer aspect of the game since that was suppose to wait till "after launch" and we haven't heard anything about it since. Or did I miss it?
 
@Prussian Havoc

Very true. That way, as @stjobe said, it can/could be controlled by the player so that those that may not be as adept at handling the AI could still enjoy the game while giving those that don't seem to have a problem beating the AI can have their challenge too.

Scaling AI based on a difficulty setting is one of the things I have been dying to see in a game. Changing the difficulty on a game should not simply mean the game just throws more enemies at the player with better stats, it should mean that as the difficulty scale increases the AI makes better choices.

In 2015 computers finally started winning at GO with no stone handicap, surely some of that know how is going to make its way into the strategy gaming world.
 
@Prussian Havoc

Very true. That way, as @stjobe said, it can/could be controlled by the player so that those that may not be as adept at handling the AI could still enjoy the game while giving those that don't seem to have a problem beating the AI can have their challenge too.

On another note, I do hope we eventually hear something from HBS about the Multiplayer aspect of the game since that was suppose to wait till "after launch" and we haven't heard anything about it since. Or did I miss it?
You did not miss it, as it has yet to occur.

HBS stated, “We recognize that Solaris VII is a great setting for BattleTech experiences, and in success, we still hope to explore different ways to bring Solaris VII to life.”

It is my fervent hope we hear something about this and the remaining Kickstarter Social Media Rewards (the four “Hero Lances” for Skirmish Mode - Sword of Light, Fox’s Teeth, Eridani Light Horse and the one already in the mini-Campaign/but not yet in Skirmish, the Black Widow Comapany) in 2020. : )
 
Scaling AI based on a difficulty setting is one of the things I have been dying to see in a game. Changing the difficulty on a game should not simply mean the game just throws more enemies at the player with better stats, it should mean that as the difficulty scale increases the AI makes better choices.

In 2015 computers finally started winning at GO with no stone handicap, surely some of that know how is going to make its way into the strategy gaming world.
Oh sure, most definitely at least some AI techniques coming from the computers that can win at GO and Chess will come to everyday games; however, the AIs that can beat grandmasters are also running on massive super computers with as much computational power as can be given them. For reference Alpha Go, the AI that can win Go without handicap, in one version (before custom AI hardware) was running on a supercomputer with 64 search threads, 1,920 cpus, and 280 gpus and was doing all of it's move calcs in 2 seconds. We have 1 cpu, 1 *maybe* 2 gpus (which aren't being used for the AI at all), and 1 thread for AI calculations over 20(?) seconds. Economy of scales there.