Forum Thread
by member
Poll: what game should i make? (18 votes)
  Posts  
New racing sim? or RTS? (Forums : Suggestions : New racing sim? or RTS? ) Locked
Thread Options 1 2
Jul 20 2012, 3:58pm Anchor

Hey guy's i was wondering if i should make a game based off the indy car series or an RTS game but i'm having trouble deciding and need some help(my third choice was going to be a game closer to something like the sims but a little more serious :P ) i've been teaching my self blender which is a little hard and i've been kind of getting ready to work on some concept art for the game

also if i were to make the game that would be like the sims

it will feature

similar gameplay but without the other stuff
graduall aging for charectors

Jul 20 2012, 5:02pm Anchor

Voted RTS, but after considering it, well, you could make something like The Sims, just with a different theme.

Edited by: feillyne

Jul 20 2012, 5:41pm Anchor

Like a Driving Simulator or something?

Jul 20 2012, 5:46pm Anchor

yeah something like that

Jul 21 2012, 7:29pm Anchor

RTS is extremely hard for beginners to mod for, a big part of the hurdle is getting the AI and Pathing to work just right. What you could do is combine all the ideas in a Roman era Racing Game with Chariots that features some social stuff on the side. (So basically Ben Hur the game :D).

Jul 22 2012, 9:17am Anchor

A Roman era racing game with chariots? Sounds promising. :D

Jul 22 2012, 10:54am Anchor

Roman era racing game would be cool, with upgrades, fictional vehicles, players able to attack, etc.

Jul 22 2012, 1:29pm Anchor

O.o no on the roman era racing game :P

Jul 22 2012, 3:24pm Anchor

A RTS game would be hard to make, with all of its army system ,oh and the AI and pathfindign for units... @_@
A RTS game could be a bit that is way more than to chew.. ;)
The Roman Era chariot racing indeed sounds interesting, could be interesting to "tune" your chariot with cool wheels and buy faster horses etc.

--

Think It. Sketch It. Create It. Repeat. 
User Posted Image

Jul 22 2012, 4:42pm Anchor

An RTS is easy to make. To get AI fast you need to get production functions right, then use randomised AI scripts ("intelligent" AI) or use scripted AI and scripted AI templates (so the AI builds everything using a specific strategy). Nothing really hard, pure fun...

Developing RTS games can be time-consuming though. -_-

Ashkandi wrote:Roman era racing game would be cool, with upgrades, fictional vehicles, players able to attack, etc.

Flash112 wrote:The Roman Era chariot racing indeed sounds interesting, could be interesting to "tune" your chariot with cool wheels and buy faster horses etc.

Indeed, plus some RPG/Sims/social elements could fit into this as well - as formerlyknownasMrCP suggested. :-P

Edited by: feillyne

Jul 22 2012, 6:59pm Anchor
Someone wrote:To get AI fast you need to get production functions right, then use randomised AI scripts ("intelligent" AI) or use scripted AI and scripted AI templates (so the AI builds everything using a specific strategy).

For your average joe this is adequate, but for anyone that knows where AI have their weaknesses its not suitable. If you're working with a game engine like UDK or Cryengine that have inbuilt AI you'll be fine, but if you have to bring your own in then its going to be a pain in the ass, further more making an AI play like a Human is extremely difficult (lot of the time you'll have to shoehorn it to force certain tactics and build strategies which isn't easy because it requires a lot of testing to get right).

thelastpharoah wrote:O.o no on the roman era racing game :P

Don't knock it till you try it lol.

Jul 22 2012, 7:20pm Anchor

Yeah, but it's still on the grounds of real-time strategies. For example, Age of Empires had relatively simple AI scripting and it worked - more or less. It just used that - scripted AI templates (rather than "human" AI).

Just like the difference between so-called static and dynamic bots. Doing AI for an RTS is hardly... hard, at least if you aim for moderately challenging AI.

PS Some developers allow RTS AI to "cheat" to be on a par with the player. :-P To allows unit spams etc. (An easy way to implement higher difficulty levels.)

Edited by: feillyne

Jul 23 2012, 4:46pm Anchor
Someone wrote:
PS Some developers allow RTS AI to "cheat" to be on a par with the player. :-P To allows unit spams etc. (An easy way to implement higher difficulty levels.)

I discourage this on every possible level, its a major hate of mine in RTS games and its bloody lazy.. it defeats the entire bloody point of game design if you're just going to go and do that (because the whole point of the game design is that both players are considered equal in ability- else it ceases to be a game rather just a boring unfair grind.) The goal of RTS AI is to throw something unexpected when the player can predict patterns in the play style- you'd have to factor in every possible tactic they could use and alternate depending on what the player does...if we're going for a fully dynamic AI anyway. Static AI only makes sense on a parent to child relationship (EG, if you were creating large formations of units similar to how AOE does it).

And IMO AOEs AI had some pretty major flaws, as a result you could still predict the AIs tactics and counter them rather simply, even to the extent of creating bottlenecks if you knew how. With an RTS the AI is the most important thing, I don't see how you can say its not hard given how many games have done it so damn poorly in the past. Even COH has its problems- the British AI for example is completely batshit retarded "DERP I'll move my base into theirs", which is important too, not all units will have the same behaviours- so factoring these into the AI is critical- fuck it up and it stops being fun.

If we're just making rush AI like Blizzard have done with StarCraft sure its easy, but its hardly fun. A competitive RTS is even harder to do due to constantly having to balance the game to suit player's feedback (if someone finds an exploit then its expected to be nerfed).

These are the main factors why I see RTS as a bloody hard thing to work on. At least with a shooter or an RPG we've got a lot of easy to implement templates, sure we can shoehorn an RTS into these engines we have available but its going to take a LOT of work to make it a great experience. I'd argue this is also a major reason why we've seen a decline in RTS and developers focusing more on action FPS/RPG hybrids instead- because the engines are more suited for it.

If its a custom engine though you might have better luck.

Edited by: formerlyknownasMrCP

Jul 23 2012, 4:58pm Anchor

Old C&C series developed by Westwood used triggers, squads (groups) and targets - plus units had specified threat levels.

Some pretty well-known RTS games had cheating AI, plus there various bonuses for AI units on harder difficulty levels. Would agree with you though that it defeats the game design, at least partially.

formerlyknownasMrCP wrote:If we're just making rush AI like Blizzard have done with StarCraft sure its easy, but its hardly fun. A competitive RTS is even harder to do due to constantly having to balance the game to suit player's feedback (if someone finds an exploit then its expected to be nerfed).

You know, that sparked an idea. How about pseudo-intelligent AI mirroring player's moves? Or mirroring player's moves from previous matches, when the player has won? On the condition that map / starting location / faction of the AI player is same as the player's was.

Also, when you got beta testers, they can record matches, some of their strategies and ideas can be incorporated into AI templates.

An AI editor is also a possibility. Maybe let players themselves create the AI? Make the templates? Test out strategies, exchange them? Completely in-game, ready to roll, and game simulation on very high speed to see how AI copes with it.

There are loads of possibilities - you can always find a method to make developing AI child's play.

Edited by: feillyne

moci
moci Is it me or does it smell funny here?
Jul 23 2012, 5:25pm Anchor
feillyne wrote:There are loads of possibilities - you can always find a method to make developing AI child's play.

Sorry but making AI and childs play do not go together, "artificial intelligence" does not mean creating a few lists and have the computer follow it. I'm completely behind what formerlyknownasMrCP said. I'm sure you can see that what you are proposing is not AI so don't call it that.

I say go for the car game at least there you'll have a certain set of physics to implement and have something going fairly quickly. If you want to do an RTS or a proper SIM you'll probably never finish.

Also, feillyne, if you think creating "AI" is so simple go help 0 A.D. . Again what you are proposing doesn't have any "intelligence".

--

- Mathiasverboven.net

Quote:If I learned anything about Valve over the years, its that they can't count to 3 ~xHotPotatox @ joystiq
Quote:If people are making shit, they're making shit not because of the engine, but because their own abilities as a game developer suck.
They'd make similar shit on UDK, Source and CryEngine if they could. ~Mr_Cyberpunk
Jul 23 2012, 5:32pm Anchor

Your attitude, skills and time matter. A simple question: have you modded any RTS games?

Jul 23 2012, 7:03pm Anchor

I get what Feillyne means and its actually an interesting idea. No matter what, AI has to be "Taught" how to play a game, allowing them to actually look through certain player tactics (so those that the QAs have used in the past and recorded them in replays) and then decide when the best course of action to trigger them is likely going to be more natural than a computer doing it, the illusion has to be that you're playing against a player, in that scenario you actually are, the AI is just shifting through what it has availible to it at the time- that way it wouldn't feel so machined in.

(a perfect example of Machined in is Total War... god the AI is so fucking shithouse in all their games, it doesn't feel like an intelligent person, it feels like a computer.. "MOVE X TO Y, MOVE Y TO X, whoops they attacked my flank SPIN AROUND 360" end of the day we want to mimic the same behaviours a player has, not give the AI a seperate set of rules all together). I think its actually a pretty cool idea- but you'd have to keep testing it in order to keep up with your players.. Still I'm actually surprised Starcraft 2 didn't use such a system, practically all their matches are by the book so using recordings would actually make more sense than their stupid as shit "Rush" AI.

Lets just say its a better idea than what we've got now lol.

Someone wrote:Also, feillyne, if you think creating "AI" is so simple go help 0 A.D. . Again what you are proposing doesn't have any "intelligence".

I don't think any AI on the planet has actually shown "Intelligence" (Intellgience meaning its able to process what it see and create a solution from nothing) lol, you have to feed them the rules of the game and they need to determine the best path/routine to use to get the solution- thats the "Dynamic" part of the AI, but everything else is pre-designed- it still conforms to a set of predefined rules and procedures. With Chess AI they put in a bunch of strategies, and the AI looks at what the other player has done and goes "Oh I know how to beat that.. Its Routine XX1245642" and then spits it out (funnily we only know that this routine would win because a player had proven it beforehand), most game AIs aren't much different- its just that they rely on dynamic pathfinding, and that's where the shit really hits the fan most of the time (because if you can't get pathing right, your AI is going to be permenantly broken no matter what you do- see Total War for example :D). Its very tricky but I do approve of using Player replays as a form of building a procedure/tactic database for the AI. At least then its actually replicating real human behaviour.

Edited by: formerlyknownasMrCP

Jul 23 2012, 7:46pm Anchor

Exactly, it needs to feel like you would play against a real player.

Considering now a few other things as well, such as counter-tactics and counter strategies/procedures, something along the lines of:
- player rushes, AI immediately turtles
- player turtles, AI instantly rushes (or player turtles, AI expands)
- player expands (boom strategy), AI expands as well in exactly the same manner (or player expands, AI rushes)
would it feel like cheating if the fog of war is implemented/present and AI knows your every move?

Glad you mentioned the example of Total War. It seems pathfinding would have to be totally combined with constant scanning of human behaviour, as well as implementing procedures from the database. So either AI scanning player's moves and production queues, so it adapts and mirrors them on the go, or replicating past player's strategies.

Edited by: feillyne

Jul 24 2012, 6:57pm Anchor
Someone wrote:Glad you mentioned the example of Total War. It seems pathfinding would have to be totally combined with constant scanning of human behaviour, as well as implementing procedures from the database. So either AI scanning player's moves and production queues, so it adapts and mirrors them on the go, or replicating past player's strategies.

It doesn't. I've studied what the AI does before very critically, pretty much all it does is has Agressive, Passive and Defensive behaviour triggers (if you've played Medieval 2 Kingdoms before, the AI controls for that game basically show you how the AI works). If set to Passive, it sits there and does nothing, if set to defensive it'll try to use ranged attacks/fortifications, and if agressive it just rushes the general/flag (ignoring anything in its way). Occasionally it'll flank but usually only to attack artillery/seige equipment. It is hopeless in seiges because its always set to Agressive AI and ALWAYS rushes the flags, to beat it simply create 1 bottle neck surrounded by buildings using a single unit... BOOM, there goes all their army (no joke I've won battles in total war non-exploitive tactics with just a single unit against a full stack because of how shit their AI is). Its the worst example of AI I've ever encountered and CA should be ashamed of themselves for claiming "NEW AND IMPROVED AI" every game despite it never being improved.

Someone wrote:- player rushes, AI immediately turtles
- player turtles, AI instantly rushes (or player turtles, AI expands)
- player expands (boom strategy), AI expands as well in exactly the same manner (or player expands, AI rushes)

This is assuming we're making a competitive twitch RTS like starcraft, which that'd make sense given that strategically those games are by the book and the victor is whomever can run the procdure the fastest- I find games like that tedious and boring. If we're making a game though that was more a RTT and more about what kind of tactic the player employs then we'd want to go for an AI that is more experimental and a lot less predictable- the goal here is to completely throw the player off, so the AI should be more "Try it and see if it works", if it fails, it tries something new, if it succeeds then it keeps doing it till the player looses. Getting recon info would be really important too, since the AI would have to keep an eye on the player and think "What's he up to"- that way our AI is actually thinking "Hey I know how to beat that" catching the player by surprise.. its the players ability to adapt to what ever the AI has done and the players ability to conceal their strategy that ultimately determines if the player is victorious or not, the AI would no doubt also be conceling their strategy as well trying to keep the player on edge. That's at least how I see a good AI.

Someone wrote:would it feel like cheating if the fog of war is implemented/present and AI knows your every move?

100% cheating. The AI should conform to any game design limitations that are imposed on the player. Else, its a cheating AI and the player will be able to tell. If the player has to rely on recon intel to win, the AI should have to as well. FOW shouldn't be too hard to work with anyway, just chop the map up into a grid with sectors and tell the AI to go explore each of the sectors, prioritizing sectors with "known" enemy activity.

Edited by: formerlyknownasMrCP

Jul 24 2012, 11:15pm Anchor

i'll make the RTS since it has the most votes and romans are cool :D

Edited by: Legionstudios

Aug 1 2012, 6:51pm Anchor

Both indie car series and RTS have the same amount of votes. :-P Picture for proof in case something has changed. :-P

So the chariot racing game seems like a more likely option. ^^

formerlyknownasMrCP wrote:This is assuming we're making a competitive twitch RTS like starcraft, which that'd make sense given that strategically those games are by the book and the victor is whomever can run the procdure the fastest- I find games like that tedious and boring. If we're making a game though that was more a RTT and more about what kind of tactic the player employs then we'd want to go for an AI that is more experimental and a lot less predictable- the goal here is to completely throw the player off, so the AI should be more "Try it and see if it works", if it fails, it tries something new, if it succeeds then it keeps doing it till the player looses.

Hmm more experimental seems like less "intelligent" in the case of AI. AI should look like it has some sort of intention of being victorious.

Less predictable, maybe, but if there were loads of precise and challenging procedures/tactics to choose from for the AI, at least tens or hundreds of them, then AI would be much less likely to predict. As long as these are highly tested ones or replicated & working player's strategies.

formerlyknownasMrCP wrote:Getting recon info would be really important too, since the AI would have to keep an eye on the player and think "What's he up to"- that way our AI is actually thinking "Hey I know how to beat that" catching the player by surprise.. its the players ability to adapt to what ever the AI has done and the players ability to conceal their strategy that ultimately determines if the player is victorious or not, the AI would no doubt also be conceling their strategy as well trying to keep the player on edge. That's at least how I see a good AI.

formerlyknownasMrCP wrote:100% cheating. The AI should conform to any game design limitations that are imposed on the player. Else, its a cheating AI and the player will be able to tell. If the player has to rely on recon intel to win, the AI should have to as well. FOW shouldn't be too hard to work with anyway, just chop the map up into a grid with sectors and tell the AI to go explore each of the sectors, prioritizing sectors with "known" enemy activity.

What/how about
1) an RTS without FOW?
2) no free base building, just building on specific fields after a settlement is founded? (to avoid walling/wrapping AI around and many other smart stratagems by real players - many of them were performed during Age of Empires II matches, one can see all these replays of pro gamers just for some serious jaw-droppery)
3) regiments/troops by default visible to all players, but that could be hidden through a specific movement type / action?
4) recon not based on unit's innate ability such as scanning area in search of traps all "by default", but rather on a movement type that has to be selected first and that has huge impact on regiment's speed as well as providing an increase of its defensive ability
5) players able to use special god-like powers to scan an area for hidden/stealthily moving regiments (something similar to god powers in Age of Mythology) at the great cost of the last 4th rare resource?

Something alike a web browser based strategy game, but fast, condensed & completely in real-time, without single recon units. (Only standard units that can do reconnaisance.)

Dunno if you have played Kohan - AI in that game was pretty difficult and challenging, many bottlenecks were avoided in a very smart way thanks to a game design.

Edited by: feillyne

Aug 1 2012, 11:37pm Anchor
Someone wrote:5) players able to use special god-like powers to scan an area for hidden/stealthily moving regiments (something similar to god powers in Age of Mythology) at the great cost of the last 4th rare resource?

this would be acceptable since its a game design thing not so much an AI thing, the AI should ideally for the very least have a basic understanding on concealment- else its never going to beat a player.

Someone wrote:4) recon not based on unit's innate ability such as scanning area in search of traps all "by default", but rather on a movement type that has to be selected first and that has huge impact on regiment's speed as well as providing an increase of its defensive ability

Depends on if the Unit is vehicular or not, if it were infantry it wouldn't make much sense to slow/speed them up, for vehicles it'd make sense though since they can vary in speed (with recon units being lightly armored and faster).

Someone wrote:
3) regiments/troops by default visible to all players, but that could be hidden through a specific movement type / action?

This is actually a pretty good idea, but it leads to an elimination of surprise (see Total War Multiplayer for example- sure you can conceal units but not during battles, only prior to battle by positioning them inside of forests)

Someone wrote:2) no free base building, just building on specific fields after a settlement is founded?

This is also a great idea but would really limit where the player and AI could build, if you were using premade maps it'd lead to constant rushes for "Ideal" areas (see Civ4 for example where everyone always rushed for areas well known to contain oil ect).

Someone wrote:1) an RTS without FOW?

Could be done, as is the case with number 3- but it might not be too challenging nor realistic. That said a better idea would be to restrict the camera, Total War actually had this option (though few ever used it) where rather than use FOW you were restricted to having the camera tethered to only visible areas. This would then mean the AI would have to be limited based on Range.. could get messy though (similar problems with FOW would appear here).

Someone wrote:
Hmm more experimental seems like less "intelligent" in the case of AI. AI should look like it has some sort of intention of being victorious.

Except that the players definition of victory is different from the AI's definition of victory- the AI would focus only on one thing and that is annihilation, whereas the player would be focusing on things like building up a surplus and fortifying their base for the eventual base rape (then turning the tables on the AI with the surplus). This kind of RTS is boring and predictable as the AI is always the aggressor never the defender (Which then leads to them essentially giving up when the player has the advantage). By experimental I mean that the AI is going to think more a long the lines of "what order do I want to hurt the player in" and then it randomizes the tactic it'd use to process the order, as opposed to just thinking "I want to kill the player, so I'm going to find their base and destroy it" or in the case of Total War "I want their flag".

Someone wrote:Dunno if you have played Kohan - AI in that game was pretty difficult and challenging, many bottlenecks were avoided in a very smart way thanks to a game design.

Yeah I've played it, it was pretty cool. That said it still focused on base rape as a means of winning but at least limiting the buildings was a smart idea since you can't create bottlenecks.

Edited by: formerlyknownasMrCP

Aug 2 2012, 4:58pm Anchor
formerlyknownasMrCP wrote:Depends on if the Unit is vehicular or not, if it were infantry it wouldn't make much sense to slow/speed them up, for vehicles it'd make sense though since they can vary in speed (with recon units being lightly armored and faster).

Infantry, environment: jungle (yeah almost completely). So traps concealed in tropical forests.

formerlyknownasMrCP wrote:This is actually a pretty good idea, but it leads to an elimination of surprise (see Total War Multiplayer for example- sure you can conceal units but not during battles, only prior to battle by positioning them inside of forests)

The whole idea is along the lines of huntsmen able to use:
1) regular movement, 4/4 speed, visible to players, no modifiers
2) stealth, 1/4 speed, invisible, initial bonus to attack
3) scanning, 1/4 speed, visible, initial but considerable bonus to defense
4) regular movement + expensive supernatural stealth power used, invisible, initial but huge bonus to attack

As mentioned above, it's all inside tropical forests. Huntsmen/tribesman would have to be actively hiding.

Also, food would be constantly consumed by regiments during movement, little by little. (3/4 food consumption when stationary, 2/4 when inside a specific supernatural/special zone, 1/4 when garrisoned inside a settlement.)

formerlyknownasMrCP wrote:This is also a great idea but would really limit where the player and AI could build, if you were using premade maps it'd lead to constant rushes for "Ideal" areas (see Civ4 for example where everyone always rushed for areas well known to contain oil ect).

How about randomly selected danger zones (quicksands, wild animal hunting zone) & access fields to resources (quarries, etc)?

formerlyknownasMrCP wrote:Could be done, as is the case with number 3- but it might not be too challenging nor realistic. That said a better idea would be to restrict the camera, Total War actually had this option (though few ever used it) where rather than use FOW you were restricted to having the camera tethered to only visible areas. This would then mean the AI would have to be limited based on Range.. could get messy though (similar problems with FOW would appear here).

... and a god power to hide an area, a miniature player-controlled version of fog of war?

formerlyknownasMrCP wrote:Except that the players definition of victory is different from the AI's definition of victory- the AI would focus only on one thing and that is annihilation, whereas the player would be focusing on things like building up a surplus and fortifying their base for the eventual base rape (then turning the tables on the AI with the surplus). This kind of RTS is boring and predictable as the AI is always the aggressor never the defender (Which then leads to them essentially giving up when the player has the advantage). By experimental I mean that the AI is going to think more a long the lines of "what order do I want to hurt the player in" and then it randomizes the tactic it'd use to process the order, as opposed to just thinking "I want to kill the player, so I'm going to find their base and destroy it" or in the case of Total War "I want their flag".

formerlyknownasMrCP wrote:Yeah I've played it, it was pretty cool. That said it still focused on base rape as a means of winning but at least limiting the buildings was a smart idea since you can't create bottlenecks.

Yeah totally agreed, AI should be variegated somehow and less aggressive.

On the other hand, Red Alert 2 and Tiberian Sun had AI organised in small groups to harm player - but these groups were much less effective than they should. Should rather AI build up some more (turtle a bit, expand a bit) then attack with a more concentrated army?

Edited by: feillyne

Aug 3 2012, 2:25pm Anchor

How will you make a game if you know a litle about Blender? Are you a programmer?
Also the RTS genre is a little crowded, it will need something really fresh to set it apart.

For an indy racing sim do you have the required programming skills to simulate the physics?

Aug 6 2012, 10:56pm Anchor

no im not a programmer sorry ): just a modeler with an idea/concept

Reply to Thread
click to sign in and post

Only registered members can share their thoughts. So come on! Join the community today (totally free - or sign in with your social account on the right) and join in the conversation.