Jump to content

Development Update - 9th November 2012


Chris

Recommended Posts

Probably time for another update here. This week I've mostly been working on the level design stuff in preparation to start iterating the maps with the community. I've got three main maps in production with their layout mostly done - one Industrial and one Farm map that can be used for Light Scouts, Scouts and Corvettes plus a small alien base map. These will be the first few maps we'll be iterating.

There's a few things holding up their release. Firstly, there's a few annoying issues with the Level Editor we'll need to resolve (should be done today and tomorrow) but will vastly increase the usability of the program when they're finished. Then we need to set levelsetup.xml to control the starting equipment of the Xenonaut troops in the mission too, as this'll be necessary for the small alien base mission. I figure the player should have laser weapons and Wolf armour by the time they encounter bases.

Thirdly, we're being held up by the AI. It's not strictly necessary to play the maps, but would add a lot. GJ has been having a few issues with the way the game mechanics work. Apparently it's unusual for a game to have random outcomes in the AI turn; for example in the new XCOM if you save the game and press "End Turn", exactly the same series of events should happen if you loaded the game up a second time and pressed "End Turn" again. The downside of this is that the AI knows in advance whether it is going to kill one of your men or not, and can plan around it. In our system, the AI has to react to whether its own shots get lucky or not, which is difficult. So he's busy with that at the moment.

Oh, and I'm playing with the layouts of the new UFOs using the new UFO interior / exterior system that was discussed on the forum. These layouts will be in the new maps for community iteration, but they're blocked out using soil blocks and generic ground tiles. The implementation of this is nearly done.

I'll talk about the plans for our iterative testing at the bottom of this post.

Next up, our ground combat coder has been working on the Desura file packing stuff. There's good news and bad news. The good news is that it works, and will reduce the Desura files down to a workable level. The bad news is that the game engine is once again causing problems. Essentially, it can only ever support 1.3gb of packed files, so we're not going to be able to pack up all of our spritesheets so we'll still have tens of thousands of files.

Our ground combat coder has therefore been working on a tool that lets us filter which files go into the archive and which folders they are drawn from. Basically this just lets us stick the .XML files associated with every spreadsheet and all the ground tiles and associated XML files in the archive, which comes to about 250mb and reduces the files number of the dev version from 120,000 files to about 40,000. We can reduce that still further just by packing up more files into the remaining 1gb of archive space we have, so I think our Desura issues may be a thing of the past.

The only complication with this is that you'll likely have to download the standalone if you want access to all the game files for modding purposes, as that won't be packed up. But that's a relatively niche thing, so it shouldn't affect too many people (installing mods is still easy even with the files packed up, don't worry).

Excitingly, I also implemented the company's accounting system on Monday. This is basically because I had to do a load of admin to get the payroll set up now Aaron is here and has a salary (I don't have a salary), which means both he and the employer have to a pay a variety of taxes on it.

We've also sat down and had a bit more of a think about the new UI for the game. Giovanni is on holiday this week but is in the process of implementing the the new GC UI background, so that should be in the next version. We're also getting the topbar of the current Geoscape UI redone by the same guy to be prettier and easier to use, which will be a prelude to the final upgrades of all of the screen.

Next week I'll probably be looking at the research tree in a bit more detail to make sure all the required items for the research tree spawn in the UFOs, and blocking out the sizes and interiors of the larger ones. This should allow the game to be played right the way through.

I'll also be finishing up the level design, which neatly leads me onto the iterative testing. Initially I just wanted to do this for the levels, but the more we've thought about it, the more applications it will have. GJ has done some work implementing a server that records gameplay stats, which I'm very interested in.

The long game we're working towards is releasing a few ground combat scenarios for public consumption, say 4-5 in all. The purpose of this would be to get some really wide-ranging feedback on the game balance, all of which would be collected automatically. How many aliens were killed and how many soldiers lost in each mission on each difficulty setting would be automatically collected, as well as loads of other stuff we can use to get a picture of whether the missions are too hard or too easy or too long or if some weapons are not properly balanced etc.

The most valuable information, though, would be on the AI itself. It's a bit difficult relying just on you forum guys for testing because you're all X-Com (and usually Xenonauts) vets and therefore you'll be better than average at the game. Discovering if an inexperienced player can beat a mission on Normal (and if so, how many troops they generally lose) would be incredibly valuable for us. We're looking into this stuff at the moment.

Also, GJ has a few concerns that he might end up making an AI that is just too good at the game to beat, despite them not cheating. I'd be very interested in seeing if that is the case - though I think you underestimate gamers at your peril. It'd be good publicity if they are indeed that good though!

Link to comment
Share on other sites

The only complication with this is that you'll likely have to download the standalone if you want access to all the game files for modding purposes, as that won't be packed up. But that's a relatively niche thing, so it shouldn't affect too many people (installing mods is still easy even with the files packed up, don't worry).

If those are in a packed file, can't a modder just unpack it to access them ? (Anyway, having to download the standalone in this case is not a big issue IMO)

Also AI, UI, and map iterations : great progress =)

Link to comment
Share on other sites

Also, GJ has a few concerns that he might end up making an AI that is just too good at the game to beat, despite them not cheating. I'd be very interested in seeing if that is the case - though I think you underestimate gamers at your peril. It'd be good publicity if they are indeed that good though!
I laugh at this supposed "M-5" AI. If it doesn't cheat it can be beaten if by nothing else than pure luck. The aliens have the advantage of starting in cover and being on the defense, however, that does not guarantee victory even when opponent is a human. I say bring it on in full force and we'll see how good it really is. ;) Edited by StellarRat
Link to comment
Share on other sites

Can't wait to get my teeth into the new improved Alien AI. I'm hoping it will be able to bait us, flank us and know that it can force us to make a stupid move into an overwatched ally for example.

There is nothing worse than the thought of an invading 'superior', 'intelligent' life form having no grasp of basic ground combat tactics and just wandering around like lost sheep.

I want to be crapping my pants at the thought of stepping off that transport.

The tougher the better.

Link to comment
Share on other sites

Realtime AI can be good too much because of it superhuman reaction and

precision in game world, it can to see you, give individual oreders to hundred units, point his gun to geometrical center of your head and pull the triger - and all this before frame on your monitor refreshes, nerve impulse reach you hand and your mouse begin to move, and there no any cheating. If you not add to it enought handicaps it can be good too much without too much brain, and human there don't have too much chances. But in turnbased it didn't have this benefits and I didn't see how good too much can not leave us possibilities too win. Well, there some defensive tactics, easy and still fair - like concentrate behind single entrance looking at it and sit here ready to rection shots, not visible from outside, far enought to not be reachable by explosives from outside - and we forced to enter to win. It still fair(by game mechanic), it's clever choice for defending, but it's not very intresing and can be very hard to beat(maybe lot of smoke and this assault shilds somehow help) but at all I still think - better playing AI(I'm talk about AI whose goal is to win, not to play) is still tend to make game more intresting and it can't be so much superhuman to be unbeatable no mater how smart you can make it. At least as far as I can imagine.

Edited by zzz1010
Link to comment
Share on other sites

That's not a good defensive tactic for a game with destructible walls, grenades, rockets, and automatic weapons. It's a recipe for failure.

The only way you could reasonably statically defend a hardpoint in a game like this is if the defenders have a significant advantage in firepower and troops that counter every indirect fire scenario you can come up with.

And being "forced to enter" a room with indestructible walls where all the badguys are camping the entrance from outside of grenade range is not an example of good AI, it's an example of lazy scripting.

Link to comment
Share on other sites

That's not a good defensive tactic for a game with destructible walls, grenades, rockets, and automatic weapons. It's a recipe for failure.

Ufos walls is undestructible, and as i say - they in places not reachable from outside for explosives, you need to enter and not die(it's trigger reaction shot), to see them and not die, shoot and not die(every you shoot will trigger reaction shot) and kill some one before you die and try to escape and not die(because on next turn they kill you if you stay there). And for to concentrate I not mean stay in pack - every one can sit in different corner.

But if it can be easily beaten(It's just examle what came to mind, I'm not tactical genius, and there can be possibilites what I don't see, or in game it will be differenly then I expect) then it's just not issue, and even if Ai try to do this it still can be easily beaten.

And being "forced to enter" a room with indestructible walls where all the badguys are camping the entrance from outside of grenade range is not an example of good AI, it's an example of lazy scripting.

It's no matter how you call this, any ai can be called "scripting"(because it is, there can be complex chains of weights and conditions in these scripts but you still can call them like this), and this not cheating it's still in bounds of game mechanics. And "camping" can be right decision for defending, no matter how you like it. You can call defenders of sieged caslte unfair campers, but if they want to survive it can be right tactics.I say before - it's not very interesting, but if we talk about ai who try to win, not to play it still viable. Do you think we must to forbid AI to camp, only player can use reaction fire mechanic? Or need to forbid for ai searching for not seeing or not reachable by explosives places, or may be most of them couldn't stay in ufo even if this is most defendend place and if there will be no one of them, only player - they loose??

Edited by zzz1010
Link to comment
Share on other sites

Also, GJ has a few concerns that he might end up making an AI that is just too good at the game to beat, despite them not cheating. I'd be very interested in seeing if that is the case - though I think you underestimate gamers at your peril. It'd be good publicity if they are indeed that good though!

Not likely.

There used to be a popular notion that XCOM EU is unbeatable on Impossible. Now there are people beating it straight up, Ironman, with Second Wave on. Players are going to find weaknesses even in very good AI and learn to exploit them.

I think a better option is to allow players to change difficulty mid-game and fine-tune it, e.g. have a menu several separate difficulty settings - for AI, for RNG, for strategic game.

If there is a good basic AI, its settings can be balanced around more than just difficulty. For instance, AI on "Maximum" might be allowed to "cheat" in having extended vision. AI on "Very Hard" would not cheat, but totally min-max its behavior down to hiding in the closets if that's the optimal action. AI on "Hard" and below would avoid camping to keep the game more fun.

Link to comment
Share on other sites

With regards to the packing of the files for the Desura version (and the final version as well I suppose):

Unpacking to access the files will probably be just the first step, as the game might expect only packed files. Repacking mods might be needed as well, or will the earlier described method (placing the changed files in a separate directory that is checked before falling back to the original files) work with a mix of unpacked modded files, and backed originals?

How will this work after the final version is released, will an unpacked standalone version be available then?

What method is used for the packing, a regular archiving method like zip, rar or 7zip?

Link to comment
Share on other sites

I doubt that was specifically because of the AI. Afaik it was mostly due to lethality, alien hit points, low to-hit chances etc.

Of course. Not about AI. It's rather that... every time the devs say something can't be beaten, in a year it usually is, and in two everyone's doing it. No one designed Doom's Nightmare difficulty to be beatable using fist alone - they even put double ammo in, assuming normal quantities wouldn't suffice.

If Kasparov had a dozen more rematches with Deep Blue, frozen from interim modifications, how do you think the last ones would go?

(With Deep Blue specifically, not Houdini with full tablebase)

Link to comment
Share on other sites

Well, i can imagine you easier can produce a much tougher AI in turn based games than you would in real-time games. Since the AI can calculate on a fixed setting and not respond to human "irrational" decisions. Unbeatable though? I would like to see it to believe it.

But a very though AI sounds really fun. Hopefully this is a true claim.

Anyway. Nice to hear about the progress! Always fun.

Link to comment
Share on other sites

Then we must remove reaction fire from game, because it's only intended to use for camping.

That smacks of silly butthurt. Think it out.

If Kasparov had a dozen more rematches with Deep Blue, frozen from interim modifications, how do you think the last ones would go?

I think he (Kasparov) would have trounced Deep Blue. How many Grandmasters were involved in tweaking the AI for the rematch that defeated Kasparov? Like 5? He wasn't just playing against the computer in the second round, he was playing against them as well. He beat Deep Blue pretty badly before it was patched to play against Gary Kasparov. And the victory over Kasparov wasn't assured until the last game IIRC.

Edited by xcorps
Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...