Do the bots justice: make them competitive.
Morning / Evening, ladies / gentlemen.
I have no intention whatsoever to complain nor whine about anything at all. I am a pro subscriber, and I am more than happy with what Quake Live has become. Like many others, I am getting increasingly irritated with people who cannot seem to understand why a game that used to be free (and still is) now is economically compelled to offer new paying content to carry the project, not to mention those among them who firmly believe that developers are being paid to release a major update each and every time that a new thread is started.
I have a suggestion, though.
What I am wishing for, since every suggestion ultimately comes down to what our wishes are, is a (far) more competitive artificial intelligence. I am very much aware that there are many other bugs and glitches that need fixing, I do not even know whether improving bots has ever been or will ever be taken into consideration. What I know is that there is plenty of room for improvement, something that the SpiterBot Team has undeniably proven to be true.
Why more competitive bots, do you ask? Obvisouly, no artificial intelligence algorythm will ever match the human brain. I know, too, that the only way to get any better is to challenge increasingly smarter (hence better) players than oneself. Why, then?
- First of all, bots being nothing more than lengthy portions of code with no life outside Quake Live, they are available 24 hours a day and 7 days a week. Not all of us are lucky enough to have friends who can provide us with a real challenge whenever we are willing to play, no matter how early in the morning or late at night.
- Bots can be requisitioned for whatever gametype that we are feeling like playing at any given time, be it a Duel, a Free-for-All or a team-based match of any size.
- Bots can be challenged on any map of our liking. Not everybody is willing to play such small and simple maps as Arena Gate and Hearth (to name a few among the least popular), which were specifically meant to be introductory maps (with one truly important item only: the Red Armour) in Quake III Arena. What if us, less skilled players, plan on learning the basis of such advanced and complex techniques as item control on those maps precisely because they offer the most basic of challenges?
- Bots are ready whenever we are, which makes them ideal warm-up partners for as long as we feel it to be necessary. They are also the best way to learn a new map without boring our friend to death or wasting ten minutes of our playtime getting spawn-fragged silly.
- Bots do not ragequit. They do not insult us, either... well, they do, but with a solid dose of humour, and we can always turn their chat feature off. This makes them the fairest opponent that we are likely to ever encounter.
- Bots' skill can be set to whatever difficulty suits us the most. Furthermore, bots are always at the best that their difficulty setting allows them to be. The same cannot be said of a friend who has spent her / his whole afternoon playing exhaustingly competitive games before we log in.
Bots can be all of this, and probably much more, but not without the solid artificial intelligence that they are currently lacking. Now, if anyone is interested in what a truly competitive bot can do, please give a look at the complete description for SpiterBot 0.52 Full. This video should give you a good overview of its performances, too.
Thank you for your time, ladies / gentlemen.
Last edited by sjn; 12-04-2010 at 02:19 AM.
Reason: #spelling edit#
At this video..I'm impressed about 0:53 and 2:40 but 2:00 is very epic Well, this bot can give real challenge. It will sharpen aim, dodging and maybe something more.
I'd like to see more lets say, active bots. The current bots just kind of walk around like zombies. Try playing ctf with 9 bots on courtyard and watch them, its quite amusing.
I use bots quite regularly for warming up before i go playing online or testing new config settings ect. I'm sure lots of people use the bots on a regular basis so i think it would be a good idea to improve them a little bit. Xaero is ok but he relies to much on his railgun+ wallhacks and autofire :P
edit: I just watched the video, i would love to see this kind of bot in quakelive.
Last edited by dru1d; 12-03-2010 at 05:35 PM.
All they did with that bot was crank all its 'ai values' to 10.
Each bot has a config file, and you can make basically any bot stupid accurate with a few simple tweaks.
To make a bot more than just a 'auto aim and shoot' machine, you need to get at the core of the AI.. aaaand that's a bit of a shady past in Q3's history.
Even i'm a bit shaky on details as this is something that's not really written/talked about much. Back when q3a was about to come out, they hired a programmer to do the AI, as it was one of those 'last second choice' things. they hired (i think) 'mr elusive' for this task. IIRC, he had worked on several very impressive bots for Q1/QW and maybe even Q2. (in fact i think the q3a bots are basically a port from his q2 work.)
At the time, the q3a bots were top notch and basically the best you could hope for in such a game of its era.
One catch however, The programmer maintained the rights to his code. The IA code became known as the 'AI black box of mystery', it was not released in the source (at first, this may have changed?), and other such nonsense.
Long story short, as far as i can remember, nobody ever got to really see inside the q3a AI untill much later, and what they got may have been a jumbled mess or something.
If you want a crazy ass insane bot to play against, you can use the ql decypter and make your own pak file with your own bot ai file in it, just copy one and tweak the values in the text file. It's pretty simple.
I used to have a bot with my own name and insane values set on him. I would let him run around inside my Q3a server when i wasn't around , much hilarity was had when people start screaming i was using an aimbot..... only to then notice 'heeeeyy, waitaminut, how is his name on the scoreboard twice??' etc.
While we are on the topic of bots and such, why was 'Pi' Taken out of the game? there is a screen shot of her floating around in the game files, but no AI files or anything to do with her, er, it , whatever.
all they really need to do is add the ability for bots to strafe, circle jump, dodge, do tricks like b2r in dm6, rocketjump, use weapons better(not favorite weapons like xaero and rail), etc
I respectfully beg to disagree with you, even though there is some truth in what you say.
Originally Posted by KETNAR
The Spiterbot actually consists of 5 distinct bots:
- 2pizza // Shotgun master;
- bot.dante // Plasma Gun master;
- O4kaPuk // Lightning Gun master;
- bot.python // Railgun master;
- spite // total master, the very bot whose skills you might have witnessed in the video link that I gave at the end of my previous post.
True. As a matter of fact, this is precisely what I did: I managed to successfully port both bot.dante and bot.python from Quake III Arena to Quake Live. Sadly enough, though, neither bot was nearly as competitive as it used to be within the Spiterbot mod.
Originally Posted by KETNAR
As for the other 3 bots, Quake Live was unable to process their item and weapon weights files, which hierarchically dicate the individual value that the bot is set to assign to each and every item / weapon on any given map. Those are the very files responsible for the bots' somewhat stupidly obsessive compulsion to equip one weapon to the absolute exclusion of all others, among other flaws.
Quake Live could not load these files because the syntax that they were written with is fairly different from the one that was used to write the files of the successfully ported Spiterbots.
Why a different syntax, then?
Because the artificial intelligence code was greatly improved, too, hence the 3 .qvm (Quake Virtual Machine) files that are included within the Spiterbot mod.
I am no programmer whatsoever, but my guess is that these files are responsible for spite's unique ability (well, for a bot, that is) to perform strafe-jumps along with the infamous Bridge to Railgun trick-jump on Campgrounds, its nearly flawless timing of the Red Armour (probably as well as the MegaHealth and the Yellow Armour), its choice of weapon (its Gauntlet frag never fails to amaze me), you name it.
You will never get a bot to do such things by merely editing its text files, even though there already is some room for improvement by doing so.
Well, like i said, I'm a little shady on the details because the quake 3 AI is something of a mystery, i never got really deep into it for any of the mods i wrote back in the past.
Originally Posted by sjn
I know however it is possible to tweak and even expand the AI somewhat with the QVM, as you said, but its limited somewhat by the whole 'black box' problem.
As for porting bot data from Q3a to QL, it is entirely possible that the 'black box' has been updated for the new engine (or the expansions made via QVM in q3 are utterly needed and or missing and thus the other files can't be loaded verbatim), i have no idea,i'm simply regurgitating what details i DO recall. its been a while..
On the upside, there is a thread now telling people how to tweak their own QL bots. Should help people from this thread if you folks are still wanting better bots.
SpiterBot is not Artificial Intelligence it's basically bots with special abilities. In the video when the bot makes b2r it is clearly visible that it's just a quick hack to make the bot actually land the jump - it does not strafe. The aim of the bot is ridiculous and not even near anything acceptable. Look at Valve's bots for Team Fortress 2: they can make tactical decisions and their aim is humanlike. Q3/QL bots never been anything more than mindless walking zombies if they are going to improve them it means ALOT of work and AI coding is not the simplest thing on earth.
I agree. Bot move characteristics should simulate advanced player movement ie strafe jumping. They should also utilize the entire map. In space CTF the bots will never jump to the rail gun and rarely go to side platforms. I need practice guarding the platform so if they would exit tunnel at different angles would be nice.