yeah, and that kind of tactic should not be possible (although its the only way i can get a good shaft %, that and him spawnin gin a corner somewhere)

LGC3

i have done away with weeks (thanks sfinx and qqshka for the idea), and have made the bot a little bit more "pure" so it should not shoot anymore rockets and the ping should be "stable" (it was always stable but the number was random)

also, you DMM4 bot fanatics, i have created something special just for you

it is called hardcore mode, and it will have its own tracking page

to play with hardcore mode, type "skillup" before loading a bot

Good work phil!

That hardcore-mode is a real challange, i would be very surprised if anyone will be able to win vs the bot on any map (maybe dmm4_1?).

Edit: just tried dmm4_1, and that map might be the toughest of the them all in hardcore... the damn bot held me in the air 24/7

so how do i turn the grunt off?

I think the grunt is the sound of a haste rune.

Yeah you are right, the coolest sound was the quad + strength imo.

heh

in the future once you lg gerbils will start getting 10,000k vs hardcore bot, i will have to make quad hardcore bot with both sounds

Omg, got lgc3.zip today to start practicing. The damn thing is hard as hell now! Can't use the old method anymore cos the bot don't shoot rockets. Hardcore mode is insane, lol! Need a couple days to get used to FPS again, but I need to finish Zelda OoT first (for the 10th time).

what happened to the demo stats? all i get is a division by zero error.

ok just saw it's under construction.

hej olo
i just went into a 30 minute blitzcode session and fixed that :>

also, you can now sort by "overshaft" and "undershaft" (just like you can sort by "pwnage"
also in map toplists menu, you have "TOTALSCORE" (how the rankings on qwnu/qwdrama rss feed get calculated) which also has average score and number of demos, i will make this more userfriendly
also there you have "GLOBAL" which shows you all the demos in one column for all maps, so you can sort all of those and see whats what (they are still separated by hardcore/softcore mode though)

yo phil the improvements you made really rock, but there now seems to be a problem with the activity timeline. for example if you take a look at dmm4_3 on softcore lethalwiz had the record for ca 14 days and then uploaded a new demo which erased all the time he was at the top and gave it to cyrax. the same thing on dm6dmm4 on hardcore happened to me, although i'm the only one who has a demo there so it doesn't matter, not that it's that important anyway.

yeah this is (sort of) by design

basically, lets say player A gets 9000
then X amount of time later, player B gets 9100
then Y amount of time later, player B gets 9200

scenario 1:
what currently happens on the site is, player A is awarded X+Y lead time, and player B is awarded only the time after he made his 9200, until now

scenario 2:
what you want to happen, is player A is awarded X amount of time for his 9000 lead, and player B is awarded Y amount of time for his old record, as well as the time after 9200.

scenario 2 is what i would call "total lead time". however, what happens now (partially because it is easier to code) is scenario 1, which would be more properly called "exclusive lead time", meaning it excludes the lead time for your own records that you beat (and, in fact, awards that lead time to the previous leader)

i'll probably recode it for scenario 2 later when i'm bored and then i will name it "lead time" and not "activity time"



in other news, the site is back on

Hi! I've just wanted to upload a few demos and I'm unable to do that. Website shows me a message with internal error.


Could you (site admin) take a look and tell me what's wrong?

Seems it's forgotten/abandoned project... noone cares about it...

I wouldn't say it is forgotten or abandoned. The site is still there and it is being maintained behind the scenes to keep it running :p

There's just apparently no interest in this kind of competition because swedes/finns get pwned by brazilians

I'm pretty sure some Brazilians cheat. I've seen some demos on chtv and they look so dodgy. One player looked like a very high senser with super high accuracy. Like 3cm/360 turning speed and 50cm/360 accuracy. It doesn't make sense. 70% + lg ? yeah right. Obvious timescale hacks.

Why is it so hard to believe? Bots haven't really evolved anywhere in that time and people have had like 5 years to perfect their game (at least). Also people sharing demos made it easier to catch up. I've never really got a great lgc score (didn't try that many times), but even I have had near perfect lg rounds for my skill level and clearly over 60% accuracy. And I'm nowhere near the shafter that many of these guys are. Since the way bots move is quite predictable, I really wouldn't be surprised if over time we could see a round that gets somewhere near 80%. At that point you are basically just shafting predictively for every spawns basic movement (assuming that you hit enough). Even now, that's mostly what it is, learning how to shaft specific spawns from specific spots, how bots move in certain parts of the map depending on where you push them and how you move etc. If you are going for a death star round, you pretty much have to control every spawn.

Especially if we are not talking about lgc rules, I think 80% is very doable, even without 'saving' shaft all that much. As stated before, bots have very predictable actions in certain situations and especially if they are using rl as well, it's possible to dominate the bot with scores like 30-0 in 3min pov. At times you might have to readjust and save some % and get a bit lower score, but still do amazing %.

The sens thing. Have you considered that professional pov-players might have two sens settings? It's fairly useful to be able to swing to the spawn immediately and then you could have a very low sens for shafting.

I lol'd.

But seriously, I agree with you. You can ignore my post above.

I can't see people getting a good score and 80% because to hit that kind of level you'd be able to waste very few cells (overshaft etc). To get 30-0 I'd imagine you need to use exploits like 'baiting' the bot into firing a rocket by moving out of LG range and then moving in with shaft while it is reloading. I mean seriously to limit the bot to ZERO frags in 3mins whilst making 30 frags yourself you'd need to do something highly unusual no matter how good your aim is, because the bot isn't going to shaft under ~35% based on what we've seen.

Anyway the point I'm making is that yes, you can exploit the bots weaknesses, but the good thing about LGC ruleset is that the effectiveness of those techniques is relatively limited. I'd bet my house that a 30-0 game wouldn't be a great LGC score.

In terms of LGC, I agree. Inter also showed that you can perfect your overshafting game, maybe some Brazilians have gotten even better. 30-0 was based on the fact that I've done 26-1 vs an RL+LG bot using LG only and having _decent_ spawns (and once again, I'm no Inter . Sure you have to do a bit of baiting etc for a domination score and it won't work for LGC. If qw was professional all the way, I think we'd see 80% vs a bot, even with LGC rules. At that point it's not really aiming anymore in the traditional sense though, you are simply repeating certain patterns from memory based on situations for most of the time.

I was trying to make points for LG percentages in pov vs bots in a more general sense, to take into account many of the things that matter. And by doing that, try to alleviate some of the concerns people have with these scores that would have sounded outlandish some years ago.

Phil, there seems to be an issue with the uploader (just tried to upload a dmm4_3 demo):

reticulating splines...

Warning: Division by zero in /home/www/lgc.quakeworld.nu/main3.php on line 494

Warning: Division by zero in /home/www/lgc.quakeworld.nu/main3.php on line 495
currently LGC3 supports these maps: povdmm4, dmm4_3, dmm4_1, amphi, amphi2, dm2dmm4, dm6dmm4, TOTALSCORE
your demo was recorded on map:
get in touch with phil and maybe that map can also be included in LGC3!

Timescale hax are ez to do in nq demos, but I donno how to do it in qw... Also high sens + good lg can be done by creating a script which changes your sens when you start shafting. I have such a setup. (I don't actually recommend it, it is really silly; but I am too used to it...)

I have asked zalon to move lgc to his box (where the rest of quakeworld.nu resides)... As it is now, it is still hosted on some dinky webhotel and I don't know what is going on with them...

(Those errors make me think that they have disabled execution of binaries, which is what the site does it calls a little qwd processor that I wrote to pull out the necessary numbers)

I'm also thinking of doing a different LG competition/ranking.

Instead of using bots, the setup would be something like this:

You play a bunch of pov games vs player x. You upload them to the site. You play some more vs player y, upload all those also, and so basically everyone uploads all their pov demos.

The site then ranks it by taking each distinct pair of players, getting the best result for player x and the best result for player y, and that is one data point each for each player. Then each player can be ranked in relation to every other player (for example using an elo type system or something else)

(The key is to disallow inflation by just one player flooding the site with more pov demos than the rest, and by just choosing one (best) for each pair of players I think this is avoided)

For this I would need a robust mvd parser (which I think we have now in qw-dev) and some free time to slap this together

Is there a mvd parser in php?

The parser doesn't have to be in PHP, it can be in C (like mine is) and then you call it from PHP and get all the values you need into your PHP script.

PHP is mostly binary safe so you could probalby do it, but there's no reason to do that

Aye, but my web host doesn't allow for it. I have a workaround in place though.

EDIT: nevermind