play #tdmpickup instead!

Give the pillar guy a call!

Link to the match please.

If they played as bad as you say, their rating will have dropped significantly after the first match.

Veterans and cheaters with alt accounts can throw off the shuffling just as much as noobs do, but without information about the players, there is nothing any system can do. Since this can go either way, the best possible guess is to put them in the center of the distribution curve and let them work their way up or down.

If you have them start off-center, this will systematically skew the rating distribution which will undermine the whole probabilistic model that the rating system is based on.

Opt-in/out is completely pointless. You can't have any skill based shuffling when there are untracked players in the game.

you assume there's the same number of actual new player vs lamers abusing. Which is not true.

What I assume is that when you add players off-center, you skew the distribution and undermine the basis of the rating process.

The amount of points gained/lost in a match depends on probabilites/expectations of outcomes, which are based on assumptions about the rating distribution. If you mess with the distribution, you mess with the whole rating process.

As good as 1200 (or whatever number) starting points sound, I'm not smart enough to foresee the effect of such a change. The Glicko rating system is based on the assumption that new players are added with 1500, and any diversion is likely to cause more inaccuracies.

And I want to spare myself of the next thread with subject "Lost 100 Elols coz some new 1200 account rekt me".

If by mess you mean likely improve, i agree. But whatver, your whole modus operandi is ignore better ways because it would create smaller peoblems you dont like for subjective reasons. If lowering the starting glicko would just start a process of readjusting every rating with time (instead of getting rid of some useless inflation), it's just another knock at the utility of the current system.

More crucially if u kick a fucker who just spent 10 dollaz for the privilege of abusing the system is no big deal, if u kick a 800 glicko guy giving his first try, probably it will make that his last try. It also generally sucks to be a newcomer and be treated as average. You are setting people up to fail from the very beginning.

You never responded to my post where I pointed out the technical problems that prevent me from using a team win/loss based rating for CA.
QLstats isn't qlranks, I'm not (as smart as) szr/sirax, Glicko isn't TrueSkill, and wishes aren't solutions.
My offer still stands. Provide solutions and we can talk about it.

If you dont know how to make it better i think you should just stop altogether so maybe someone else will do elo instead of glicko. Thats my opinion

Is Predathor's system better than nothing? If so then it should continue.

As long as hthis system is in place nobody will feel the urge to make a new one. So i am of the opinion he should shut it down.

Yeah I don't think someone will shut down their website because if they do, the chance improves that somebody somewhere eventually makes something better. In any case you don't actually intend to help with this hypothetical new system, right?

Right? It would be oh so morally wrong if i didnt

There are currently two major issues in CA with the current system.

-The one that has the most effect is when two 1500 players playing their first game(s) end up on the same team. In a lot of cases their rating should be 800 or so. The shuffling should distribute players with less than 10 matches evenly between the teams when possible.

-The rating-updates don't seem to be instant so a player with 1500 rating will keep it for a few games on the same server. Not sure if there is a way around this. Also not sure how aggressive your initial placement is. Bumping or dropping players up to 250 points on the first match would probably help a lot. Then reduce the adjustments after that. Skilled players would jump to 1750 after first game and newbies would drop to 1250 while others would end up somewhere in between.

When QLstats has no rating for a player, it doesn't provide any value to the minqlx balance script. That script then uses a value of 1500 for new players.

When the match result gets posted to qlstats, then qlstats will independantly use 1500 as the initial value for the new player and calculates the new rating, which is an instant update. New players have a high RD value and can easily gain or lose hundrets of points.

I guess that the minqlx plugin caches player ratings for some time, so there might be a delay before it loads the updated values from qlstats. That's something that should be brought to the attention of the minqlx / balance.py author.

the scaling is completely arbitrary you nonce

^^ Yeah. What does it matter if the noob gets 1500 on a team with an 1800 average, or if he has 1200 on a team with 1500 average? Why not just have new players gain elo based on their performance against other players, but have their elo not begin to affect anyone else until they've played 50 games or so?
Besides...you're just as likely to gain elo from having a new player on the enemy team who doesn't perform up to his 1500 elo, as you are to lose it from having one on your team.
In fact...you're MORE likely to gain it, since theres one more free slot on the enemy team that could be filled by a new player/account. You fill one of the slots on your own team yourself.

I'm pretty sure the fact that the scaling is arbitrary does not mean that the whole curve of players depends on what value new players start with. Current values would mostly persist if you gave new accounts a value of 1000 ELO. If the following is the curve (I'm quite sure that's how the distribution is, predhator said 1500 is the mean):



then changing the default value as indicated won't make the whole thing shift to the left. Not sure though.

In any case, automatically giving new accounts the avg value does seem counterintuitive, they are more likely to be worse than the avg (as they are new).

i see the argument now. it was poorly phrased because he kept saying that 1200 was better even though it was pretty much the same thing on a different scale

predathor, i understand your concerns that people CAN subvert the elo shuffle. however, this requires malice on the part of the person, and there's not that many people who behave like dicks because, seriously, very few people who can play quake properly have fun railing nabs who can't strafejump.
but, EVERY nab who joins a game has this elo put on to him.

now, ppl who have been playing for months now have a pretty solid elo, it doesn't change that much between games - if you are a 1400 player, you'll pretty much be a 1400 player for life. if you are getting better, you get better SLOWLY.

players who have earned 1500 .. YEAH I KNOW that some of you have lost touch with reality, and you think elo 2200 is like, meh, nothing special, but to most casuals it's hax.

i have nearly 4 years of play, and i'm 1300 to 1400. i can play with these new players and wreck them, and they get 1500 ???

1200 was fine. it worked well because, if you lose and you go below, like 1180, we immediately know you are a new player and should be ranked about 900, also, the numbers kinda work and having a 1200 on team was not the end of the world.

but a 1500 means the other team gets that extra player who can carry, and all the sudden we're playing 2v4, instead of maybe 3v3 with team A having a 1200 who is a 900, and team B having a 1200 who really is a 1200.

players who actually average 1200 are not bad, they won't win a proper T4 game, but they can do their 300dmg in a 8v8.

[-]
You should really read up on stuff before having an opinion about it.

This is such extremely poor logic that it hurts my brain.

if you could kindly point out the incorrect assumptions i have made, instead of shitposting, that would be great.

I wouldn't know where to start... Basically you could just google elo-rating. Start at how it conceptually works and then realize that what you're saying is making as much sense as saying you'll put a square triangle hat in NicholAss.

I'm sorry, but when you say that someone with 60 ping has an advantage over someone who has 30 then you're one to talk about logic my friend.

At least I know why and how the differences in pings comes into play and how and why it affected the game to the point where it is today.

You can check that if you reread my former post and have a clue about the game and how the game styles 'evolved'.
Or you can sit around and place useless insults towards me.

I'm not insulting you, but you cannot sit there with a straight face and expect me or anyone with some sanity to believe that having a bigger ping is an advantage, it's objectively wrong.

it's situational. I've been thorough in my explanation before and hardly anyone understood that...

Meanwhile you're saying that "you're one to talk about logic" about my complaints on the netcode.
I take that as if you think that I'm not logical on the subject, which I find insulting.

Objectively speaking I am correct though, you're not using logic when you're sayin it's "situational based".

It's a fact that the more latency one has, the longer it takes for shots to register to the server, so objectively, the lower the ping, the better the hit registration.

How is there any benefits to having higher ping than the other guy?

I didn't mean it to come across that way as an insult, sorry.

Since is completely impossible to hit ppl with 20 ping with projectile, and he has 20 ping, he wants it so also with the unlagged weapons. I bet you have been railed by a highpimger, the server changes your position because the shot hit u a tenth of a second in the past and your coordination gets altered. This happens with lg knockback too with a lower degree. What i dont get is why doesn't he think that happens tothe guy with the higher ping, only on every shot.

You should rewrite that post into English....
While you're at it you might also consider thinking about that 'my proposed solution' would make it more possible for crap ping players to hit projectiles onto other people more easily (not as easy as the low ping guys but easier than it is atm).
But you probably won't realize that, because you're too upset that I have in general played with a low ping within Europe...

Is there any proof of what you're stating?

I've always found it easier playing a person who has a higher ping than my own, never the other way around. I don't want to look at things from a subjective view, other things could have contributed to what you're discussing.

Lower ping is generally way way better no doubt. Getting hit by unlagged shots messes you up a bit because the knockback is different. Basically the hit registers where you are on your opponent's screen rather then where you are.
Generally the same applies to your opponent recieved shots even if the weapons arent unllagged. And you will shoot faster and less randomly due to the lower input lag. It's not really an advantage.
The game is a bit different than it would be with the perfect lan conditions because of the lagged/unlagged weapons balance, so online heavy hitscan hitters are usually favored a bit more. It is still way better to be one of those while having a ping advantage no doubt.

I've always said stuff about +forward with lg, so it has always been situational.
The higher the pings go (of both players) the more powerful +forward lg becomes (for both players).
The problem here being that the guy with crap ping always plays games where +forward LG wins so he's used to it. The other problem is that whoever wants to play +forward with LG can fuck up his ping. And yet another problem is that the guy with good ping can't help that +forward LG suddenly becomes so much more powerful when his opponent is on a crap connection.

The reason as to why this is is explained in my previous (netcode) posts.

If you wanna see what I mean... Go play a ping 60 opponent while being on ping 60 yourself and +forward into fights with LG out....

Sure. It's funny tho that out of the top duelers the people with higher ping (ash and pavel) have exactly the opposite gameplay. Meanwhile the most unforgiving +f lg player is evil (who has the best setup possible and a perfect connection) and the other famous lg player never wanted to play with ping higher than 30. Maybe it's because even tho when u hit with higher ping you might throw off your opponent (if you dont get hit by a shot that throws you off the same) but it's a lot harder to hit?

ouch

Only reason there is to prefer low ping is the more responsive hit sound. (For which I also said that 'if ID was going to stay with unlagged they should put a cg_predictHitBeep into the game').

Evil pings (at least he used to ping) a stable 50 - 60 to almost all EU servers. He even pinged 30 to Moscow until I stopped playing.
Evils game plan was so build around LG when he came on the scene... It even eventually made Cypher whine about how Evil was aggressive, online, with LG while it made no sense at all to do so.
Now it is true that since Evil also went to LANs his gamestyle has gotten much more balanced than it was before hand, which makes him loose a lot of 125fps nowadays.

Ash doesn't actually have a high ping, Ash warps like crazy and as a result is a heavy rail player who switches to LG to finish people off. The rest of his game plan is to spam rockets in order to get to the rail. Pavel is a more balanced player, but even when he gets into trouble he'll do almost all aggression with the LG out.

Strenx just was an idiot with crazy aim... He shouldn't be using timenudge either. Basically he puts a handicap on himself... However for other people his lg does feels really really weird which might throw them off.

Turning off hitsound or crosshairhitcolor has pretty much no affect on accuracy once you get used to it.

depends on the person I guess...
My LG went from 30 ~ 35 to 40 ~ 45 when I turned hitsounds off.
But I started to fail estimating how much hp my opponent had, so I played all of my officials with hitbeeps on.

I guess my Q3 OSP style never fully went away ;)

I've always heard people say their lg improved with hitsound off but every time I've tried it I lose about 5 percent even after 15+ games of getting used to it =( Guess I'm just super dependent on sound.

it kinda goes the other way around imho.
If you're (by nature) dependent on sound then turning it off will improve your lg in QL (because the sounds are too delayed).
If you're aiming more visually then having it on feels natural and gives some sort of confirmation or something... I'm not really sure as I'm the sound type ;)

Dude. Players could start with 500, 1000, 1200, 5000... but the whole rating distribution would follow as well so the result would be the same. Don't try to compare QLRanks' ratings with those. 1200 in QLRanks isn't equivalent skillwise to 1200 in qlstats.

im not trying to compare it to anything.

and, why would you say that; there already is a population which keeps the scores consistent - pros are 2300, good player 1800, decent players 1300, nabs 900 and under.

at this stage in the life of QL we really, really need to assume that if a new player joins, they are more likely to be someone who is new, rather than some Q3A lan wizard who somehow forgot that quake existed - i.e. most people who get QL through steam are in the 900 to 700 ELO range. starting them from an arbitrary value of 1200 is better than starting them from an arbitrary value of 5000, when there is already said pop count which is keeping the scores as above.

1500 is not an arbitrary value.
It is the mean of the distribution curve on which the Glicko rating update formulas are based on.
They only work when the curve has a certain shape (similar to a normal distribution or Gauss curve) so that win expectations (=the probability that player A will have a higher Perf-value than B, given their ratings) can be calculated and then the amount of points a player gains or loses when he beats/get beaten by another player.

Skew the curve and the error in the calculations only becomes larger.

1500 is the starting rating because it is the mean and that's logical. But 1500 is the mean because it was decided from the start that it would be the mean. Do you happen to know why exactly it is 1500 and why the scale is 400/ln(10) ?
I think it's either because the author thought those would produce nice values or because he determined they were the optimal parameters. I found a document which might be able to answer but I'm hoping you can save me some (a lot of) time!

The chess Elo system had a value range of 0-3000, so Glickman used the same scale for his Glicko system
I think the 400/ln(10) and some of the other "magic" values in his formulas are an approximation function for a normal distribution with a mean of 1500 and a standard deviation of 350, so that the implementation doesn't need a lookup table.

I see. Really, unexplained magic numbers annoy me greatly.

theres a certain way reddit users speak and ur doing just that

Interesting. Perhaps I should create a reddit account eventually.

As long as values don't get into the negatives the choices for values don't really matter much. You'll end up with a gaussian (like) curve regardless of which values you take.
1200, 1500 and 2000 are fairly regular, but as long as the scaling is done sanely you can also use 1 or 9000 or any other value as mean.

What does matter however is how many people leave the game on both sides of the curve. In QL it's easy to see that most with elo > 1800 continue while a lot of people worse than 1500 are being insulted and raped beyond believe and thus leave.
Eventually the elo will even out though, but at a higher mean than the 1500 among the still playing players. (assuming that noobs quit and non-noobs continue)

at last a logical post :P

Fairly wrong... The distribution curve is created through the Glicko-rating algorithm (or however I should be calling it, it's related to elo which I do know a reasonable bit about ;p).
The mean therefor is arbitrary at the start (although I do think that Glicko is a specific implementation that always uses 1500.... I could be wrong... I never did read up too much on the difference between glicko and other ratings).

The mean isn't arbitrary because it's not just the starting value but also a part of the rating update formulas, which use an approximation of the distribution to calculate win probabilities and the resulting rating updtes.

You can change the mean, but you also have to change the formulas too. If you don't, the actual and the approximated distributions will differ a lot and the rating updates become increasingly erratic

That's true, but in the papers I read (about ratings in general) "same mean value" (mu) is used everywhere.
So, to speak in code, the global variable 'mean' can be used in all formula where it is needed.
If you were to change that mid-way through an already active rating... all hell would break lose :D
(not necessarily if you've split all the variables into local ones and only change one of those... but it's definitely not wise to do so)

How would it make the errors bigger? Not only i think it is arguable, since having newcomers play with a closer elo to their actual skill level just takes away some inflation, i also think there's already so much noise and error in the calculations because of the other major issues that this additional alleged new error would be close to nothing compared to eveeything else.

but there is your problem; you assume that a new player will be average to the current QL population.

This is mathematically fugly.

Didn't even realise there was elo until some person called HELL said I was apparantly hacking in warm-up before we even played.

And yes I'm calling you out because apparantly after some research you were a top quake player, you should act professionally instead of unprofessionally.

The player known as HELL (in Q4 and parts of Q3 dignitas HELL) hasn't played under that nickname for a really really long time. At least as far as I know.
Chances are you played against the same fag who abused a shitload of other old pro players nicks to hide his own cheats. The attitude also fits...

I think he has come back letely. Pld vs him too and looked like the real one.

really?
With his real nick? That doesn't sound like Hell...
But who knows it might be. I'm inactive so I don't follow it too much...

hell_uk http://qlstats.net/player/25451#overall

He recently played against Garpy... He'll know if it's the same guy or not.
Tbh 113 games played with 3 lost of which only 1 game took 10 mins which is lost by 1 frag...
I'm still leaning towards the BS cheater guy who abuses nicks of previous players. But then again... Hell was pretty good and it seems like all of his games are on his best map (Bloodrun) so it could actually be him.
If you wanna know for sure then ask Garpy :)

http://www.esreality.com/?a=post&id=1760489

Whatever, the rating system is absolute shit anyway.

the rating system is absolute shit anyway.

...and still we have quite a lot of people on the servers who fap to these mystical elo values and would look down on anyone who has an elo lower than them. Even though that doesnt mean jack shit.

Its quite amusing, actually. Guess if someone has no real life achievements, these numbers make them feel important, or some shit.

Problem is, its completely unreliable and breaks the matches. But hell, its a SNAFU either way. No elo = bullshitto. Elo = bulshitto. What to choose? Hm.

If ELO is taken in moderation, then it's a nice addition for fun and balanced matches.

Current system is well implemented, but it's utter shit since it only accounts for the players individual performance for team modes to determine the amount of rating given. Win and loss should be the primary statistic to base the rating distribution on, the players individual performance should be only a small portion.

I can say with absolute certainty that this whole rating system is complete garbage if it places a player higher than me, who completely ignores the team and plays purely for individual statistics. That's why I completely ignore this rating.

Funniest thing is, if in CTF you cap all 8 and still have a low dmg ratio you lose elo. Yep. Nobody cares that you got the flag, or got most assists, defs, etc. These things are apparently pointless. All that matters is dmg, as if CTF was a CA with flags.
I know that dmg plays its role in CTF as well, yet it is utterly ridiculous when capping, defending, assisting etc play no role in your rating. Somebody might say: "It's not true as real men have no problem dealing damage AND getting caps etc". Well I started to play threewave ctf in Q in 1997(grappling hook FTW). Then played CTF in other Quakes. And all I remember was that the point of the game was getting the cap. I'm probably dull or thick in the head, but I dont remember any bs about making high dmg = good at CTF. Whats more, attackers used to have even lower dmg if they wanted to do fastcaps and had to RJ, GJ or whatever to boost the movement.
Now dont get me started on people who have a fixed elo and ruin games. Or people who STARTED playing the game already having a huge elo. What for, only the creator of this system knows. Also, I dont get the server admins setting the elo manually for some people - based on what? Or maybe it's a "hey friend I like you so ill give you more points" kind of a thing? This is all so artificial it baffles me. Either it's a completely unmodifiable system with semi-objective ratings or it's a completely biased system based on the whims of server admins, weird variables (dmg instead of caps) and irrational default settings (new players getting higher elo etc)
In my extremely nooby opinion, something is not right with elo in CTF.

lol and I thought it can't get more retarded, looks like I was wrong.

Its the OMG server, they manually set the ratings and don't use the ELO that's done by Qlstats.net

do you think its a good idea?

visit qlstats.net and look at ctf's leaderboard, it is like a horror show of retarded mid ffa chase warpers

CTF rating and ranking doesn't work well.
But I'm not wasting any time on it b/c the only populated CTF servers in EU use their own rating and shuffling system, independant of qlstats.

What in the world is wrong with people who chase kills in FFA? There's so many other players to kill just chill the fuck out lmao

hahahahhaahaha try being a known player.

After a few hundred games, the value lets you roughly guesstimate the player's abilities, which is a good thing.

i'll concede that median values is too hard a concept to apply to the real world but what about this simple observation - the initial score is too high.

until last week we had initial score of 1200, what was wrong with that? it worked far better than initial 1500.

until last week we had initial score of 1200, what was wrong with that?

I'm pretty sure it was always 1500 for qlstats since this is what Glicko is presented with. Are you talking about the numbers displayed by some script for unranked players?

when the steam migration happened, the first elo score you got was 1600. after the change, about 2 weeks in, it went down to 1200. any server which polled qlstats had 1200 as initial score for new players.

i played hundreds of games with brand new players who all have exactly 1200; i didn't dream it - i got so used to this that when i see someone with 1150 i immediately think, "here is a guy who just joined quake, and has maybe lost his first 3 games".

then 10 days ago it jumped to 1500.

you're hallucinating.

Yeah I'm pretty sure it's been 1500 a lot longer than 10 days unless I'm very mistaken

I just got confirmation from the man himself. The starting rating has always been 1500 on qlstats. The different numbers you remember are either all from a deficient memory or values used for unrated players (no matches on qlstats returns nothing, not 1500) by the shuffle script you use. I believe it's the later.

I am personally in favor of scripts using a lower than average value for new players but I don't think it would be wise to contaminate the system with a non-mean starting rating.

predatHOR I usually don't agree with the criticism over ur elo implementation, but in this case, there is some truth on what the guys here say. The algorithm u use is designed for a common game with equal share of new and veteran players. QL is nothing like that. Imo, even qlranks 1200 was a pretty high starting number for a game like ql.

I am not saying to change it. Since, even like that it really helps getting decent games, track progress and organise elo specific tourneys. I am not playing ca though…

Some of the criticism is actually justified, and I'm well aware of many flaws of the system. I also agree that a brand new account is more likely to be sub-average than average.

Unfortunately many good suggestions, like changing the start rating, are not fully thought through and end up with undesired/unpredicted consequences (others are technically impossible/impractible or come from a specific point of view or ignore inconvenient constraints).
I'm not saying it's a bad idea to change the start rating, I'm saying "please think it through". Because guess who gets the blame and the work for fixing it again.

But since the majority seems to agree with the O.P., I'm willing to make this change. If it makes the people happy, then "mission accomplished". If the rating process gets (even more) broken on the way doesn't seem to be a concern.

I'm saying it's a bad idea though.

Democracy can be a bitch when ignorant people want to change something. Just like the dutch vote against the Approval Act of the Association Agreement between the European Union and Ukraine.

It doesn't matter if they initially are rated according to their true skill, because minobot's team average based shuffle is trash anyway.

On topic:
It is a wild guess in the middle of the spectrum and a lot of algorithms use this robust approach, see "bisection method" for example.
The reason you place new players at the average/middle skill value is to obtain a faster rating algorithm, that needs less matches to determine an accurate rating (it works faster than placing a new but good player at the lower end of the skill spectrum and waiting until he worked his way up).

The problem with QL is, that the average QL-player can strafejump, uses direct weapon binds and has years of experience, but still will have the same rating (the value of 1500, 1200, 1337 or whatever random number represents the average) than someone who plays the game for the very first time.
So yes, placing new players at the average value of the skill distribution indeed is stupid regarding matchmaking in QL.
For this reason I tried to convince sponge before the steam launch's high influx of new players, to place new players in something lower than tier 3.

Except those on Steam trying a new arena FPS have on average played tons of FPS's before trying out Quake Live and on average tend to be as good as or even better than someone who's been playing for years yet is still only rated at 1500.

i never ever played one of those people, while i play to a straight up newcomer once a day at least. Also sometimes you see the expert cs player, the guy moves like it's cs, has ridic low sensitivity and he's far below the 1500 rating as well.

When steam came, it was like mowing grass.

Same shit as the loloadouts update in standalone times.

Except those on Steam trying a new arena FPS have on average played tons of FPS's before trying out Quake Live and on average tend to be as good as or even better than someone who's been playing for years

wut.

Experienced FPS player > long time QL player who's still only holding 1500

The former unlikely to stay as low as 1500 for very long at all.

The "experienced FPS player" still cannot strafejump (or do any movement beyond WASD), doesn't know how any of the weapons work or that you're not supposed to switch between them with mousewheel, simply doesn't know what an item is and has no map knowledge whatsoever. It may be possible to have adequate performance in FFA like that, but not in other modes.

Your idea of experienced and mine are quite different.
Sure there were a ton of players that seemed like they were still on a console as they played when it launched on steam, but they're absolutely not the ones that I would ever refer to as 'experienced FPS players', there were also tons of players I met that came from TF2 or similar, had played other FPS games that DID have strafe jumping, and absolutely knew to bind weapons individually.

But anyway, the only really important thing here should be if Predath0r looks at the data and the averages show that 1500 is too high to start on now or not.

Sorry for the late response, but after my last posts I have seen in another thread, that by now they started to autoban players from the qlstats deactivated list, based on subjective cheating accusations, having cheated pre-steam/in other games and general offending behavior, like racist or excessive aliases.
The weather in the last days was simply too good to waste time on the elo wankers' 4th Reich in a dead fps.

I think such cases are rare, because like the others I can remember a ton of overrated new players through the years and especially during the steam launch, but I personally can't recall a single underrated new player from another game.
And even when those fictional entities that you describe, played a lot of other fps, these games would be less about homoerotic skillsets like jumping around in a pretty manner, +back spamming the map with an overpowered noobtube and farming items.

My idea before the steam launch was to trade some of the rating algorithm's speed in order to protect noobs by placing every new player at something lower than tier 3.
What happened instead was, new players landed in tier 3 in order to have a slightly faster rating algorithm for the mystic fictional case you described, while the newcomers by thousands got slaughtered and adviced to uninstall on tier 3 servers.
The overloaded server infrastructure, the broken loadout gameplay and the retarded matchmaking were the keys to QL's "successful" steam launch.
Back then, my idea was to protect newcomers, now it is just to stop old players from raging about the ruined matches delivered by a system, that ironically is supposed to help matchmaking.

And yet, my statement about the higher scaled problem stays unchallenged:
It doesn't matter if you would magically know the correct rating of a new player's true skill, because minobot's average based shuffling is utterly insane anyway.
That problem not only exists at the very first matches of new players, it continues to fuck up matches all over the spectrum.

But like my previous suggestion this is so far ahead of time, that there is nothing else to say than:
/Sitting back and waiting on you to realise that regarding matchmaking, an absence of this system would be a better system.

Except it was shown absolutely in the months that we didn't have any system, that the absence of the system was a total disaster for CA and CTF the main game types it's used for.
I don't know what happened with TDM but that's hugely niche and is almost exclusively the realm of #tdmpickup anyway.

Lolca: The most casual and therefore the last gamemode that needs a competitive rating concept for tournament results of virtual win probabilities.
Ctf: The system overrates damage and therefore sucks, like the author admitted in comment #103.
Tdm: Public tdm died years ago.

The months without a system rather showed that people started to use common sense in order to shuffle randomly until teams were balanced for a gg, without mixing up penis length with virtual win probability ratings.

That was quite a contrast to the latched shuffles based on a retarded balancing criterion, fueled by stats wanking chase noobs like you, who ready up instantly although the previous game with the same latched teams led to an eazy blowout win.
Let alone the cases where new unranked players were involved or old players came back from a longer period of inactivity during the start of the ratings.
I am curious what your weird stats-glorification is, for people like komap who systematically ruin matches in order to work their way up in the rankings.

Funny, how people complained about the bad shuffles in the omg server thread, the virtual damage based lolca performance ratings in the qlstats threads and now about the shuffle results in this thread; and still you drop trash like "it was shown absolutely".

Anyway, on friday the 13th I bought Doom4 and the only reason why I waste my time on the inferior +back rl spamfest for warping subhumans that QL has become, is because I am waiting until it gets dark outside.

Lack of a rating system was crushing higher level CA, players were leaving, there were lots of times no games were available, people rarely stayed on any server for more than one or two games, if one left it was filled with an even less experienced player and then game over.

Now there are still tons of servers that don't run minqlx, no one that doesn't like the rating systems has to play on a rated server, no-one that's auto kicked from a server is left with nowhere to go.

But most games appear to be happening on rated servers because .... people overall like the shuffles that they give, despite the unavoidable flaws.

"SmackGyver" doesn't exist on qlstats, I don't know what level of player you actually are, but calling CA the most casual and noncompetitive mode makes it sound like you're a 1700 glicko player at best yourself, 1900+ players that haven't intentionally joined a very skill diverse server for a laugh couldn't be any less casual about it unless they were playing duel for money at the time, in fact half the time you'd think there was a real gun pointed at their heads.

This player "komap" you mention also doesn't exist when searching on qlstats, but there will always be outliers and a few maybe intentionally gaming any system.

I hate the way t1hol plays for example, he's an incredibly strong player, but he basically sells his entire team for Elo half the time by running early or not showing up, but he's strong enough to make up for it... at least half the time anyway.


To reiterate an earlier comment, I have ZERO issue with the start glicko rating changing, if the actual stats show it's a sensible change.

he joins #ctfpickup at weekends as 'smackgyver' but plays as 'smackalodon', strong aim and knows what he is doing. i cant say about 1v1 and ca but in ctf imho he plays better then what i have seen from u in pubs. sry m8. komap got famous for abusing qlranks, you can find it at esr by search 'komap' and 'antiflag'.

Thanks, I just had no idea who he was, wasn't looking for one upmanship or anything it's just it changes the outlook on things if it's just some newbie with dunning-kruger syndome that's really just complaining as they're unhappy they're kicked from higher games.

1900+ [glicko] players [of clanarena] ... couldn't be any less casual about it unless they were playing duel for money at the time, in fact half the time you'd think there was a real gun pointed at their heads.

Seems to be serious business, in the gamemode that gets called "lolca". Thihihihihi.

So I called you a "stats wanking chase noob" in comment #108.
You reply by describing for 3 sections how lolca got obsessed and zombiefied by stats wanking, then you try to chase me on qlstats, after this you try to chase komap on qlstats, after that you ambivalently complain about how stats wanking ruins the game, and afterwards it turns out, that you even judge an argument based on stats instead of content.
Do you notice how I nailed it and how your replies unfortunately advocate against the system that enables stats wanking?

Calling me a chase noob was attacking me and not any of my points. I haven't attacked you.

Your arguments about non-rating based shuffles being better are opinions and in order to understand opinions it helps to find out about the experiences that caused those opinions to be formed.
If your arguments were facts/statistics then sure, it would be wrong to check into the background of the person presenting them and not just the validity of the data itself.

I did not slight you for your ability, or your ratings, I didn't know them, I hadn't found them, nor did I in any way state I was invalidating your points by doings so.
I *DID* suggest that the terms you were using for the game made it sound like you were not at the upper end of the ratings because those that are tend to be taking things more seriously, but I had no actual information about who you were / what level of player you were, and 'sounds like' is not definitive.

Apparently you're pretty good at CTF, I still have no info on your CA game as it seems you use a different alias for that entirely as there is not one single CA game tracked for your "smackalodon" alias.

I categorically did not say you're wrong because you're bad, I did not say you were bad and I did not say you were wrong, I just suggested your level would change your outlook.
I had no actual information on your level.

Not a lot of people I've seen call it 'lolca', I've only ever seen it a few times on ESR and a couple of times on servers where in the latter case they were clearly referring to teamplay-less CA games turned FFA with widely varying skill levels in the team and they'd given up on any idea of playing normally. The ratings were designed to help people avoid such messy games if they wished, and I believe most people that care for that find it works a lot better than when there was no system, if you're within about 150-200 gliko of the other people on the server then chances are you're at very roughly the same kind of level, that is all these systems were designed to try and achieve, not absolutes.

Bringing up disliking t1hol's play was an example of AGREEING with you, yet you use that to slight me more.

You use me trying to look up komap as a put down as well? what else am I supposed to do when you mention a player I've literally never heard of, and say they "systematically ruin matches in order to work their way up in the rankings", funnily enough I'm going to try and find them via the rankings you say they're abusing.

There's no point trying to converse with you at all if you can't interact with someone who's trying to understand your side of the argument better by attempting to look into the information you present, and even showing they're agreeing with you some. But are instead just looking for ways to attack them for doing precisely that.

This whole thing started as an argument about the entry rating of a new player, and I've twice stated changing that is fine by me so long as it seems sensible when looking at the average rating stats, it's just wrong to change things for a tiny percentage of users complaining no matter how loud if the actual statistics disagree.
But you ignore those comments too, and just look for things to pick on.

So, basically, bye.

I still have no info on your CA game as it seems you use a different alias for that entirely as there is not one single CA game tracked for your "smackalodon" alias.

Why so complicated? Have you never taken into consideration, that I didn't pay twice for a dead game and I also don't play lolca at all?

Section-wise replies:

[1]
No, it is a fact, I can send you the demos: You are a chasing noob, which makes your whine about chasing in comment #106 absolutely ridiculous.

[2]
No, that is a fact aswell: The average rating of 4...5 players/probes is an utterly retarded balancing criterion and latching the shuffles based on it, is even more retarded. Your "opinion" gets disproved by blowout games that were balanced to the same average rating of both teams.

[3] [4] [5]
The thing is, I am not as stupid as steffo to fall for your self-righteous justification nor does it matter who the better player is for my argument:
Chasing players on qlstats, admitting that lolca got zombiefied by qlstats, players victimising their mates to gain elo and then judging statements based on stats, this all does advocate against a system, that enables such stats wanking.

[6]
Lolca, is a standard expression of people that see this gamemode's complexity as inferior. On a server where everyone has about the same skill the rating/shuffling algorithms don't matter at all. The full madness unfolds when you have players with different skill levels of the whole spectrum on the same server.

[7] [8] [9]
Komap was just a prominent example for the negative sideeffects of stats, just like the guy you named. It is just ironic that you advocate for stats, but mention a contra-argument by yourself. In the same manner you won't be able to deny larger scale effects beyond specific examples, like tierslumming in 1v1 and people being ready although the teams obviously are unbalanced, just to farm stats. Everybody has seen people ragequit during countdowns, just to save their meaningless patched up ratings. I won't give you credit for looking into that perspective, because on the topic at hand the alternative would be to deny obvious facts. Also I think you are overrating your empathy, considering you assumed I am "some newbie with dunning-kruger syndome".

[10]
Actually this conversation started by me complaining about the retarded balancing criterion in the first sentence of comment #62.
I didn't ignore the comments about evaluating the data, I just thought further and realised that it would cause more trouble than help:
While back then at starting the system speeding up the algorithm by making everyone start at the middle/average made sense, it now makes no sense to place bloody newcomers at the same average rating at which veterans with years of experience have settled on. The only subset of data that would be relevant for such an analysis, is the sum of all smurf-account-free newcomers that appeared after the system had enough time to place all old players with enough sample matches to reduce their margins of error, good luck on extracting that data.

[11]
Yes, judging by the latency of my replies those virtual win probability ratings of a dead +back rl spamfest got irrelevant to me aswell, so bye.

I am not as stupid as steffo

I won't explain you the difference between opinions and facts, your nazi attitude speaks for itself.
You know what's stupid? You being butthurt for my plus to Yakumo.

See it how it is, I read his comment to the end and then laughed about the moron that is stupid enough to fall for such a self-righteous justification: steffo.
This made you so butthurt that you had to write a content-free whine reply, ironically calling me butthurt for laughing about your stupidity.

In the same ironic manner you support such a nazi stats system, but feel entitled to call others nazis. That is why it is better when I explain you the difference between opinions and facts:
Opinions are what you write, that is why you resort to vague claims about pills and attitude without anything to back them up.
Facts are what I write, that is why you don't even dare to contradict any of them.

Now excuse me, I got some stuff to do and you are just an obtuse waste of time.

Well, CA is indeed not Quake, but Yakumo's description is pretty spot on nonetheless. Because winning is everything (even in lolCA) and you have to carry 1-2-3 vegetables if you want to win. )

"It may be possible to have adequate performance in FFA like that"

Most definitely not.

DOUBLE DIZZLE!!!

Didn't read the whole thread so if it's been mentioned before /facepalm

Why don't we have a system whereby you start with 0 elo and after 5 matches awarded elo based on that?

When a match starts at least we know some new players with unknown skill might unbalance the teams and they can be suitably evened out and all have fun =)

Wht should be done is average the newcomers elo after a bunch of games and set that the starting value

neither knows what bridge is

Aaaaaah. Ok.

The legendary Bridge. A staple of every CA drama.

Why not force new Players to play 10 mandatory bot matches that will decide tour ELO rating.

Beacause then people would whine about THAT.

Wouldn't that just balance out after a couple games with bad people at 1200 and good people at 1800 ? If you consistently beat 1500 ELO ppl then you are bound to have more than that.

yes it balances out in some games (not sure how fast). The problem is the collective effect.

Example: out of 10 new players 7 are completely new (below 1200 real glicko) 2 are somewhat average (around 1500) and 1 is a decent player (1700+).
Let's say you set starting glicko at 1500: in 3 games everyone has his real rating. It's a total of 24 games gone badly (because two of them were already rated right).
If you start at 1000: it takes 4 games for the average players to get rated correctly and 8 for the veteran. Total of 16.

So with the system in play every single player is less damaging than the worst case of the other one, but the collective effect it is much worse overall.
And to be fair i used a very optimistic estimation of how a popoulation of 10 new player would be made of, by my experience is more like 20 are newbies and 1 is not.

Where are you getting these numbers from? Unless you show that your intuition actually matches the math this is pretty meaningless

It's not meaningless. It has meaning as an example as long as the distribution of new players is as onesided as it looks by my experience (and not only mine according to how many people mention this in game). It would have meaning if it took ONE game to rate people properly. I know that one's experience does not equal to reality necessarily.

BUT the motivation behind the starting elo, given by the guy with the numbers, has nothing to do with how the new player skill level is distributed. So of course i'm prone to think there's a fair amount of truth to that example and it is not meaningless.
There's also the fact that while it made sense for the initial months of steamquake to have the starting glicko centered, it is less and less (enourmously less) likely going forward that a new player is average. Quake is one of the hardest games out there even in its watered down versions like CA.

The best way to go in my opinion is to check what's the average elo of a few hundred new players after something like 5 games (depends on how fast the rating system works) with an algorithm and then use that as the starting elo.

All this doesn't even take into consideration the detrimental effect that has being rated badly for a new player vs for a veteran.
If you get kicked or harassed or you play an imbalanced game -in your favour btw- and you are a veteran, big deal, you'll still play quake.
If you get stomped and you get the blame because you are rated twice your value you are just much more likely to quit quake for good and wish you spend 10 dollars on something else. And as much as i try to never do so to newbyes and be helpfull, this stuff will happen and there's just so much you can do. And can you blame that much people who just want to have a decent game if they play a handfull of matches each week?

Whining about CA elo...#retarded

When people don't care about their glicko/elo/wtfever this system balances pretty nicely after 30-40 games played. It's all you fucking weirdos who actually give a shit about what their rating is that cause the problem.

It's too high, if the team average is higher it makes an even bigger impact, new players are realistically 1000 on a good day.
If they have no PC experience, 700 is generous.

IT IS ALL RETARDED

Thats why I play this game. everything is retarded there; players, updates, servers, games, drama, you name it.

The elo discussion will outlive us all, I'm afraid. Both sides have their rights and wrongs and neither option is good. There always will be casualties of any system.

So my advice, honest advice, is to skip elo and play. Seriously guys. Let's try to get any fun out of this game before we are all intermingled in Elo Wars and we forget what made Quake fun in the first place - fast paced, funny, sometimes completely insane gameplay with wicked people and lots of idiocy.

I mean, we are serious in our normal lives (at least we try), so why do we have to be serious in a fucking VIDEO GAME?

It all went south anyway since all of our stats were arbitratily deleted after move to Steam. that was very nifty by the way. I still admire Sync and Sponge for doing that.

Everything you post is gold, lolzy :)

From what i saw, players rated around 1500 specificaly in CA (not tested other game modes) are very much better than the average experienced FPS player that would try QL for the first time.

To help new players have a better first impression about QL you might want to lower it.

predathor, i want to thank you for being understanding. you are really helping the QL community and i really appreciate what you do.

You guys have a problem with everything, dont you?

Do you got a problem with us having problems?

>tfw you finally get to play quake with somewhat normal ping because of self hosted servers
>the only populated one is a ca server

just kill me now

ppl whining about 1200 starting ELO rating, where in fact algo is GLICKO.

Imagine TEH SURPRISE when they realise that glicko is dynamic, and it goes DOWN if you are inactive.

Predath0r, please don't pay any attention to this, you are doing a GREAT job, please continue to do so.

well, fuck me, i though he said he's bring it back to 1200. instead it's still at 1500.
and sorry to those who get hurt by my misusing the term ELO, when i should be using the term GLICKO, while it's perfectly fucking clear that we mean " the stat that comes up when you use the !ELO command ".

The way it was said, it really didn't sound like it was sure to be changed or like it was gonna be soon.

But since the majority seems to agree with the O.P., I'm willing to make this change. If it makes the people happy, then "mission accomplished". If the rating process gets (even more) broken on the way doesn't seem to be a concern.

By the way, it's impossible to bring it back to 1200 since it was never 1200 in the past! :p

AFAIK, it was 1600 the first days, then it went down to 1200, and now 1500.
Also, predathor said it updates right away, i think he means right after a game. so, if you !ELO during the second game, it should have updated, or maybe a short time in. Instead it does't do that at all, even the next day ppl still have the same rating.

I just got confirmation from the man himself. The starting rating has always been 1500 on qlstats. The different numbers you remember are either all from a deficient memory or values used for unrated players (no matches on qlstats returns nothing, not 1500) by the shuffle script you use. I believe it's the later.

As for the update thing, are you looking through the website or only from the potentially cached !elo command ingame?

im only looking at the !elo command ingame. As i understand it, that is the value which the shuffle uses. Ofc we are talking servers which have the appropriate plugins.

Yeah. You do realize that PredatH0r doesn't control shit regarding this plugin, right? No way for him to stop it from caching values for a long time. No way for him to choose which values it uses to shuffle a player which has no rating.

actually, i didn't. i assumed it polled the server once per game / map. i also do not know how long the caching period is. but, you are implying that, because the player is not listed on qlstats, then the plugin chooses 1500 as a starting value. is that true?
i mean, if it is, then the issue is with the plugin - and it's still an issue. i haven't accused predathor of anything, i just want this 1500 thing to get resolved, i.e. to set a starting value of (considering what happens in game) 1000. I have yet to see a "1500" player play better than that, most hover around 800 to 600.

because the player is not listed on qlstats, then the plugin chooses 1500 as a starting value. is that true?

Yes.

https://github.com/MinoMino/minqlx-plugins/bl...nce.py#L31
https://github.com/MinoMino/minqlx-plugins/bl...ce.py#L153

"I have yet to see a "1500" player play better than that"

plazma with his 1400ish. : D

fine then. the issue still remains, though. easy enough to get the various servers to edit this value, but hoping to see the plugin get updated too.

At first, how about asking new player several questions about his experience of Quake and other fps games? It could be somewhat helpful, couldn't it?