no gametype ranking? :(

oh nvm found it. GW! website could do a bit of work tho.

if you disallow to tdmpickp use your rating so public tdm became alive

how will disalowing tdmpuckup use rating help boost public tdm?

people will elowhoring on public and not on tdmpickup.

now you want to cheat in pub?

we dont use it :)

lulz skulp is 6th in south america!

Something changed in elo also.

I still notice that for the FFA games you are also including the stats from Instagib matches. The most active server is a Russian Instagib one and the third most active is a US instagib server. Scoring these players with Glicko/Elo must throw off the stats a bit when mixed with the standard classic mode FFA player data. It's a different game type really and maybe should be treat as such, with it's own rankings, as it's a popular mode.

I noticed that a few days ago that some instagib/fun servers are getting rated. That's because they use standard game type factories and change individual settings instead of using a custom factory with the modified settings.
I can filter out instagib, but other "fun stuff" is unfortunately undetectable. Only very few settings are part of the ZMQ stats or server browsing data.

"Most active" is a different story. These matches should not be rated, but nevertheless stats are stored and when people play a lot on these servers, they will continue to be on the "most active" list.

And no, I'm not adding any more ratings.
There are already 36 and there's not going to be 1v1 CA, 2v2 CA, 8v8 CA, classic instafreeze, vampiric infected pql loadout FFA, ...

It's mainly just those insta matches with the extra 'pql' settings that are not filtered, not a massive deal really. Thanks for your efforts on the website and making time to reply. :)

Hi Predath0r! Can you make player search by GLicko range
Ex all players from 1000 to 1200 rating from Europe region?

Why not just open the ranking list and go the page with ratings 1000-1200 ?

Main idea is to find opponents for adding them in friendlist.

Can you realize smthing powerfull like that?
http://www.esreality.com/post/2773424/re-quak...pid2773424

does elo work for 2v2/1v1 CA like before ? Because i had ~3000+ elo before the 'update' and now i'm stuck at 1819 and i played like 20-30 games since then but the elo doesn't change.

:(

Team games require a minimum of 4 players.
The primary purpose of the rating is for matchmaking / team balancing, and throwing 2v2 CA in the same pot would make the data useless.

it wouldn't, actually.

imagine heartless playing 2v2 for half a year and then deciding to join some 4v4 - his rating won't reflect his skill, therefore the shuffler will mess up the teams.

it's hard to imagine :D

so there is ranking for 4v4 CA and duel but not 2v2/1v1 CA ? :D

blasphemy! :)

are there any plan to make it work for 2v2/1v1 CA but not the same as 4v4, i mean, if you play 2v2/1v1 you gain elo points which don't translate to 4v4, and vice versa.

I think what predator says makes perfect sense. You never ELO shuffle a 2v2 CA game, so ELO for 2v2 matchmaking is useless. Also the ELO he is implementing isn't something to 'work towards' or 'gain' it is just an important aspect of team balancing in 4v4+.

So yeah if he puts it in 2v2, not only is it redundant and useless it has the opportunity to fuck up 4v4 match making.

"You never ELO shuffle a 2v2 CA game"

reason?

Usually most players know each others levels in a 2v2 ca. If it's one sided you just switch teams around after. If there's a player you don't know, after 10 seconds of speccing them you know their level.

You never ELO shuffle a 2v2 CA game, so ELO for 2v2 matchmaking is useless.

actually you do. I used a lot the !teams and !elos to make the 2v2 teams more balanced. And it worked quite nicely.

Also the ELO he is implementing isn't something to 'work towards' or 'gain' it is just an important aspect of team balancing in 4v4+.

So why is there elo for duel then ? Why elo for duel but not for 1v1 CA ?

So yeah if he puts it in 2v2, not only is it redundant and useless it has the opportunity to fuck up 4v4 match making.

I didn't ask to make the elo the same for 2v2/1v1 and 4v4. I said it would be cool to have something sepparated, just like elo for duel is different than elo for 4v4 CA.

You have elo for duel. You have elo for 4v4 CA. So why not 2v2/1v1 CA ?
Make it so that if you win like 10 games in a row at 2v2 CA, it doesn't affect your 4v4 CA elo, but only your 2v2/1v1 CA elo :)

I think 1v1 CA would require it's own ELO completely separate from 4v4 etc CA. It's the same as Duel and FFA, they have to be separate for the same reasons. Oh just re-read your comment and that's what you want.

My only guess is this: the amount of people who actually play 1v1 CA is so minimal it's not worth the effort. And even if it existed, it would only be for bragging rights. In duel it is a very helpful indicator, but in CA 1v1 who cares? It's the most casual of engagements, it doesn't matter if you destroy or be destroyed. Also 1v1 CA isn't classed as a 'professional' mode of QL, there are no tournaments etc. So there is no need for a professional ranking.

Oh btw don't get me wrong I love 1v1 CA - it used to be my thing! I think I'd give you a run for your money if I was still in europe.

note: But I do think that 1v1 CA is far more deserving of its own ELO over FFA, for instance. Why on earth qlranks ever decided to rank the scrubbiest / noobiest / shitest mode of quake I will never understand.

If 1v1 CA is small enough you would have a good idea of everyone's relative skill. Also in 1v1 CA you can tell someone's skill a lot faster and more accurately than in duel - basically as soon as you get in an LG fight with them :p

Qlranks didn't used to rank anything lower than 4v4 even anyway, or was it 3v3, I forget... either way, 1v1 and 2v2 certainly weren't covered.

Why not 3 like it used to be? 3vs3 is played a lot in CA, some consider it the most balanced amount of players for the gametype. Even Rapha said it himself in the year of the Qkcon CA cup that 3vs3 for CA is much better than 4vs4.

Layout and everything else, even colors should be the same wich qlranks.com had.

Qlstats could have work with a copy of qlranks.com, even domain should be the same.

This is the stupidest thing I've ever read.

This, the current design is hideous, but I suppose it's not your priority right now.

There's probably 100+ people on ESR who could knock together a nice and clean theme for him within a few minutes.

And I thought that I am delusional... but there's always a bigger fish :)
I asked several times. Nobody wants to touch the python/pyramid/mako stuff that the site is made with.
If anyone is willing to rewrite the UI, I'm all in for it.
But it's more like 5 weeks than 5 minutes.

It actually looks fine. If it got some lame hispter pastel colors neckbeard valley starbucks macbookpro design that would be a downgrade in my books.

Nice, but will be great if there is a light/white theme to choose. Too much black now. :D

Yes let's make the entire internet look like facebookgoogletwitter.

There is a reason behind that, don't you think so? ;)

I tried a light theme, but that doesn't work well with colored nicknames and neither with the colors in the weapon charts.
Tried a medium gray too, but that is even worse. Red, green or blue text in nicknames just gets destroyed by sub pixel font smoothening.

Hell yes I can fap to my stats page again, I was getting pretty bored of the screenshots of my qlranks page I took before it went down!

See, and screw the doctors who said you'll never have an e-rekt-ion ever again!

noticed many changes the last few days and it started shaping pretty good (fap worthy)

great work, keep it going

still some bugs though

like not all values are updated on players page.

new games are added, but the stats are not updated

The matches not added to statistics probably don't qualify for some reason. Maybe they didn't have enough players or weren't using one of the factories PredatHor follows.

lol, obviously I don't mean those. There are new games added that say +x amount of elo, but the stats don't count them.

I am sure he knows what I mean and will fix it when he finds the time. That's why I bothered posting.

Hahahaha, i thought i was the only one

I was lightly taking the piss actually :( Also masturbating over your own statistics is highly abnormal and suggests a serious personality disorder.

Ummmm..... so was I. *ninjaface*

qlstats has added a new incentive to play quake as your glicko decays with time. See, stats have positive effects on quake = D

I haven't played in almost 3 months now, I'm curious to play just to see how high I can get my CA glicko but at the same time I can't be bothered... I'm too lazy

follow your dreams!

Does the rating of the other people on the server get taken into account? All other things being equal, would you gain more Glickos for the same relative performance, if you were on a server full of high rated players than if the sever were full of noobs/smurf accounts?

Once again,
Thanks PredatH0r,
You rock!

First of all, thank you for the amazing work.
Some questions and tips:
- I see the site shows the stats and glicko score as soon as the match has finished but It doesn't update the rating. Why? When is it going to update the rating?
- When you search a player the site gives you back all the players that have what you write in the nick ordered by number of "registration" in the site. The feature is ok but I suggest to change the order: put first the player(s) that have the exact nick that you write and then all the others. Why? Because for example the player "id" join my duel server; I want to check his stats so I write his nick in your site but I realize immediately that I must check 79 pages of players to find him...
- The graph of weapons accuracy is pretty messy: too many lines! Can you provide a feature that shows only some selected lines? For example I want to see my lg acc so I select it and see in a clear manner the trend.
- If you can check the nationality of a player by means of the flag I think the nation ranking is a cool feature to add.
- Another cool feature of qlranks is the possibility to find the clans and check all the players of the clan togheter. I know that with steam we don't have a clan tag so you can't check it with bot but maybe you can add some sort of registration for a clan.
I don't really know, something like clan leaders contact you, give you some sort of garantee that they will add a real clan (for example with a link to houseofquake site for euro clan) and then they can make a clan in the site and add players in it.

Clearly these are only some ideas so take them for what they're worth.

"- weapons can be toggled on/off in the charts"

Just click on chart legend to turn off all unnecessary weapons

cool

it just needs a hand pointer, so that we know they are clickable

The Ranking isn't updated automatically yet.
I have to execute a sql script which is somewhat expensive to compute. At some point in the near future I'll automate it to run at a specific time of the day, so ranking are updated once a day.

Please keep in mind that the rating process is a statistical system which operates on probabilities and that only works over larger data sets and not single events (match results). In chess they have rating periods, which average out the individual ups and downs and take away a lot of the volatility.
People want to see rating results immediately, and the price to pay for it is the volatility in the ratings.

An alternative to the player search is the server search. In the "Now Playing" box you will see the top 10 (or 5 per team) rated players who are currently on the server.

I don't know the nationality of players. I run a geo-ip check on the server IPs, which is only good enough to assign players to continents.

I'm not planning on adding support for clans. My motivation for the site is to provide rating data for balanced shuffling and some stats for player amusement. I just don't have the time to add more stuff and maintain it over time.

There's a dedicated QLStats forum now hosted by plusforward.net.

I wrote up a rather lengthy Rating Q&A article, which explains some of my decisions and certain properties of the rating process and values.

I hope everyone checks this out if for no other reason than you don't suffer for trying to support a bad team :)

Can you pls put back and count instagib ffa again? Or atleast do a separate gametype leaderboard for that. There are tons of players on this kind of servers and we want to get counted too:) Thx

Works again now \o/

These stats are very wrong.I cant possibly drop over 100 ranking in 1 day playing a few hours?

Why couldn't you?

I sucked yeah but not that hard.Something is flawed

Something like how you play? :p

Seriously, how do you judge if it is flawed or not? Do you have actual data, maths behind it or just gut feelings we could discard immediately?

Please read the Rating Q&A.

If someone with lower rating has a better game performance than you, your rating drops. If you outperform someone with a higher rating, yours increases.
The amount of points gained/lost depends on the rating difference and the confidence in those ratings.

What many people don't understand is that this is not an open-end XP system. If you are already rated rather high, you will often gain a small amount and then ever so often lose a big amount and your rating is back where it was before - your estimated skill.

In chess they don't immediately update ratings after each match. Instead they have longer rating periods, which average out the ups and downs.

Putting a cap on the gains/losses would break the system. If you have an expected 99% probability of winning over your opponent, you would win 99x 1 point and then lose 1x 100. If you cap the 100 loss, you skill would be overrated.

Ofc the system isn't perfect and takes some time to adjust. And even then it will still have significant flaws. New, underrated players can put a dent in your rating, but after a couple matches, it recovers.

In chess they don't immediately update ratings after each match. Instead they have longer rating periods, which average out the ups and downs.

but when? instead of checking every now and then and see weird numbers that don't match, would be nice to know when the ratings are freshly recalculated. It could be every night, so we could check them the following morning or even once per week, like Sunday night. It doesn't matter when, but has to be a specific known time.

Another weird thing I saw is the number of duelers in Europe, which even though was already very small (about 990), yesterday became even smaller 886. Why is that so? I am sure there is an explanation, but I would expect to increase over time and not drop.

btw very sad to see so few duelers. After id dropped quake and the old qlstats became obsolete lots of ppl lost their enthusiasm over the game. That is also why, what you do is very important. It gives blood to ql.

imo having stats is very important for duel. Better matched and better quality games. You have a sense of how you performed. It is another thing to win a noob 30-0 and another to win a good player even by 1 frag. With some kind of stats, you can know that. Also, duelers are competitive persons and therefore knowing how they progress in skill day by day makes even more sense for them.

I know u probably doing this mostly for the team games, but have no doubt that duel benefits a lot.

The ratings are updated after each match, the ranking (players ordered by their rating) is updated daily at 12:00 CET since yesterday.

I'm not a chess player and don't really know how they handle things. I take what I know from the papers about the Elo and Glicko systems. Glicko 1 supposedly works best with a "moderate count of matches around 5-10 per player". But duel players can be idle for a week and then play 10 on one day, so picking a rating period is quite arbitrary.

Eventually it could be done to recalculate ratings over a time period, but the results cannot be broken down to individual matches. And when the numbers in the matches don't sum up with the rating, it will cause even more confusion.

I think that 0 glicko gained should be shown in green instead of red.

Obviously you are penalized by jumping into a game and therefore end up behind?
I have dropped over 150 in rating...this is ridiculous

feel free to ask ID to fix their bug which would circumvent that but good luck with that ;)

Player performance within a match is normalized regarding to partial play time, either by number of rounds when available, or by time.

If you provide your Steam nickname or steam id, I can take a look.
The qlstats.net forum would also be a better place for that.

Does this mean that if I join a match and play, lets say 3 rounds, dealing more damage/round than the rest of the teammates (including a higher rated player) which all had been playing since round 1, I get and increase in ELO, and this highrated player loses ELO?

It has been changed a few weeks ago that you are no longer compared to your team mates.
Regardless of that, you need at least 50% match participation (rounds in CA,FT, time in TDM,CTF).
Ideally only 100% should be counted, but that could be abused by quitting to avoid a bad rating and also would prevent that a player who had to move to spec to balance teams from getting a rating.

And as I mented a few times already, quitting and moving to spec are not distinguishable from the data sent by QL. Both look like a quit.

Quick question. Does this go by in game name or by steam id?

steam id. you can check aliases next to the nick on site. just click on "aliases"

Just a humble suggestion,
It would be nice if you could add the players weapon averages under the weapon graph headings or somewhere, so we don't have to hover over the graphs when you just really want to see them all in a quick glance. Thank-you!

It would be nice to have a skill graph. I would often use it to see if i was improving.

Would be cool to have some sort of log in to the site, where you can track your stats.

For example if you're trying to improve with a certain weapon in overall/long term date range, the ability to then reset those stats and try again would be pretty cool.

So far the system in place looks pretty awesome, UI could be improved slightly but it's not going to be that easy considering the complexity of the code.

Good job gents!

Please add the average net damage in the general stats!

There's a feature request thread in the qlstats.net forum.

Atm I don't have much time for new features, so it may take some time before they are added.

I doubt that a skill graph would be very useful. All you see there is a lot of zig-zags around your estimated rating, which on average doesn't change a lot. It's not an XP system where you constantly gain points.

Before I can add stuff like average net damage, which is expensive to compute (high disk I/O and CPU load), I have to get some of the already existing slowness under control.

what's up with still loosing points for duel, when the opponent leads and then just quits if he's afraid of a comeback?

i'll take a look. There's no special "quitter" code yet.

Whats up with qlstats? since 3th March it stopped tracking matches.

Yeah something is off.

he went for cigarrettes

nice work.

by the way, euro FFA'ers, who is Cook? is it Bunnye? beastly player.

also, top 11 TDM in NA....

Yes, it's Bunnye.

Some kind of a wishlist for qlstats.net :
1. Would be great if you could add leaderboard for B-Rating in FFA aswell. Mb instagibbers are on another boat from nw players but it doesnt mean they dont want to compete in stats too :P
2. In detail page for specific server it would be nice to have smth like its top players. Much more representative and interesting table then just most active players sorted by n of hours played
3. Dunno if there is some tech limitation behind this but its much better to have player stats like acc,win % etc to be based on all matches played like it was done in qlranks not just 100. Yet it can give only match details for last 100 of em while counting all of em in overall stats of a player.
Thanks in advance.

I don't know the particular flaw behind this but I play duels against only one player who has 1800 rating and the fact that the system ranks me at no. 7 shows something's not right.

The flaw has to do with clustering.

Glicko has no way of knowing an "objective" true skill of a player.
It's a form of statistical computer learning and all it can ever know about your skill is looking at your results and your opponents.
When you provide such limited and biased input data as playing only against 1 opponent, then your rating reflects only how good you are in comparison to that player ... and not against anyone else.

That is why I discourage people from comparing the rating values ... and why I don't like the ranking lists.
For meaningful rankings it would be required that everyone plays on random servers against random people all the time. And that's exactly why you can't join particular servers in games that enforce matchmaking.

Servers with skill restrictions are another source that reduces the quality of ratings. A relatively good player who plays on servers with a high min-skill, but loses there often, will have a lower rating than a worse player who beats noobs on open servers.

The ratings are definitly flawed. The Glicko algorithm can produce some anomalies, but the biggest "flaw" is that the data fed into the system is biased by how players select their servers and opponents.

Thanks for the clarification.

PredatH0r: if I wanted to collect some match jsons for data analysis, where would I go now?

I'm trying to select an "active" server from servers page and build URL like this
http://212.224.101.146/api/jsons/2015-11-24
or
http://74.91.121.184/api/nowplaying

Both coming up with tcp error, tried with port 8081.

Is central server an option at all?

I'd need to know a bit more about what data you exactly need.
There is an API to pull raw match result JSONs (without any rating information) for a given server, but no API to get matches for a particular player.
Try contacting me on IRC then we can discuss the details.

Granted, you can't really tell who is playing well/smartly - and who isn't - from stats alone. This is particularly true in CTF. Still, I think too much importance is given to damage in that gametype.

You can get + ~30 points for a defeat as long as you make ~25% more damage than any other player, and you can get - ~5 points for a victory despite having the most captures (and points) overall, and the most damage dealt in your team, just because in the opposite team one or two players (more or less 'useful') are spending most of their time hitting easy rails at mid.

Something's not right there and that's why there are several cheaters in Europe's CTF leaderboard. And I'm only talking about the top 10.

I don't even know if that's fixable tbh and it's not a big deal but it had to be said nonetheless.

Anyway, at least these rankings makes more sense than QLRanks' ones, so thanks for your efforts Pred'

In general for team modes it is better to base the rating and ranking on winning or losing (with weight of opponent strength) than it is to base it on personal stats or oersonal performance. Not only does the personal stats way of doing things open up for exploits like you mention but it also does not give credit to players "doing what is needed for the team to win".

One cant really blame a player as such for having a different objective (gaining rating rather than winning the game) so you will have contradicting goals and incentives. If the only way to gain rating is by winning the game then it is not really possible to exploit it in a way that is detrimental to what "should be" the general objective of the game.

It would be impossible to count how many times I was MVP+MVDP (and sometimes MVOP on top of that) in a game where my team lost huge (so to make this clear, we lose a game like 8:1, and I'm still doing the most "CTF" stuff on the server against a much stronger team)...yet this systems punishes me for it so hard, it's even worse than the other "old" system was (with the trueskill/elo thingy)... and that's something I would have never imagined to be possible . Damage is nothink in CTF tbh.

If you are a defender, you can only deal damage if they come to you, and attacks are well planned and sometimes slow-push/buildup like, and if you are a carrier, you are doing it right if you don't fight but cap.... yet it seems to me that the systems rewards damage the most.

And I didn't even mention the scenarios when one swaps to the loser team a few times because of the quitters.. that will show big time too.

I love what PredatH0r is doing, and I couldn't care less about this whole eyo thing, if it wouldn't be used to actually balance games. And it's still much better than nothing ofc, so don't get me wrong please:)

Personal performance in Quake CTF is virtually impossible to track, unless you coded some special system to monitor player decision-making in certain situations, even then it'd be far from accurate. I've already said all of this back when QLRanks first surfaced... :)
That the only way to make sensible ratings is if you're able to register teams of 4 or 5 of FIXED players, and then ranking is based on weighted win/loss. Outside of that application, I find it completely pointless. So I simply ignore it, just like QLRanks.
The website(QLStats) is quite useful for many things and I appreciate the work that's gone into it. I play some FFA and ratings/rankings there seem ok I guess.

For current public CTF servers, the player ratings there are often way off, this is proven by the bot making undeniably horseshit teams from time to time(based on those ratings), and then you're stuck with it... Honestly, what's always worked there best in my opinion, is random shuffle until you get something that looks okay. It's the same few dozen ppl playing anyway, we know each other.

The only problem I could see is that QLRanks encouraged a few bad individuals to winjoin/teamstack purely for the sake of rating, and in the same way QLStats could encourage some people to alter their gameplay for the sake of rating, even if it goes against what they're supposed to be doing. But I think QL is just too old now and past that point of people caring. If people refuse to teamplay, that's probably because they don't care/don't want to/incapable, not because of ratings. I could be wrong though, a few exceptions perhaps.

I fully agree... rating individual performance in CTF doesn't work. It was an experiment, based on formulas from ctfpickup.

The omage CTF servers use their own system for rating and shuffling.
QLstats is basically only an observer and create ratings that aren't used. Therefore I didn't spend any further time on improving CTF ratings.

Team win/loss would have more meaning, but unfortunately there are a couple road blocks preventing me from using it so far. Glicko is designed for 1vs1 comparisons and provides no means of updating individual player ratings based on a team win/loss. There are some proposed workarounds, but they also have severe problems.

And trueskill is not an option, because it is patented and even if it wasn't, there is no working Javascript implementation that could be integrated in the nodejs process. (There are some nodejs modules, but they are either unfinished or produce incorrect output).

a good way to sort out of the +B playstyle that this elo system has single handedly caused would be to reduce the amount of elo you get for dmg but add a multiplier for the winning team.

please consider this, pred!

No matter what I do, stupid people will continue to do stupid things.
Gaining "high elo" is shooting yourself in the foot. All it does is give you weaker teammates. Yes, you gain a few positions in a ranking list that compares oranges and melons.

I did consider using team-win/loss based input to the system.
But I also had to consider a dozen other factors too which are not so obvious. And I wrote some of them down in the Ratings Q&A in the qlstats forum.

How good is a team-win/loss rating when you have no guarantee that the server will shuffle teams, maintain balance while players leave and join the match, playres kicking other players because they drag down the team, players not filling up uneven teams because they are about to lose, ...

Win/loss is good for a league setup when you have control over the servers and the data the game sends you. Since I don't have that luxury, I can only try and make the best out of it, which involves a long list of compromises.

it is true that there is no guarantee that teams will be shuffled but theres a couple of problems with your assumption here..

Just assuming old qlranks aka trueskill then lets say you dont shuffle and you get a "shit team" well then you wont really lose any precious rating either. You are not likely to win but the rating system knows and gives you the minus accordingly. Players joining and leaving you could deal with whatever way qlranks dealt with them.

Problem is now you assume the worst about everything and create a rating system that has the wrong incentives and creates "worse shuffles" in general. So yes some people / server have games where people join/leave in the middle but the main cause of that is poor balancing to begin with and also you create a whole bunch of new problems and incentives instead. All the problems with people joining/leaving etc can be figured out and the "you cant help getting a bad team" / servers dont shuffle is a non-issue to begin with.

Well if stupids do stupid regardless of everything i wonder why the situation with +b now is 10 times worse than it was with elo rewarding teamplay. I dont know where u get that impression and why u keep sticking up to a damage based system in a game where dmg is not exact and its inflated if u shoot ppl with low health and u tend to kill ppl ith rail or directs.

on an additional side note, the system is clearly too biased towards the top end as the starting glicko (1500) puts newcomers ahead of ppl who played for years and routinely completely kills teams.

dmg is definitely a good way to calculate who the better players are - the best CA players are all high elo therefore the system works.

you should keep that as a metric of skill but also add a slight multiplier for the winning team! that is the objective of the gametype! if anything replace the 'kills' multiplier that you have at he moment with it!

that's a false assumption. You cant compare the best of the best with the average joe.
The flaw in that reasoning is that the best can win basically 1v4 AND deal a lot of damage both. While average dudes won't have the ability to win unfavorable positions and they can be the problem of the team despite being top dmg.

Also, again: the damage is not exact. winning team routinely has an average of 350 damage. And if you are the last man standing every time maybe something like 10% of your damage output never happened.

Public CA was an empty enough experience before you took away the reward for winning. I am glad you made the change away from comparing against your teammates, but I would really like it to be that the goal of winning the match isn't secondary to the goal of doing high damage.

About the oranges and melons thing, its just fun to see a number that goes up when you play better. And its not 'shooting yourself in the foot' that getting an even higher number means having to play even better.

+ everything qwih has been saying

+B is just the smart and efficient way to play QL. Even more in CA than in other game modes because there is no items to rush for. Being the last one means you are elo farming? The rare times i play 3v3+ CA I am often the last one because my team mates attack alone 1 by 1. Does this mean i should end "lost" games sooner?

I think his new elo calc is worse than before. But it's still better than nothing.

It is just false that the better way to play ca is relentless +b from the start. CA has the objective of helping your team win. The problem u expose comes up in 1v1s and 1v2s moreso than from the start, at it was partly mitigated by hp based rounds and players based rounds (altough it was done rewarding the stronger player so it is still not ideal). Ppl have played this mode forever and veterans dont do it for a reason. It's far from me to consider ca as a good game mode, but the rationalizaton of your poor play tilt me to no end. It's time to consider that ur team dies 1 by 1 because u +b from the start and leave them to die and not viceversa, as coordinating +b teamplay is way way harder than coordinating +f play in public.

With "new" and "before" do you mean the latest change where it only compares you with your opponents (and no longer with your team mates) - or do you mean qlstats vs. qlranks?

That last change was an attempt to remove an incentive to play against your own team, but the rating quality seems to suffer from that change.

can u put "1337" instead "disabled" at my qlstats status?
http://qlstats.net:8080/player/12851#duel
kthx

you look disabled

eat my donkey kong

That's probably the longest list of aliases i've ever seen

nah, he's just a wannabe faker.

here's the official faker ranking:

#1: http://qlstats.net/aliases/76561198267048362

player_id | count
-----------+-------
18146 | 371
121 | 363
1843 | 356
1857 | 330
1530 | 255
386 | 221
12851 | 158
8156 | 142
1254 | 97
2087 | 91
13707 | 89
1146 | 89
12883 | 87
95 | 83
48544 | 80
40994 | 78
1656 | 78
6414 | 76
50837 | 75
619 | 75
3435 | 72
5856 | 72
5264 | 72
2587 | 71
49813 | 69
36484 | 67
7198 | 65
7117 | 64
6208 | 63
6331 | 62
3901 | 62
6314 | 62
5548 | 62
2116 | 62
9503 | 61
23626 | 61
4381 | 60
17197 | 57

So i just played a game with a guy who purposely makes his team lose then he starts shooting. He got up to 1700 ;)

Any you blame me now for him playing like that?

You're implying that he has an "Elo OCD", that this type of playstyle is his strategy to increase that "Elo", that his true skill is less than 1700, that he would otherwise not play like that and that stats padding didn't exist before qlstats.

If the rating is really as bad as you think, server owners will realize it one day too and will use the superior 50/50 random shuffling instead.

I blame you for having a team with a guy with glicko 1700 (above average) that should have elo 600 (absolutely useless to make your team win) and wasting my time.

Pre-steam there was no way for this dude to ever affect me, if he's bad or just a troll i don't care because teams will end up fine. This had a chance to happen only if all the 1700+ elo servers were empty.

And i'm implying that he plays that way purposefully to troll, thanks to your system he can completely ruin the game, and he can do it up to the 1700+ servers :D.
But since the alternative is only random shuffle thank god this viable system has nobody winjoining anymore! My freakin hero!

It's laughable how stubborn you are to the amount of people and evidence against that system.

Here you can see why no one contributes to PredatH0r's projects. He does not accept any contributions or ideas. It is either his way or no way. Wow, do I miss the days of the old community coders before this self-professed savior of QL came along. They may have been slower but at least they welcomed feedback from all

Getting called Cheater everyday only because
This faggot deactivated my qlranks Profile ...
But Real quake pros dont think I cheat...
Get over it and learn to random dodge wannabe t4 fags

You should show some appreciation and be grateful he took the time to make the site. I'm not a ca player, but on the few occasions I have played it with and without the stats, the latter was much worse. He doesn't have to accept every idea that's presented to him, but he always seems happy to listen. If I was him and I read complaints like this, i'd just shut it down and turn the site into a mirror for 1 man 1 jar.

You should show some appreciation and be grateful he took the time to make the site. I'm not a ca player, but on the few occasions I have played it with and without the stats, the latter was much worse.

Oh great, your gibberish is based on your experience from a few occasions you got lucky playing lolca.

What you shitbirds don't seem to gather is that the main purpose of such a ranking is to assist fair matchmaking.

Wait on a duel server, someone connects, warmup takes one single combat then he suddenly is idling while he stalks you on qlstats.
After that he will:
a) instantly disconnect.
b) instantly ready up.
Repeat it a couple of times and later on you can evaluate those players' ratings:
a) had same or lower elo.
b) had much higher elo.
Suh gud matchmagink, mucho wow, apreziation pliz.

On team gametypes you additionally have the problem of retards like komap systematically ruining games in order to fake their worthless ranking, but the main problem you missed:
It doesn't really matter what ranking algorithm (elo/glicko/trueskill) is used or how accurate it is, the real problem is located at the shuffling algorithm of minobot.
From what I have heard it uses several procedures (like compensating the best player with the worst, the second best with the second worst and so on,) but from the multiple compositions it picks the one with the smallest difference in the average ratings.
Great, 4v4, 4 good players versus 2 very good players and 2 shitnoobs that can't strafe.

The average of the individual players' rankings is a horrible criterion to judge team balance.
Instead, a much more intuitive and straight-forward shuffling algorithm could have been implemented from oldschool pickups:
Balance top-down by sorting 1-1-1-1-1-2-1 (depending on which pickup ruleset your prefer) and flush the smeared average value down the toilet.

Back then, people used common sense to judge if teams were ok and if not, they could shuffle again to obtain a new random constellation.
Nowadays, people blindly trust minobot's ranking based fixed teams and even when teams obviously are bad, they can shuffle as much as they want without any significant change.
Ask your plusser yakumo, the pet that instantly readies up on unbalanced teams and then chases the only good player from the enemy team with the highest ranking on the server, like my ass shits honey.

So regarding matchmaking, glassass would be an improvement over qlstats.
And predathor indeed is a nazi that doesn't think through suggestions contrary to his believes, assumes people can escape his ratings without any indicator and didn't even disable my tracking after I claimed that I use cheats.



Anyway, ql's gay +back rl spamfest of item farming tierslummers is dead and now bow to doom4's mighty 8-10dmg-lg roasting.

Wow not even reading past the first line, 8/8 for effort m8. Let me know when the next chapter of the salt chronicles is out.

Doesn't surprise me:
Centuries of small gene pools within the conquered island zones made cognitive skills, like reading, really rare in the uk.
I am afraid the next chapter would be too mentally challenging for you aswell. >:)

0/10 for the sad comeback attempt:
You commented on something you didn't read? That is like admitting your reply is irrelevant.
But calling it "chapter of the salt" accidently implies that you know the content of my text.

predathor indeed is a nazi

you insulted a nice guy who works for free on quality content, while it seems to me you can't just accept how some ql players behave.
or you've just run out of pills? :>

Ask him how to avoid getting tracked and he will tell you to play on unranked servers without having an indicator which servers don't get tracked. I even found myself claiming that I cheat in order to get deactivated at this totalitarian system.
Or try to convince the austrian that playing ctf def gets underrated while damage from mid gets overrated, you will realise that "nazi" is more of a fact than an insult.
Scroll up to comment #108, where he hands out Stars of David.

The quality of his content? You mean like fair matches, while qlstats serves the direct opposite?
Or do you refer to tracking server ips of specific players, in case a retard wants to ddos online tournaments?

Yeah, high quality content, just I never asked for the content.
It is like an unwanted dog shitting on your carpet, with really high quality creamy shit, but at least you are happy that it is for free.

The whole idea is retarded by default, because the elo concept was designed to be applied based on official tournament results only instead of random public fun games on unequal pings. And you talk about quality of such a fundamentally flawed system?

What pills and where are your arguments? Seriously, do you have any arguments?
Do something useful and answer me the following question:
(Assuming that I remember correctly, that you had mathematics knowledge from an engineering background.)
Do you think that the average rating based on 4...5 probes (players) is a statistically good comparison criterion to judge team balance?

As i see it there is a much better chance to improve the rating by him shutting down his site so someone else would bestimulated to make a system that works.
The guy absolutely killed the fun out of the only mode that made sense to play with my setup. And i should be gratefull? If tomorrow im gonna come to your house and fuck u in the throat should u be thankfull coz i did it for free?

As i see it there is a much better chance to improve the rating by him shutting down his site so someone else would bestimulated to make a system that works.

Nothing stopping anyone from setting up another system as it is. If there is truly a better way of doing things, then someone could implement it and server owners would be free to pick and choose which system to run.

The guy absolutely killed the fun out of the only mode that made sense to play with my setup. And i should be gratefull?

Player rankings for the top players seem to generally correspond with the previous qlranks ratings. There's potential for abuse in any system that could be implemented, and overall this system works fine for most people.

If tomorrow im gonna come to your house and fuck u in the throat should u be thankfull coz i did it for free?

Wow is this really your analogy for what predathors elo system did to you. Sorry.

Im gonna say it again coz its a simple concept, if qlstats dies tomorrow theres a much bigger chance somebody does a new ranking than now. And hopefully its not damage based.

Also the logic behind " it works for some then it always works" is flawed. This system rewards baiting so much that a guy with no skill who prposefully trolls by letting his tteam die got up to 1700 instead of having 0 elo. And the system is damage and kill based in a game with no exact damage. If my team does 299 damage to eery player and i kill everybody with a direct i will have score 8 vs score 4 of everyone else despite doing the least. Saying this works is as wrong as u being gratefull regardless coz it was free.

guess what? I was thankfull for a while. Then found out what happens with new system, i told predathor and he said i was talking nonsense and treated me like an idiot. So hey thanks a lot.

self-professed savior of QL

Source?

When I get feedback from 20 people saying "thanks" and from 5 saying "FU" I go with the 20.

I appreciate any feedback, but it should be obvious that I can't change the system every time a few people wants it to be changed.
I can't implement suggestions like "quitters should be punished by losing points" simply because the game doesn't provide any data whether a player quit, got kicked or moved to spec (because someone in the other team quit or got kicked). These cases are indistinguishable. Now, if I just punish "quitters", it means that I would also punish the weak player who was sent to spec.

And that's just one of dozens of examples why the sytem is as it is: flawed, but the best I can come up with, considering all the factors that most people aren't even aware of.

What you also might not be taking into account is that at the time when qlstats was built, there was no minqlx available to shuffle teams. So team win/loss was no viable option. Then there are aspects like how fast a system establishes a useful rating for a new player, which would be a lot slower when based only on team results. Then you have the 1500 players which ruin 50 matches till they have a plausible rating.

Whenever I get feedback about something that makes sense, is doable and I have the time for it, I do it.

thanks, nice work

I played 14 matches tonight, won 12 of them, and gained 4 elo. Contrast that with a guy who plays 4 matches, baits his team into losing all of them, and gains 20 elo. I am not saying 'FU' but please consider improving the system to avoid cases like these.

You have a win percentage of 54%. qwih has 53%.
In an ideal world a perfect system would be able to shuffle teams so that every player has a 50% chance of winning. That's the purpose of matchmaking.

It is a common misperception that good players should have > 50% win percentage. If you have 1 "pro" on the server, then you also need 1 "noob" to balance things out again so everyone has a 50% chance of winning. In a perfect world with perfect matchmaking even k1llsen would only have 50% wins. I know this sounds totally wrong, but it is correct. The reason why he has more than 50% is because there is nobody weak enough on the server to drag his team down enough to create a 50/50 chance for both teams.

I looked at your matches from last night. Out of the 14 matches there was only one 10:2 blowout and that was because of a quitter and unbalanced teams. So the matchmaking part seems to work.

The rating numbers aren't an open-end XP system, they try to form a win probability distribution. The system has a certain expection about your performance and that of all your opponents, compares them and adjusts them. If you performed just as the system expected you to do, then there will be no change in the rating.

A system with "team wins -> you gain points" is certainly easier to understand, but not automatically better. In the current system you gain points by performing better than the system expects you to. For CA this means that you have to deal more damage than lower rated opponents and/or more than a higher rated opponent.

"In the current system you gain points by performing better than the system expects you to." That is false.
You gain points by having more SCORE than the system expects you to. That's different than playing the better CA game than the system expects you to.

In the end, which performance metric do you suggest? Only win/loss?

Ye. Qlranks was good. With highest elo ever reached too. Will take a bit to balance tho if it cannot be done retroactively

Almost right.

The system doesn't expect a particular score value, only that you have a higher score (=Perf value: damage/100+0.25*frags) than lower rated opponents and a lower score than higher rated opponents. Except for a small margin to consider the matchup as a draw, the amount of the score difference is irrelevant.

You gain points against opponents who had a lower perf and you lose points against opponents who had a higher perf. The final aggregated amount depends only on the prior ratings + deviations, but not on the actual perf values.

No they dont try to perform a win probability distribution now do they? They try to perform a dmg probability distribution?

That is correct, it's the probability of you having a higher PERF value (which is damage/100 + 0.25*frags) than another player.

In the case of CA there is a high correlation between the damage output of a team and weather it wins or loses, this indirectly gives you a win probability (+ some extra prediction error).

it is kind of interesting that you say "win probability distribution" since that is pretty much exactly what "trueskill" does.

Isn't that the case with any skill estimation system?

no in his case he uses an indirect approach to determining winning chances, their dmg relative to opponents. Winning is not a part of the equation for determining the rating used to balance out the winning chances.

With trueskill/qlranks it was more determined by "winshares", what strength opponents you had managed to win games against (not outdamage on an individual basis).

Correct.

Trueskill is *designed* to calculate rating updates for individual players of a team. Elo and Glicko are not, they only work for pairwise comparisons.

Since trueskill is patentend and there's no working javascript implmentation either, I had to work with whatever is available.
Some use Elo/Glicko in a way where each player is compared to a single averaged oppenent, but IMO that doesn't work well either.
A player rated higher than the average would lose more points in a lost match than he could gain by winning against that lower rated average opponent. Repeat that a number of times and all the ratings converge at the average of 1500. Please correct me if I'm wrong. Another problem is to calculate an averaged rating .. when people can come and go or switch during a match.

It's not that I'm not willing to use team win/loss, it's more that there are a bunch of road blocks that prevent me from doing so.
TrueSkill is not an option, so any comparisons with it are moot.

If the problems can be sorted out, I'm all in for changing the system. Unfortunately most of the suggestions and proposed solutions opened up more problems.

Anyone serious about improving the system is welcome to talk to me on IRC so we can work things out.
Some of the big problems are:
- rating updates for individual players
- partial play time (late joiners, quitters)
- team switches (red, blue, spec)
- quit, kick, conn interrupt, moved to spec ... can't be distinguished
- no guarantee that teams were shuffled at the start
- matches that became unbalanced by players leaving/joining
- players with new accounts (either noobs or smurfs) and and unreliable 1500 rating
- ways to exploit/abuse the system
- player behavior (bullying weak team mates vs. CFFA)

Have you tried contacting the old qlranks people and getting a hold of whatever they used? I think wdx has a working implementation of "trueskill" if nothing else, talk to him.

I did talk to szr, about his experiences and also about the prospects of updating and continuing qlranks.

qlranks is based on Windows technologies (ASP.net, C#) and the infrastructure required massive computing resources which cost a significant amount of money every month.
Also, the source code is not public available and I didn't want to work on a closed source project.

A working Trueskill implementation is one thing, but Microsoft only grants licenses to Xbox titles. Reports that Trueskill can be used for non-profit projects are not backed up by any Microsoft sources.

2 or 3 months after qlstats was up and running a statistics PhD student and QL player helped my analyze the collected data and it looked quite ok. He did some more research and came along this:
http://jmlr.csail.mit.edu/papers/volume12/weng11a/weng11a.pdf
It contains a proposal for a non-patetend multiplayer rating system, but I have no idea if it is suited for QL and I didn't have the time to implement and test it.

A brave new world where math algorithms can be patented.
Also I don't think you can actually patent math in EU, so as long as you don't host it on us servers you can use trueskill. You just can't use "trueskill" trademark

lzw, gif, mp3, mp4 ... all were patented in the EU.
When the patents expire after (20?) years, the algorithms can be used freely. Before that, a license from the patent holder is required.

There is a way higher correlation between winning and winning.

Of the matches i played in that stretch of time, 6 out of 14 were what i would consider poorly balanced, with scorelines of 10-6 or worse. That's not terrible, but it doesn't make me overwhelmingly enthusiastic about the system.

I still think that on the simplest level, a ranking of team mode performance should reflect your ability to make the team win. Otherwise, what is the point of winning? The success of QLranks tells me that such a system works for matchmaking as well as being, as you say, 'easier to understand.'

Anyway, thank you for listening and for your work.

Whats really on character is that you dont consider why they say thanks and why the others say fuck you. If 20 people say thanks because any system is better than nothing and 5 ppl say that this system is worse than elo for 10 logical reasons, not only u see it as a fuck you, you assume the first group would be against the change thesecond group wants and just discount the arguments based on how many back them actively instead ofactually thinkin about them logically.
You are by far the person who makes me come closer to quit quake.

I read somewhere that online on average per 1 negative review of a product there are 80 silent satisfied customers.

Those squeaky wheels...

You would think that if you're criticizing people who devote their spare time to improve the game FOR FREE that you would at least have the balls to not use an anonymous account

"improve" xD

how would you pad stats at "qlranks"? By winning games?

"winjoining" was the only way to exploit the system. Basically somebody very dedicated could wait around a lot to steal 5 points or so if he really wanted.
A problem mostly fixed by the command !teams to swap two guys and make teams fair again.

but I think you still had to play a certain % of the game. Sure you could join a winning team that had "good chemistry" but yeah, compared to how you can really actively pad your stats every game now it is not an issue whatsoever

Ofc i agree 100%, with the risk-rewrd remark too

"kick your weak team mate" was common practice too.
Two teams balanced by average rating, then the team with the weakest player kicks that player for a stronger player to join (or the other team to get rid of one player too) will leave you with a higher average than the other team.

Or wait out the shuffle and join later, when a weaker player comes along to join the other team.

There was not a single game where nobody got kicked (granted, i never played on any 2000+ servers ... but neither did 90% of the other quake players)

So this is what it is? You got kicked when u started? Its funny coz it wasnt that common, not every server has vote on midgame and it actually served the purpose of letting newcomers go play other newcomers when it happened. And the new system doesnt stop it from happening, people still kicks players out of their tier or just the bad player whos always last and ding nothing. But hey at least new players now have a glicko 5hats above average.

if you change players then the relative ratings and the risk/rewards changed accordingly.

qlranks was atleast 10 times better than qlstats. Why cant it exist again?

why the fuck do people care about a rating
wich a t3 nab made? srsly? he has 0 elo at
duell and 1500 at lolca...

Then why do you play a game which has a 16 years old engine written by a coder who would even struggle to win a game in t4, let alone t3?
Seriously, you are so dumb, my monitor almost refused to display your post:(

fail

Still amased at the amount of ungrateful people.

Is it possible to create new account in stats without changing steam id?

No automated way, but I can update the database manually if there is a good reason to do so. I'm not assising serial fakers and cheaters in covering their tracks.

are there any criteria to be ranked? Not that it matters much, but I feel left out :'(

>= 50 rated matches for a given game type and your steam-id must not be on the "deactivated" blacklist.

ok, i see.

I'm sure I'm not on any blacklists.
It says I have 52 matches played in ctf tho. 98 matches overall.


Rating: 1615 ± 68 (ctf, 38 games)
Not sure what that means.
38 rated games then?

yes u need 12 more

Just a question.

I spawned a vanilla duel server, played a few games with my friend and later on, the matches didn't show up on qlstats.

What might be the cause of this?

You need to add it at qlstats.net to get "polled" for stats, it is not automatic.

^^ that, and you need to set some command line options for your QL server so that the game open a stats port where qlstats can connect to.
+set zmq_stats_port 27960 +set zmq_stats_enable 1 +set zmq_stats_password yourSecretZmqPass.

I recommend that you use the same port number for stats and for the game port (one is TCP, the other is UDP).

Now playing ,not showing any people anymore on server.is it just me?

Works fine for me.
Is it a particular server that you're having this problem with?

i have that on firefox and chrome ,,,, not showing any players on server anymore :/ dunno why this happens ,,, problem since 2 days

Should be working now.
There was a problem with the HTTPS support I added 2 days ago.
No idea why it worked in my Chrome, but I could replicate the issue with FF, IE and Edge.

Fixed! THX

for some reason qlstats stopped tracking my duels since 2016-03-11
ca/ffa and other gametypes works fine.

Is it possible that you are playing on unranked duel servers?

dunno. I doubt that every duel i played since then was unranked.

It just so happens that a server where i ping best is unranked, so most of my matches are not tracked. Figured that could be the case with your games. Maybe not

That's my only explanation too.
Maybe the server owner changed IP but didn't update the information on the qlstats tracking panel. There's quite a lot of unresponsive servers on all four panels.

qlstast down? :(

Minqlx server ops need to run 'git pull origin' in the minqlx-plugins directory to set the new minqlx defafults,

then either restart their servers or just manually change qlx_balanceUrl from http://qlstats.net:8080 to http://qlstats.net on any running ones they don't want to restart yet.

8080 is now a blocked port.

Yes it has also had the same problem qlranks.com did for awhile, too much traffice at certain times for the bandwidth it had alloted.(just a guess because of 15 second load times.

Hi,

Is there a way to reset my elo stats? I've been playing forever but for some reason, I think my elo is too low. I was playing last night and the admin of the OSO server actually kicked me saying I was "trolling" b/c I was playing with 1550 elo, but being top of each game. I do see his point in that the teams would be imbalance if my elo did not reflect my actual skill level. I have no control over elo but I've had over 800 games for the elo I currently have. So I ended up putting on a handicap to help balance the teams and ended up being around middle scorer. This results in negative elo every match for me now so my elo further plummets and makes the situation worse. Is there any way I can reset it and begin anew? Maybe it'll be more accurate if I can do that.

Regards,

Gr4v

Don't sweat it. The people accusing you of trolling are just being idiots. This game amplifies tiny differences in skill into very large differences in score. It is also very top heavy, with tones of super high skill players around.

Placating people by turning on a handicap is nice and all, and if you've got something good going it might even make sense here and there. Kind of pointless in the big picture though.

All that said, new elo system punishes inconsistency quite a bit more than the old one. If you really need it to go up, you need to avoid playing the occasional bad game.

Using handicap is exactly what you shouldn't be doing and it's totally unnecessary when playing with elo shuffle. I mean that's a legitimate reason to kick someone, if you've been using handicap to decrease your elo and then suddenly turn it off. If you hadn't used it your elo probably would have increased and even If this game was unbalanced your next game would be more balanced, now it will be even less balanced.

Just play in other ranked servers where noone force you in handicap for 10 games. Rank goes super fast up if it's a lot lower than your skill. You can gain 100+ elo per game.