http://qlstats.net:8080/

Where are the leaderboards for highest elo by gametype or countries? It seems to track stuff, but does nothing with the data so its basicly worth nothing...

Did you want stats or a big e-penis?

http://qlstats.net:8080/ranks/duel
http://qlstats.net:8080/ranks/ca
http://qlstats.net:8080/ranks/ffa

Seems alright, but the rankings page takes a long while to update. As an example Rapha has one of the highest elos but has not yet been updated on the duel ranking list

I have to run the update script by hand.
And since I know the numbers are pointless anyway, I don't want to feed people's OCD (indluding my own).

Although you personally don't think it's useful, a lot of people are into this type of thing. Please consider keeping it up to date and visible in the spirit of openess / open source. (:

read up on posts by the guy that put that site together - it's still just in a "data gathering" state, so your assessment is completely accurate. however, he has plans to improve and provide actual rankings, etc. just be patient, or see if there's something you can do to help.

How is this thing gathering stats? Should I configure my servers somehow to send stats to it ?

Has to be ranked

It connects to every registered server's live stats feed to gather data. You can launch your server with :
+set zmq_stats_enable 1 \
+set zmq_stats_password "mypass" \
+set zmq_stats_port $gameport \

...and contact PredatH0r with the relevant info for him to connect if he wants to.

No need to contact him
http://qlstats.net:8081/servers.html

Thanks! I am gonna enable zmq on my servers then.

Hey, I just started playing again after taking a break in march.

The site lists my CA elo as 140 after couple of matches, but on some server I joined it displayed my elo as 1000 after using "!elo" command. Fot other players it also didn't match with their qlstats.net elo (it was in 1800's), so what's the deal here? Where do those 1000+ elo values come from?

idk maybe the source value is manipulated in some way to fit the ranges people are used to.

minqlx balance plugin uses PredaTh0r's qlstats.net site, but has options to let you redirect it to another xonstatt site, and people could write their own plugin to do the equivalent of balance with a different source entirely.

I don't think the Omega CTF servers are using qlstats.net for a start as they were discussing a custom rating system.

i did some more comparisons and it seems, that the value is just multiplied x10 so it's as you said. The thing was that the values were not responding to changes, maybe they were cached once I joined the server and didn't update while I stayed.

its the best thing ever. no more elo quitters and scared people.

Worst thing ever its impossible to get good teams and everybody quits after 2 rounds

lulz plussed by mistake

It only helps very sensitive ppl with issues. At least for duel, was almost perfect before I stopped playing, before summer. Almost, no elo whores. Among the hundreds I pld, found only a couple.

Most duelers at some point, tried the elo whoring thing, saw the vanity in that and stopped. You could see it clearly in their qlrank profiles, where their current elo was much lower than their all times highest (at the pinnacle of their elo whoring). At that point, was super rare to find someone elo whoring. The opposite was more frequent, ppl with lower elo than their real one, because of prac accounts or by playing very skilled opponents only. That wasn’t good either, but there weren’t many and you knew them.

On the other hand, elo was giving u a measure of how well you were playing and how u were progressing. On a game like ql, the score only can’t tell u a thing about ur performance. You can win a noob 30-0 and play a crap game, or lose a close game vs someone really good and make the game of ur life.

i also think the fact that the site kept track of your best elo made it so you just needed to elowhore once and you'd have the recognition 4ever

Read:
https://m.reddit.com/r/QuakeLive/comments/3w9y3x/new_elo_system/

It's based on score /time now instead of win /loss, far less quitters since you can be on losing team and still gain elo. Just needs more time have enough games to properly rank everyone.

I had 2 objectives when I released qlstats.net:
- provide better data for matchmaking than a random number generator
- provide statistics to players and server owners (what matches you played, your opponents, your accuracy, ...)

I have NO interest in a player ranking (as in a top-1000 list), but some sort of skill rating is a necessity for matchmaking. At the time of this writing, qlstats is using the unmodified Elo code from xonstat. This is likely to change in the near future, and so will all the numbers.

qlstats is still in an experimental state. The collected data goes almost unfiltered in the database. With that data at hand we can make more educated decisions about what games should affect ratings and which should not. And what sort of data anomalies we have to take care of.

All the ratings will be recalculated some time soon. Maybe with a different algorithm or parameters. The "Elo" numbers will certainly change. Xonstat starts players with 100.0 and never lets them drop below 100.0. The API multiplies the numbers * 10, so minqlx will probably show you 1000 instead.

Anyone who wants to contribute is very welcome to join #stats.ql on quakenet IRC. The source code of the whole site is on github, the website part is written in python, the connector between QL and the site is written in node.js

Bro, we need weapon accuracy leader boards. We gotta see who has the best LG and the best rail. Find a way to rank weapon proficiency by competition, accuracy, shots fired, time played, ping, map, etc.

xD

That won't work. For one thing, you can get much better accuracy on some servers than others. Not just because of ping, but because of the players. You can get better accuracy on some maps than others, and in some situations. How many times have you seen guys camping the top of a jump pad with LG? If they were there to do damage, they could hit a rocket and still rail the guy before he hits the ground. Or just rail him twice for 160 damage. You won't do that much with LG before he's pushed out of range, so there's only one reason for it.
And why encourage people to camp, fully zoomed in, and never take a risky shot, just to protect accuracy stats? The game is about killing the bad guys and doing as much damage as possible.
You miss EVERY shot that you don't take. The guy who did the most damage was the most accurate, because he hit more. It doesn't matter how many he missed, because we're not paying for the ammo, and there's not exactly a shortage of it.

thats deep

Status of matchmaking, predicted almost 2 months ago: http://www.esreality.com/?a=post&id=2781228#pid2781565

is very true

gg

I forfeit a game early with a tied scoreboard, but I had more damage dealt, so it gave me elo.

neat

I recalculated CA, FFA, CTF on qlstats.net with Glicko-1 and started to update the UI for it. It's all under construction atm.

IMO the results look plausible so far:
http://qlstats.net:8080/ranks/ca
http://qlstats.net:8080/ranks/ffa
http://qlstats.net:8080/ranks/ctf

These are not final value, they may be recalculated again.
These glicko values are not yet updated automatically after each match. I hope I get that done later today.

Duel is currently being recalculated...

These values are poop. I dropped from rank 38 to rank 180.

http://www.quickmeme.com/img/9d/9d3ef935fa12e01e0f80c27da35687456a7fc7811262ad87269b4fde78d80ac5.jpg

Guess I'll have to offer a premium service next.
For a small fee you can register a user account and when you're logged in, you'll always be rank #1.
Seems to be the only system that works well for more than 1 person.

Pfff fucking Pay2Win.

Take a clan

great work, PredatH0r

Not sure if you are doing this already but do you bias the results by the most recent win/loss/etc stats of a player? In other words does the system favour the current form of players for more accurate results?

PS Thanks for your hard work on this! <3

Instagib FFA really needs to be separated out from the current rankings IMHO. It's not really useful to have those stats mixed in with regular FFA like it is a t the moment. Also It would be neat to see who's at the top of the gibber list :))

I'm running 2 system in parallel atm.
The old Elo numbers are updated automatically and they include almost every match, except aborted ones.

I still need to run the updates for the Glicko numbers manually.
They ignore a lot of machtes which don't meet certain requirements.

Some requirements depend on game type. For example, I'm only counting matches with certain known factories. There is little to no information available about server settings, so using the factory is the best thing I can do to rule out whacky fun games. Instagib and PQL are also excluded.
Some more rules: https://github.com/PredatH0r/XonStat/blob/mas...ode.js#L76

Some general requirements:
- matches with bots get ignored
- players with < 50% playtime get ignored
- AFK players are ignored (dmg <500 or dmg taken/given >10x)
- matches where total play time in teams differ by more than 2mins get ignored
- minimum player count per gametype
- score and/or time and/or roundlimit per gametype

You are counted toward the team where you spent the most playtime (not the last team you joined), but currently the team has no effect on the performance calculation of a player.

There's also a criteria for "draw" outcomes between 2 players, depending on gametype. E.g. in FFA if you are within 5 frags of another player, the head-to-head is counted as a draw.

Wait, someone actually named their factory mg_ft_fullclassic? This is amusing :D.

As for the other mg_ factories, a few might fit :
mg_ca_7lg - if you accept 7 dmg lg
mg_ca_nohmg - if you accept the lack of hmg
mg_tdm_classic
mg_ctf_classic
mg_ft_uft - maybe, I would need to know what your other FT factories do

I'm not asking you to include them, just do as you see fit ;).

Glicko-1 is that used by qlranks?

No, different system. I know, the pure numbers look similar, but you have to resist and not compare them :)

This is still only an early attempt. There may be bugs in my code, there may be criteria I miss, there ARE inherent problems when applying Glicko/Elo to multiplayer situations ... and it will never be a perfect. Maybe not even a good one. But I hope to beat a random number generator.

Was just curiosity. I don't care so much about my ratings. I just hope this feature in future allows to have some matchmaking.

I think this system is more accurate that that from xonotic, because before a lot of players i know and i play with, was below me and had lower ratings, but they are stronger than me. Now seems everyone is rated pretty well (always talking about players from my circle)

Glicko is more accurate and faster than elo imho.

The Elo system had all players start with 100 points on the bottom of the list. The Glicko system starts everyone with 1500 (-350 uncertainty) in the middle of the field.
That's also why in the first couple matches new players gain Elo and may lose Glicko until they settle somewhere in the lower half of the field.

There are people running around with 4000+ elo right now. Think the uncertainty thing should have a lower and upper bound so it doesn't go crazy.

Actually that 4000 is 400.0

No one have even 1000 Elo or Glicko

http://qlstats.net:8080/ranks/ca?page=1

The Elo number gets multiplied by 10 when displayed in-game from what I saw on the few servers that used Elo instead of Glicko.

Nobody has a Glicko rating > 2500 in CA.
What you saw are the old Elo ratings. I talked to Mino yesterday, he'll update the shuffle script to use Glicko ratings. Will take a bit till all servers run the new script.

Pred, put our ranking back on our page.

Do you mean that it doesn't show a rank on the player profile page anymore? That is intentional.

Throwing European, American and Australian players in the same ranking does not make sense. They don't mix with each other and therefore each region creates its own rating distribution. They are not seriously comparable.
They had the same situation in chess where ratings in Australia drifted away from ratings in the US so they had to introduce a conversion factor.

And that does not only apply to continents, you have the same effect with closed communities where the same people always play against eachother. A local hero may have a lot of points but that doesn't mean he's better than someone lower rated from another community.

For matchmaking the numbers are good enough, because they are applied within the scope of the servers and players where they are taken from.

So...what are glicko ratings based on? Because Ive seen players switch places, where one had higher Elo than the other, but now the other is rated higher. Is it back to team result being the only factor? Is personal score still a factor? If so, are damage and frags weighted in the same way they are in QL (too heavily towards frags, IMO)

Most importantly...will it encourage people to bait their teammates! Yeah, I'm a baiter hater. Political correctness be damned. I hate those people! They all look the same to me! And you just know they're plotting something when you see their beady baiter eyes.

Don't EVER compare Elo and Glicko numbers.

For the Elo rating only aborted matches are filtered out, everything else, including instagib, quadhog, pql, ... 1v1 CA is included. People started at the bottom with a rating of 100.0 (shown as 1000 in minqlx) and had to work their way up.

For the Glicko rating I am using a lot more criteria to filter out matches or single players. New players start around the middle of the playerbase.

Both system treat all possible player pair permutations in a match as head-to-head match and decide a winner/loser/draw and adjust the ratings. That's done for all pairs, regardless of teams.
For CA score/time is used as the metric to compare 2 players. The game calculates the score as damage/100+kills.

I know the argument about players hiding and then collecting easy kills at the end. I disagree with that. If you make 200dmg on one player and 200 on another before you die, you get 4pts, but you left 2 players alive who are shooting at your mates.
If your mate deals 100dmg to each of your targets and kills them, he also gets 4 points. He stopped 2 enemies from shooting at your team, so his score is justified.
If you believe that hiding is a good strategy, try it out. It's likely that your mates get slaughtered and you face 3 of 4 enemies alone. You may hit 2 rails and get 1.6 points and then die.

So glicko is your score, adjusted due to every other player on the server? Cool. Is the result still a factor? Not hating, just trying to figure it out. Personally, I think it SHOULD be one of the considerations, cus the point of a team game is winning the game, and it's not just about raw damageand frags. But either way, it's vastly better than the old Elo where it was JUST result that counted. And starting players off in the middle makes sense, cus it will stop ppl whining too hard about a new account with 1000 elo who rapes everyone and hit's them hard right in the elos.
I. Don't. Hide. Unless it's in the middle of a bunch of green Keels. I don't even have +back bound. Smoker, so...quick deaths afford me opportunities. Especially as I roll my own.

Whether your team won or lost doesn't matter in the calculations.
I experimented with giving the players of the winning team a bit of extra score, but that just adds undesired volatility to the system. Ratings and rankings will become more random and jump around instead of converging at your true personal skill.

Factoring in team win/loss is pointless in casual settings where teams cannot be forced to be balanced... You would get punished if you get bad team mates or if they just fool around instead of trying to win.
It would also punish players who join a 3vs4 match when the smaller team is behind and about to lose.

If you have bad teammates then they also have low rating so you dont lose much if you lose but can gain plenty if you manage to win. If winning/losing is not a factor then you are better off playing for yourself and your score rather than playing for the team. The best way to encourage teamplay is to make the rating win based.

Someone with web design skills please code PredatH0r a nice clean UI for this site =)