Good job, so how is the elo being calculated? Same as duel?

Edit: also will clans come into the calculation? Would be nice to see a ranking of clans.

Nah its different to Duel, its based on Elo but its not the same, its a Bayesian skill rating system. It can deal with an infinite number of players and calculate individual skills from team results. In other words we will be able to apply this to any team game such as Clan Arena.

So if you get a high score in the game you get better elo then a lower score? because you can be an important team player and not get the most points.

The same goes for any sport or game, really...

The athletes you see in the headlines and earn the title of "one of the best" are those that break records or score the winning goal, not the guy who assisted him or helped him achieve it.

So perhaps the elo rating should apply to teams that all players share the same clan tag. There are various issues with this but it might get players in the habit of all having the same clan tags when they play together, even in public games.

Edit: Just to clarify what I mean. A TDM match where every player on that team is in the same clan would count towards elo no matter what the other team they are playing is. As soon as someone on the team has a different clan tag it doesn't count. Probably a bad idea but I wanna see what you guys think.

it's a bad idea. if you want to rank clans, go take a look at the various tdm leagues.
what you want in the end is to rank the clans according to their practice games (mostly), i don't see the value and would rather have something for the invidiuals.

tl;dr: qlranks is different and it's a good idea (imo).

if you want to rank clans, go take a look at the various tdm leagues.
what you want in the end is to rank the clans according to their practice games (mostly), i don't see the value and would rather have something for the invidiuals.

if you want to rank player go look at the various duel cups.
what you want in the end is to rank the players according to their practice games (mostly), i don't see the value.

i really think a tdm ladder thing for clans would motivate a lot of players/teams and it would be a cool thing to have and compare teams. sure it doesnt mean that much but the same goes for the duel ladder and everyone is goping crazy about it nevertheless. people care more about their matches when they will show up in some form of ranking. it would be a good thing for the tdm scene.

if you want to rank player go look at the various duel cups.

that's were i would go to see how good someone is doing at duel, not tdm. some people play one, but not the other .)

maybe the tdm ladder for clans isn't that bad a idea, but really, if i want to know how good my team is, i'd start playing competitions, being mainly ranked on practise games doesn't say much in tdm (imo).

i was trying to point out that everything negative you said about a clan ladder applies to the current duel ladder aswell.

ranking players for tdm might be interesting aswell but i fear that it will encourage players to focus even more on their own stats instead of their team :/

i agree. pubstomping will reach all new levels if this gets enacted.

a quick google tells me you either:
(a) used the xbox live ranking system
(b) made your own fairly complicated rating system entrenched in probability theory.
(c) completely misunderstood the word "bayesian"

which is it?

I find this sort of thing fairly interesting, since im studying for a maths degree.

If you are using the xbox live one, why dont you use it for duel too?
The difference between 1 vs 1 ranking and N vs N in xbox live is that they add up the teams ratings to compare team strengths, and that is very easily applied to Elo, so any benefits over elo in team games also apply to 1v1, the main one being a much faster convergence speed without sacrificing accuracy, which would be very useful, since there seem to be many qlranks profiles which show a general trend towards a certain ELO, but are still quite a way off even after hundreds of games.

still thinking this is pretty worthless unless you take clans :p
good job never the less.

Once I get TDM to Duel level, CA is coming next (then FFA)..

Just curious, ctf is last based on what?

Well CA is played more then anything probably. Though CTF should be before FFA. Actually I don't care about the order because sirax and szr are doing an efficient and punctual job at bringing these out, they will be out sooner then saying "QL Update".

true, they're efficient. I just thought ctf had more interest than ffa or tdm. ah well, it'll come when it comes.

ctf has never had more intrest than tdm D;:

on ql its' a diff story:
http://www.quakelive.com/forum/showthread.php...ould-it-be

it's the true story

tdm is pld far less than any other mode, but the ppl who play it are hardcore and thus more vocal. They also hold key positions and roles in the community (like trance and the level up group, 2gd and probably even sirax).

imo ffa should had been 2nd for many reasons that I won’t bother to mention (it won’t change anything anyway). sirax and the rest are the ones putting the hours and effort and they are fully entitled to follow their preferences and continue with this project the way they envision it. Therefore, I am glad we have this and I have no problem w8ing.

In other words TDM players are proactive

I won't object to that, it's true

i can see that. Since i'm the ONLY vocal ctf player around on this site lol. Good point.

Let CTF be last then, and really at this point, without anyone else stepping up to voice the concerns of the CTF community, they might as well forgo adding it all together. This would allow more time to perfect what they have already created to date.

haha ingame forums, ok.

I am defeated.

indepther explained the difference. popularity wise, overall ctf is more popular, but tdm has a more proactive community, despite it's small size. And so with that reasoning i concur that tdm > ctf, relunctantly.

Time will show, we will get a better idea when sirax has it tracking both tdm and ctf.

To make you go "zomg its a conspiracy!"

because they suck at it

My guess is it's last because it's the hardest one to calculate. The ranking in it won't be as accurate as other gametypes (like duel) mainly because some offense players (I'm using off. as an example because I play offense and it should be fairly easy to figure out what I mean) are the bum-rush type players that would rather continue applying pressure to the opponent than have a stack and run in and fight for the flag. Plus, you've got the positions that you'd have to worry about (which you can't actually calculate because there's no "set" positions.

However, I'm used to having to wait for stuff. This isn't the first time :p

Because without aircontrol CTF is a shitty fucking gametype.

Data was being collected for the past few weeks, correct?

Cool stuff man. It may be plain, but it gets the job done. :)

Hey its Pretty Awesome.

I might even be 2GD.

2G 2B True

But still Pretty Awesome.

Or Some one pretty.

Clans, please :)

It would be good if it could list the clan tag of the player and even steal the formatting? Just an idea.

Clan tag at time of the match is stored.

I.e. if you play for clan X during match A, that data is stored and can be isolated.

This is in order to allow clan data to be recorded for multiple players of a single clan. This data is even stored for duels which is largely useless :P

Clan rankings will be based upon performances where the majority of players on both teams use the same clantag, so keep that in mind if you want clan matches to be ranked in future..

Good to know, thanks for the reply.

It's hard to know what those numbers mean aside from larger being better. Something like plus-minus ratings would be easier to interpret, and are basically the same thing. They're also called roland ratings.

Plus-minus is something pro basketball/hockey teams use. In the context of tdm, adjusted plus-minus would be the additional number of frags an average team would get or give up when a specific player is on the team.

Also, the duel elo scores have the problem that if someone plays a lot of games against slightly weaker opponents their score becomes unrealistically high. Plus there is probably a net flow of points into the system due to people starting with 1200, losing a few games, and not playing more. Online chess elo has that problem anyway.

The reason these rankings work well for quake is because players don't play for specific teams (i.e. #tdmpickup). Hence they have a higher analytical value over a roland rating style system, which is designed for a player's value to a specific team.

All that matters in these ratings is 'larger is better'. That simplistic view is the fundamental beauty behind all rating systems.

All that matters in these ratings is 'larger is better'. That simplistic view is the fundamental beauty behind all rating systems.

Huh? So this is only supposed to be an ordinal scale? I would have assumed it's an interval scale.

The second part of your post is simply wrong. Some rating systems allow for measuring on an interval scale.

I don't think 'larger is better' has much to do with the scale.

And it was a typo before, I should have said 'all rating systems that I would consider using' :)

All that matters in these ratings is 'larger is better'.

This implies an ordinal scale, imo. Because on a higher level of measurement, some other things matter, too.

But wait, isn't the ELO system measuring on an interval scale? I'm pretty sure it is. Someone with 1200 is better than someone with 1000 by the same amout as someone with 1400 is better than someone with 1200.

Yes that's true - but that's transparent to most people. All that matters is assessing relative strength from a quick glance.

This is quakelive player ratings, not statistical analysis..

So you can get more information out of the ranking besides "player x is above player y, so x is better than y". You sounded like this was the the only thing it's good for.

There are adjusted plus-minus ratings that normalize for always having good or bad players on your team, so it can be done independent of your teammates. The only real problem is that if you always play with another specific person on your team then the two of you will have the same rating.

Here are the papers, I think. http://scholar.google.com/scholar?q=Joe+Sill+...i=scholart

dont wanna sound too negative but i am not really a big fan of encouraging players to focus even more on their personal stats instead of their team.

Clan rankings will be based upon performances where the majority of players on both teams use the same clantag, so keep that in mind if you want clan matches to be ranked in future..

but i am very very glad to hear that some form of clan ranking is planned aswell :)

It doesn't encourage personal stats. If your team wins you gain points, if you lose you (mostly always -you can gain) lose points :)

oh okay, thats good i guess. would be cool to know the actual formula ^^
maybe i am just to pessimistic cause overall it will probably be a good thing, but i can already see players unwilling to even try to get fair fair teams in pickup then for example. no matter how you do it, it encourages abuse. but even elo doesnt seem to help a lot there seeing how many people seem to farm +2/3 point wins.

I guess thats true cos I played some tdm recently which I didnt get the high scores but I controlled items and positions that lead to the team winning. And Im ranked 16th thus far.

Guess TDM and CA is pretty much the same so its just a gamemode copy ;)

LOL NO ARE U STUPID LO!L!!!!LLLOL!!!!!!!!

DOUBLE REPUKKEN!!!!

Excellent stuff, role on the CA :)

looking forward to FFA ranking, great great great job!

nice work! Well done and thanks:)

Good job!

why ca or ffa next? why not ctf?

FFA is a bit more complicated, because all players are on their own team. CTF could go in tomorrow to be fair.. dunno if there is much interest

there is.

in all honesty, cater to the euros first, they want tdm, so give em tdm. There's nothing going on in na now anyways, and if na wanted anythng it would be ca.

I care mostly for ffa. Maybe even more than duel, cause I got the impression it would be harder to pad ur ranking there and the ranks will be closer to reality.

also there is so much speculation about who are the top players in ffa (have seen many threads about that). qlranks could end it with an answer.

plus we will get a good indication about the size of our community, since everybody plays some ffa from time to time, unlike duel and tdm (ca also could do that)

why even bother with ffa? such an unimportant and useless mode that only gained popularity cause 2gd made fun of tony most of dreamhack ... rly :D

ctf is much harder to rate since it got positions

it's the hardest and most subjective imo

Good try. 5 players currently sitting in top 10 (including #1) are never even to be seen in tier 4 servers (not to mention pickup/clan games). #10 tdmer in the scene has 17% profile accuracy.

I really hope this is only because game sample is too low or something and they will lose lots of points the first time a game with a proper tier slummer is processed. But so far it looks horribly skewed...

inb4 q_adrian ;)

in regard to which gametypes you should do in terms of how useful the information will be, which is pretty much inversely proportional to the amount of work required to implement it::

FFA ratings will converge fastest out of them all: a rating system seeks to order all players in skill level, and an FFA game orders all players on the server by skill, since the only aim is to get points and get as high on the scoreboard as possible.

CA will be next fastest. Since each round is completely separate, you can treat them all as individual games, so it will converge >10x faster. Plus, there will be more variance in the results, so it will have a greater accuracy, i.e. less chance some tier slummer who wins every game will top the list, since losing 1 round is going to hurt his rating eventually.

TDM and CTF will be comparatively slow to converge and easy easy to exploit. You cant include player scores or final scores in the equation, since it is possible to exploit that, e.g. stealing all armors and weapons and going on a rampage will earn you points, but the rest of your team will be fucked so you made your team lose. So you only get 1 bit of information (which team won) shared between however many people play in the game, which isnt much.


However, there is something missing from your system that would improve it drastically: forced matchmaking based on ELO. That would completely eliminate the "tier slummer" problem, but, unfortunately, this isnt possible.

the current #1 Baron_Winter sure likes to play terminal heights (that shitty space map) with abusive teams on publics
254:-27 http://www.quakelive.com/r/profile/matches/Ba...fd90/tdm/1
or all the 1 min long games that he started 1v1, way to go!

I'm repooling the data locally at the moment, and looking at addressing the rankings issues. I'm fully aware of some of the crappier rankings.

Perhaps I should just rank tdmpickup channel games or something..

and looking at addressing the rankings issues. I'm fully aware of some of the crappier rankings.

well, filtering out matches that did not last maybe 10 minutes would be a good idea i think. that would stop a lot of potential abuse and it should be very straight forward to implement. besides that i dont think there is much you can to filter stuff like that. or can you somehow tell if a match was private/public?

Perhaps I should just rank tdmpickup channel games or something..

is that possible? how would you do that? would be really cool to have that as a filter option (like you have for regions in duel now)

The problem is that people abusing the system, intentionally or otherwise, would eventually be managed out by better players.

With TDM the community is so small and numbers of actual games played is so small. 'Good' players are vastly outnumbered and play far less games than the 'noob' players. There is little to no interaction between these two groups.

ye, it would also be great if the scoreboard stats were available.
If you look at expanded statistics there are armor and mega pickups per player, PU time per player, lot's of important factors when rating a team game.

Just asked in quakelive:

chubz • sponge, SyncError|work question, how could someone get to expanded statistics that are covered on the scoreboard. ie. player PU time, items he picked up etc? in browser or over some exposed sservice?

@sponge • chubz: once that data goes away it's lost, it's not recorded anywhere atm
@sponge • there is a screenshot button on the new scoreboard to easily snap a pic, but that's all there is for now

for starters just rank ppl that manually want to rank themselves i.e. type !update tdm ischju or w/e - you get very few players in but at least you can figure out problems better.

I love this. Great idea. I know that you guys have a lot to do etc. If you have the time it would be lovely if you made this with a manual input for #tdmpickup (@QuakeNet) http://tdmpickup.eu

yea do it for #tdmpickup pls it should be very cool. We want it. We need it

one way to weed out abuse and get a more proper ranking is to use only maps that are in the competitive mappool, ie ruling out dm6 for example. obviously some 2v2 games would be lost that way too, but its not such a big loss i think, especially considering its supposed to be a 4v4 ladder?

it's a tdm ranking not a ladder as such.
Would be fun to branch it by type ... normal or 2v2 tdm, with ranking maps for each.

But I don't think it'll work unless quakelive releases those map statistic info hidden in the scoreboard.

item pickup stats per player with the given and taken amount of damage with overall frag net would really make an accurate math for tdm elo.

I updated the algo. Only tracks tdmpickup.eu map list.

http://www.qlranks.com/tdm/

How does the top 100 look now, in terms of placement?

Retribution isn't on tdmpickup's map list, never was :) (but it's a proper map, leave it) Same for Lost World.

I get that map restriction was to remove Terminal Heights abusers (namely Tiboca), but why is that even needed? The guy gets owned every time some semi-decent player is in the mood to join his server. I still get points from instagib duels on Thunderstruck in the normal QLRanks and there you don't care ;)

Top 100 still looks unimpressive (check where l1nkin sits for example), but at least there are more recognizable names there. At least it's better than before!

Site claims there are 1139 tdmers tracked so far, but I think less than 70 have rating above 1200. The split shouldn't be this drastic, right?

The people at the top have played more games.

l1nkin for example = 7 games tracked this week.. no enough to be considered for the top..

Is there a way to update the TDM ranks? It isn't automatically updating.

can I be a taster

http://www.qlranks.com/tdm/
seems to be down.