Bug reports
Please put your bug reports in this thread.
baSe |
vs
St0n3 |
Estoty Duel Tournament #120
| agent |
| 72 |
| QLSTATS Activate account |
| QLSTATS Bug reports |
| QLSTATS How it works? |
| QLSTATS Install QLstats on Server |
| QLSTATS Change CA Elo Calculation |
| QLSTATS Account deactivated |
| QLSTATS Reset my elo please |
baSe |
St0n3 |
St0n3 |
Purp1eF1sh |
PASITA |
Purp1eF1sh |
PASITA |
Artemka |
Purp1eF1sh |
klyb |
baSe |
St0n3 |
Artemka |
imijkeee |
klyb |
PepCo |
baSe |
Purp1eF1sh |
St0n3 |
PASITA |
how many games required for getting a ranking?
I use the term "Ranking" for the player list ordered by their "Rating", like a top 100 list.
This ranking list is only updated once a day at 12:00 CET.
Ratings however are updated after each match. On the match detail page you can find information about the rating status. Matches require a minimum number of players, similar play time in both teams, ... to be rated.
for example it says 46,566 players since start and they are only 2734 players ranked in CA that is the most popular game type. Even if the requirement is 5-10 games, the number seems small.
Sorry if I am wrong, but wanted to get it under your attention.
Out of the 46k players, only 14k have completed 10 matches combined over all game types (including unrated matches).
And 1/3 of that is plausible when to take away FFA-only, Duel-only, FT-only players and those who only play on customized servers
One hour comes in, one hour drops out.
At least you can tell if someone is just strafe-racing or using weapons.
How come?
link to this match: http://qlstats.net:8080/game/155639
It's possible that the match finished within the few seconds it takes to restart them.
I ran the rating code for all the "not rated yet" matches and they are rated now.
However, the rating appears out of order now. It used the last known rating (as of today) as input for the rating update.
I dont know if it s a bug or not. I have 2 server on port 27960 and 27961.
The first on 27960 work correctly, i see it on servers list.
i have an issue with the second on 27961 i dont find it on server list. Telnet on 27960 and 27961 are ok. Do you have an idea?
less likely are deactivated players or AFK players (with less than 500dmg or >10x more dmg received than dealt)
without knowing what games you are talking about, i can't really tell much
and some player have "Rank: - " under player information
a profil with rank:
http://qlstats.net:8080/player/43330#ca
without rank:
http://qlstats.net:8080/player/17726#ca
http://qlstats.net/game/284500
I switched teams during the match and in the final i had double -19 points .... that's hurts a lot :)
The rating change is calculated per player and match and cannot be broken down into parts for team red and blue.
If you played for both teams, you see your rating change twice, but it still only counts once.
I'll see what I can do to avoid showing it twice, which is indeed confusing. It will be random though for with team the change would be shown.
http://qlstats.net:8080/player/20581#ca
Thanks for the work you do.
http://qlstats.net/game/580301
Я_ХОЧУ_ЛОМÐТЬ won this game, and qlstats reports a loss
https://api.qlstats.net/api/jsons/2016-04-17/f31426f2-4d1e-4756-9f9e-30f70c76732e.json
In duel a quit is automatically a loss, even if it's in the last second and the player had a lead.
For example !a just before round starts and then complete entire map and still get a - next to your name.
Time spent on RED 00.00, BLUE 12.17. no ranking. Keep up good work!!
The rating can only be calculed for the whole match, not partial games when switching between red/blue/spec, therefore the rating change is only shown once in the results (wherever the player appears the first time) and not multiple times.
Another perfect example of a -28 elo drop when you agree to even teams right before round one starts.. 0dmg, 0 seconds played 0 0 0 0 :/ its weird man
You played in team blue and your Perf value of 35 is equal or worse than every player's in the red team.
That's why you lost points.
The fact that you switched and that there is a record with time=0 is purely a cosmetic problem.
And as I mentioned before, the rating change is only shown once (since there is only 1 change), which happens to be on that 0sec record. I agree it would look better to show your -28 in the line of team blue instead and "-" in the line for team red. Wouldn't change a thing about the -28 though.
shouldnt duel pql be b-rating too?
If server owners use the standard duel factory and then manually change cvars to turn it into pql, there is no way QLstats can know about that. If the server uses a different factory (which includes the PQL cvar), then QLstats will put it in the B-rating.
What server are you talking about?
I play in south Africa... but my region says Europe. How do i change it so it appears on africa region?
http://qlstats.net/player/31719#ca
it determined the region based on the servers you played on.
Just check this tdm games with Pref -200 he gets elo -2 -3
http://qlstats.net:8080/game/761082
http://qlstats.net:8080/game/761012
How it is possible if other players get elo -10 or more with Pref +100 and more ??
i'll need to take a closer look at that when i find some time.
unfortunately i'm quite busy atm.
I'm also from South Africa and mine also says EU. Would you be so kind as to help a brother out please?
http://qlstats.net/player/21857
Thanks man!
kreaper
I noticed that it counts rounds when you spec as actual playing time.
Look at this game http://qlstats.net/game/904146
I lost connection when it was 3-3 (so 6 rounds in total)
I was able to reconnect about the end 8-5
Spectated for 2 rounds until 10-5 and got huge drop on my rating as it thinks I played more than a half of the game (it says I played 8 rounds)
I think the problem is that it took maybe a minute or to for you to actually time out. During that "999"-time the server still listed you as being connected. Only after you timed out, the server sent a notification to qlstats that you left the game.
There isn't anything I can do about this, only the game servers knows about this.
It's probably just a coincidence that it was 2 rounds until you timed out and also 2 rounds that you specc'ed.
http://qlstats.net/game/937484
my ELO is comming down 19 points and the reason is my teammate "+! pengelephant" in that game? WTF guys... fix that pls...
here my "opponent" shown:
http://qlstats.net/player/37077#overall
http://qlstats.net/game/939665
And other times i win and still lose glicko also to players with much higher glicko than myself
http://qlstats.net/game/948521
How are you supposed to climb if winning vs better players doesn't do anything and winning to players on your same "glicko" has the same effect ?
-Dexxa
seems like my first rated game got me good
Could you revert the elo back ?
Your average CA rail acc over the last 100 matches is 24%, your LG acc is 17%.
Most players have higher acc than that and perform better in matches. Therefore most players are rated higher than you, wich means that your rating of about 1250 is actually quite accurate, when 1500 would be the average.
If you have missed it most of my "enemies" which i get compared to have 1.8 - 2.0 k rating no matter how you look at it the system is flawed
http://puu.sh/pXUa9/49eb97dc3f.png prime example
i dealt more damage had more kills less deaths and same play lenght
yet i get awarded less, intentional right?
the screenshot you posted is understandably confusing. to understand it, you have to know that the game's SCORE includes real damage received by opponents, while the DAMAGE_DEALT value returned by QL includes excessive damage (100dmg for a 1hp rocket hit, 80dmg for a 1hp rail hit, ...). So while it may seem that you dealt more damage, you might actually have dealt less damage that mattered.
qlstats subtracts 0.75*frags from the SCORE value so that only 0.25*frags go into the PERF value. That is to reward damage dealers and not frag collectors.
so yes... the system has flaws. but there is a lot more logic involved than what meets the eye.
You should not be losing any points to anyone if your Perf is 5 points higher (+/- 5 points is treated as a draw in FFA).
Players with a small number of games have a high RD value, which means their own rating adjusts a lot but only have little effect on other players' ratings.
The problem is, I currently don't have the time to work on qlstats.
In this game http://qlstats.net/game/980576
I have about 400 dmg and 10 frags more than the closest opponent and still we have around the same performance? Why fragging is considered a bad thing in CA at least?
I subtract 0.75 * frags so that only 0.25*frags go into the performance value. This should motivate people to do damage instead of letting their team mates do the dirty work and then get the cheap frags at the end.
"capped_damage" means that when you hit a 1hp guy with a rail you only get 1dmg. The other damage numbers in the stats are all uncapped, meaning they do give you 80dmg for a 1hp rail hit. So I had to use SCORE as the basis and work my way backwards to the desired result.
In my opinion, which you may or may not value, game rules should not be altered in any way under anyone's personal vision of the game.
The game's SCORE formula was not designed to be used for a rating system. It's a simple formula, designed so everyone who plays can understand it.
The way qlstats uses individual player performance for a rating update adds a couple more requirements for a "good formula". qlstats is often criticised to "destroy teamplay" because individual performance matters for the rating. And the more you value "frags" the stronger this effect becomes. Using only 0.25*frags was a compromise made by several people, not just me.
When joining a game 5 rounds late you catch up with the 2nd on scoreboard who has loads more rounds than you you still get punished for not being first on scoreboard no matter how much you caught up
Also what is the point of maximal 2 people gaining Elo and the 2nd one maybe getting a whopping 2 elo (WOW!)
It seems to be pure luck whether you get alot of GLICKO while being #1 or not
or the fact that you're compared to players with 800+ elo (big surprise you can't beat him who would have thought?)
If you try to pull something so complex off maybe you should add some variables that actually make sense and not just blindly go for Damage and Kills nor should you be compared to the highest elo player on the server
Some people say this was an upgrade to the old system
Making a half assed "complex" system is hardly an upgrade
i can shut it down and you're back at a completely random shuffle. that was the situation before i started the site.
old qlranks is no more and the all-knowing super-accurate system that you dream of will never exist.
- being on the worse team
- not having 2k elo and playing like it
- winning and still losing elo
Is THE alternative to the former System?
I'm saying that the current system needs to be adjustet if you play decently vs people way above your rating you shouldn't get a loss in ELO. Neither should there only be 1 player gaining elo at a time
The algorithm basically choses the best player, he gets elo , 2nd place might get a few points, that's it everyone is losing elo not because they play worse than usual but because maybe the team fucks up or you're simply unlucky yet you get punished like you're terribad
I've had games where my "opponent" was some guy with 2.1k elo, well long story short I lost elo, not because I played bad but because he has 2.1k elo lol
He was on my team as well which means I had even less opportunity to deal more damage as my teammate basically played REALLY well
these are things you cannot influence
This System MAY work if there were alot more players and alot more Servers for lower level players of certain skill levels but as it is in every game you face players above your ELO, you can do well you lose elo you can do bad and you lose some more elo
I know that perfectly balancing out the system will never work with so many unknown factors such as teamplay ect. but as it is the system just takes ELO if you manage to outscore your 1.8k elo players or your "opponent" by the time you manage to have a decent team / players you have lost what you have earned from the previous time
Maybe adjust elo gain / loss a little and add a penalty if your enemy is some 2k elo monster
You can be the 2nd best player in the world, but if you only ever play against the best player (and nobody else), your rating will be worse than some average Joe's who occationally outperformes some other average Joe.
That's why playing on servers with skill limits biases the whole rating system.
Your PERF value (calculated in CA as real_damage/100 + 0.25*kills) is pairwise compared to each of your opponents'. If yours is higher, you gain, if yours is lower, you lose. The actual amount depends on the prior ratings and RD values. This is summed up for your overall match rating update.
In the past qlstats also compared you to your team mates, but that has been changed months ago.
http://qlstats.net/game/992147
http://qlstats.net/game/992215
Oh look I gained even tho I couldn't do much because the 2k guy went to fucking work, pretty random no?
please shake off your obsession with these numbers when you don't even know how they are calculated, what they mean and how they can be interpreted.
as a rule of thumb... pretty much all "common sense" assumptions you may have are irrelevant. glicko was developed by a university professor for statistics who's also a chess geek. do you seriously think you can come up with something smarter?
The matches you pointed out are perfectly fine. Your perf is within the tie-boundaries with your weakest opponents. but since they had a higher rating, you gain a few points from them. but you lose to that other opponent who had a better perf than you. i don't know WHAT you expect, but in this case the system works exactly as it should.
This rating may work for Duel / Chess but for CA? Apparently not
Also "the common sense assumptions" aren't only expressed by me go to X ca server everyone will tell you the fucking same
Maybe YOU should come up with something smarter seeing as it doesn't work for shit no matter how many times you tell me how the goddamn rating works when most of the time it doesn't, on paper your rating system works nice and fine except when you actually see how it turns out for a team mode
Also losing almost 500 "GLICKO" doesn't seem to bother you maybe another feature?
freeman_w here. Before everything else, thanks for your efforts buddy.
I already posted a few postings before, pointing to the problem where someone loses ELO/CLICKO against a teammate but you didn't answer on that. Anyway, not as important atm.
I agree on some parts with the Anonymous person on this actual conversation. I also think that the system needs a bit of twist for the team-modes. But I wanted to make another suggestion. Could you describe in your frontpage somewhere, HOW you exactly make the calculations? There should be a link that shows how you do that on a player-view aspekt.
For example, you could use P1 "the actual player" (P2, P3 etc for teammates if need), OP1 (OP2, OP3 etc for opponents) and then explain whatever else is being needed for your calculations (like "perf", maybe P1_oldElo, P1_newElo, DMG-D, DMG-T and so on). That would be a nice thing to see in general for the site.
And I also have a question. Why does the "Recent Games" page, sometimes show one opponent/teammate, and sometimes more than that? If you are being measured against all players that you hit, then there should be entries for all of them and maybe a MATCH entry for the whole match (another suggestion here).
I have my most recent match here,
http://qlstats.net/game/1016199
where I only get compared to the best guy (and i expectedly lose points, even though he is 600 ELOs away). I don't think that I only played against him, so I would like to c how I did against the others. And speaking about this match, some other questions maybe:
1) why is erk losing so many points on my team? he does that against Flugsvamp, who has the same ELO (2016). OK, he has a Netto of -1 but he has more DMG-D than DMG-T? (You say that you try to reward people who make more damage rather than the frag hunters). How is someone supposed to play in order to get better? Make more kills? Make more damage? As long as the calculation is not transparent enough, you will have questions here.
2) Why is Zilda gaining so much elo compared to the TEAMMATE (what is this kind of comparison anyway?)??? Zilda has a Netto of -5 and about same DMG-D with DMG-T. There are matches where I win with my team, I have positive netto, more DMG-D than DMG-T and i still lose points... Here is when I say WTF??? WWWWWTTTTFFFFF?? (I don't have now a link for that, but believe me, it happens). So you should understand, why some people are not so happy with that system... It is not transparent enough. That's why you should maybe try to make this explanation page of how the thing is being calculated. It would probably also result to less questions here. And also, on the "Recent Games" page, there should be a complete listing of all players involved, and preferably a concluding MATCH row, as an accumulation of the individual other rows. Or even better (because the above would make the "Recent Games" page a mess) ONLY ONE row for every match, and the individual comparisons within the "View" page. (In that case, you wouldn't need the "Opponent" column under "Recent Games")
Just some suggestions that would help all players (I guess) understand better what is happening. Thank you for your time man... Stats make a game more interesting... AND understandable stats make it even more :)
Have a nice day
your perf value of 36 is compared to the perf of all your opponents (not any of your team mates). there are only 2 opponents who had a lower perf than yours but they also had a significantly lower rating than you, so it was expected that you outperform them, thus you gained only a handful points from them. But at the same time other opponents who had a significantly lower rating then you outperformed you, which causes you to lose a lot of points to them. You also lose a few points to the higher rated opponents, but that was expected, so it's only a few.
Zilda gained a lot of points because he had a low rating (~1300) and outperformed opponents with a significantly higher rating (1500, 1400)
erk lost his points mostly to bnkrz, who had a lower rating but got a higher perf.
When you hover your mouse over the Perf column name, you see the forumula how the Perf is calculated. The qlstats source code is also available on github, if you want to know it in full detail.
The rating process is form of statistical learning... it will never be 100% accurate. Sometimes you lose too many points, sometimes you gain too many. Over the long run the rating should give a good estimate, nothing more.
On my site there is one match being shown "FOUR" (4!) times, always with the same "opponent". What is the idea behind that? The match is from the "2016-07-29 17:58"... Isn't one time enough? Is it a bug?
My profile:
http://qlstats.net/player/37077#overall
It's kind of a bug, but one that can't be fixed due to the way the database query and the paging system works. Eliminating the duplicates would require different database queries which cause a lot more CPU load.
Since this doesn't happen too frequently, I'll let it slip.
This is the prime example why it doesn't work, I get matched up vs Warlock i do very well, he does very well too and gets 83 points for being matched against ME 1.8k elo vs 1.2k elo and he gets a fuckton of points for outscoring me while I gain a mini ammount of 22 for playing on practically the same level (wow), you're still telling me you can't see the problem?
Also check warlock's first 3 games. Apparently gaining 580 Glicko in the first 4 games seems to be pretty normal, just as normal as losing 500 on the first game.
you are compared to warlock, who had a significantly larger rating and higher perf, so you lost a few points to him. you are compared to baky, which is considered a draw and you gained points because you had a lower rating. you are compared to kottmur and gained a lot of points because he had a higher rating. you also gain points from kiryu.
so, aside from one player getting too many points (which is a known characteristic of Glicko but not a severe flaw, since it will correct itself) .. this match is rated as expected.
i'm done wasting time on this subject. i recommend you stop looking at the stats and be happy about your low rating, because it means that you will get stronger team mates.
The next map was Trinity. Went all the way to the end but a bunch of guys were losing and started dropping out towards the end.
Does that affect whether or not a game is recorded?
Just curious!
Thanks for your work, it's helpful to review the stats, accuracies, etc after the games. very nice.
I don't know what's causing it and unfortunately I don't have the time atm to look into it. Whenever I find the time to continue working on qlstats, this is the first thing I will look at.
How did I lose ELO?
Why did I lose ELO in this game?
Found maybe a bug I guess? Look at http://qlstats.net/game/1323246
Your Perf formula is "Score - 0.75*kills", so perf SHOULD NEVER (in normal cirmumstances) be higher that the score? Then pls explain why sacrum on my team and Bojanjes in the enemy team have a higher Perf than their score... WTF? o_O
Real formula is (score-0.75*kills)*timefactor. Timefactor = (match duration) / (player time). For your result timefactor = 1
>sacrum on my team and Bojanjes in the enemy team have a higher Perf than their score
Timefactor for them is more than 1.
http://qlstats.net:8080/server/303
Image of result: http://imgur.com/jR8BYty
Link to match: http://qlstats.net/game/1473092
But I guess this is nothing that can be fixed. Just reporting :D
please take a look:
http://qlstats.net/game/1490252
server got offline for some reason after 1 minute or something, but this match got ranked.
http://qlstats.net/game/1484812
Cheers,
niewi
The sorting of the players is incorrect and I think the calculated Glicko Change isn't correct either.
The table is sorted by Perf, which is Score adjusted by the time played.
Cheers,
niewi
thx
link
and even in this one with all positive paramethers kdr ,damagr d/t ... again negativ points
link
fr0ggy
Cannot add server on any panels
http://qlstats.net:8084/servers.html
It says panel is full for each.
Thanks in advance :)
-=/
not that it matters just informing seen section of the website.
My name is Cow. It says I have played 15 rounds, died 0 times and only have 2 kills. In reality I did play 15 rounds but had way more damage/kills/deaths. It made me lose 46 elo, which is unfortunate. I think it's caused by a bug. Just letting you know and I hope there is something you can do about it.
This is the raw data received from the QL server. All it contains about you is 2 kills and 0 deaths.
Can an admin have a look at this game? http://qlstats.net/game/2452008
My name is Cow. It says I have played 15 rounds, died 0 times and only have 2 kills. In reality I did play 15 rounds but had way more damage/kills/deaths. It made me lose 46 elo, which is unfortunate. I think it's caused by a bug. Just letting you know and I hope there is something you can do about it.
Ej
http://www.qlstats.net/game/2581172
http://www.qlstats.net/game/2562001
Studying the stats for some recent games it seems like Qlstats is giving people on the winning team a boost to their performance rating compared to people who played equally well on the losing team.
i can win or loose the game with highest elo, but elo change is only -
http://qlstats.net/player/12622/games?page=64 - error
http://qlstats.net/player/12622/games?page=75 - error
http://qlstats.net/player/12622/games?page=99 - error
I suppose it happens due to unsupported chars in nickname of an enemy
not changing...
see the pm_your_tits score of -75 with 1 round played, no team swaps, nothing.
https://qlstats.net/game/5276004
i played 3 rounds yet it says I played for the entire duration of the game. damage taken 900 damage, yet i somehow played 18 rounds.
Is there also a form for feature request somewhere or a topic?
My idea is:
Make an extra ranking with players that have been active in the last half year.
This way i can see my real ranking in the active community and not have to 'contend' with people that have retired. I am hoping to get higher than (top 74.29%) -don't laugh- but it also may go down, but it will be a more honest representation i feel.
It would only have to say 1 little thing extra (top 74.29%, ½yr 14.91%))
It's just about the percentages, not the absolutes.
I would figure your code would be identical except for the selection criteria being 'has played within last 6 months'.
Oh yes, this is all vor CA.
Don't think it matters though, should be same value to it for all game forms.
Thank you for this nice website btw.
As handicapped as balancing on ELO can be sometimes for balance-bots it is realy helping a lot for balancing CA games. I think the CA scene was saved back when you made this. SO thank you for that.
Is there any answer to the question, why I lose/gain ELO/GLICKO to my teammates? It was also posted before but I didn't find any answer to that yet. I almost only play CA atm and the last games in the overview show drops/gains of ELO to teammates?!? Don't know how far this goes behind but I am talking about the games 2020-07-08 00:25 to 2020-07-17 01:55... Almost all the loses (and gains btw) are done to teammates... What is wrong here? If this is a bug then it seems to be long there (as I said there is another post here mentioning that, posted 4 YEARS ago)... Any explanation is appreciated.
Greetings