UnFreeZe Servers vsp players stats/rank

[adminless]

ok, I'm back with another good update here. first of all for the guys who asked about the stats algorithm, that was finally answered at this post here so there you have the full explanation behind the original vsp skill algorithm (nothing special as you can see) but note that that's just the original one, from now onward it has been (slightly) modified as the original one was way broken in many ways already.

then coming to the update this time, I think it's kind of a mayor addition, the other day while I was working on some database stuff at the new host came into the stats site and decided to patch the awards stats at the site that were broken from the start and that due schedules and different issues I had never got the chance to pull off so now they have been finally added to the site as you can see on the attached screens.

I believe they are a awesome addition and that they should definitively bring much more depth to the game than just a single artificial skill formula that no matter what it will never suit everybody and every single situation anyways and help make the game not just exclusively only about the top players, there's awards for all kind of players and all kind of skill levels as you can see.

so check them out, I feel they all should be pretty much self explanatory but in case someone have any sort of question or want some further explanations feel free to answer this post. and that's it here, I feel like this definitively should do it here at least for a good long time.

have fun guys

[adminless]

well this time reporting that just as final adjustments to the rank site I finally increased the cycle for the UnFreeZe rank from 10 days to 14 (i.e. two whole weeks) in order to homogenize it from day of the week to day of the week as not everybody has the chance of being active (or that active) in week days than on weekends so to avoid variances between start week ranks and end week ranks now every rank will include at least two weekends making it more consistent overall. in addition this will also bring a bunch more of casuals to the rank to complete that second page more or less which is never a bad thing either and so far with the modded formula elos seems consistent so far (peaking at 2400-2500). also for consistence did the same at the FreeFUn rank but here this time decreased the cycle from 75 days to 70 (i.e. ten whole weeks).

and as a bonus feature I added accuracy game to game standard deviation (i.e. in common words, how much your accuracy is statistically expect to vary from game to game usually) at the accuracy rank page to see how precise is everyone's (well that of the 100 first ranked guys) accuracy value across games which can also come quite handy to spot those casual cheaters (i.e. someone may have a average accuracy mean but a statistically abnormal variation). note about this value that it's not the absolute maximum variance (it's the root of the mean of the squares of it instead), it's just how much is expected to vary usually from game to game.

and nothing more by now, I believe that is that, I know that there's still some things left here and there about this rank site and that there's still definitively some stuff to polish and review but for now I think that this rank site finally turned out pretty solid and for everybody.

[tar]

Hi @admin,

there still are some irritating things:

1) What has a K:D weapon ratio to do with weapon accuracy?

Actually, the counts of kills and deaths are put into consideration for the weapon accuracy which is not very logical. In extreme, I could hit all of my rails, get no single kill and die of a rail shot. That would result in a 0% rail accuracy which is just nonsense. Instead, the real single weapon accuracy should be shown - but this is not possible at the moment, I guess?

2) Easiest preys (worst enemies)
Both lists are ordered by actually number of most kills (most deaths). Instead, the K:D ratio should be used here (which is already shown next to it - but it is not ordered by it).
EDIT: Perhaps there should also be used some formula that takes the kills (deaths) into consideration - as in extreme, 1:0 (and 0:1) counts would show a more or less meaningless list. A bit complex, here.

[adminless]

elo

1) "Actually, the counts of kills and deaths are put into consideration for the weapon accuracy" ??? no they are not, that column at the players profile I guess that you're asking me about, has nothing to do with accuracy, it just shows the so-called efficiency with the weapon (I mean, it writes eff at the header not acc).

2) "which is already shown next to it - but it is not ordered by it" ??? again, yes they are (or at least they should). that has already been fixed here (as can be seen on the attachment) so I don't know perhaps there's a problem with your browser, may be you could try cleaning its cache and double checking again if they are still unsorted because so far I believe that has been fixed the past month.

ps: it's curious you actually properly answered yourself the first question here in addition you also have the overall accuracy on the "General Stats" and combined at the accuracy award

[tar]

1. I see, so the overall accuracy is actually the only accuracy there. I also thought the "best killer" thing considers accuracy but it considers kills with weapon per game. Alright.

2. Ah, I forgot to allow Javascript. Now, it is ordered.

Thanks.

[fernandinho1337]

can you explain how things like best killer are calculated?

it says on the ranking page: Kills/(games+(37-1))

why is there a +37-1?
best regards

[adminless]

that's a good question there. that number, 37 in the example, correspond with the actual standard deviation of the number of played games among the ranked players (in simple words how many games normally vary from the average number of played games) and I believe that the original author added it to the formula in a simplistic (questionable to some extend) effort to normalize the results so they are more even and "fair" among all the data variance. like I said it's pretty much how originally it was there and that constant is also used on other places thus so far I think that it's not a very good idea to try to outsmart the original design right away. overall it gets the job done and prevent for example some guy playing just a few games or having a good streak and getting the award which I understand that it was the main goal there (who showed to consistently be the best killer overall through a large number of games).

[fernandinho1337]

yes thank you. makes sense.

best regards

[fernandinho1337]

Hello adminless,

I think I found a bug.

please see screenshot. the match on dm11 on december 5, 2018 ranks as one of my worst and one of my best games which does not make sense. what is put into considering the best and the worst matches? kills (best) and deaths (worst)?. shouldnt the kill:death ratio or something like that be used for both?

best regards

[adminless]

hehe that's another good spot there, yes, currently the criteria for the best and worst games are just absolute kills and death values. yes, probably a kill:death value would be more correct but it's also somehow a bit tricky (ex. one can join a game in the middle, or a small game, and have a nice streak, 10:3, and that would rank at the best game instead of a massive 70:30 game that could be more meaningful, so it's even arguable) so I'll see. anyways that's some of the things (among others too) scheduled to review on future revisions of the rank sites.

[hurrenson]

Howdy!

Just a quick question. Is there any reason why my super great meme nick "UR MOM GAY" isn't in the award stats anymore? It's still listed in the regular stats. My other current nick "NoShootImGril" is listed correctly. Bug, or did i miss something?

(Just as an unrelated side note: played with the new client and so far nothing to report. so far, so good.)

[adminless]

hello mate

mmmmmm current award game deviation value to entry the list is of 37 games unfortunately that nick (UR MOM GAY) only has 36, so play a bit more games with it and it will appear (or eventually if the input data normalizes a bit more it will appear anyways with even less games). that value (the game standard deviation) is used in that, a effort to normalize the values among the different players (and their consequent difference in number of games played).

as for the new version yes cool thank you for testing it out, also nobody complained about it yet so I guess that that means that it's good. this week I've added a new native dedicated single player (don't know if your tried the qagames libraries), yet even more fixes (I think today I finally finished with all the fixes on the book) and a bunch of security patches to also guard the files against rogue attacks (even if thanks to the people they were defeated, you never know when shit can get real again). so yeah go checking it out, I usually re-upload new builds frequently and by this week I hope I can finally bring everything together into a package and probably for next week after test it a bit more release the final definitive package (I mean I really hope so after so much time).

[MaRk05]

ich wünschte ich wüsste wie man das alles übersetzt haha . hab keine ahnung was hier so geschrieben wird

I wish I knew how to translate it all haha. have no idea what is written here

[adminless]

try here also you have a multilingual sub-forum here if you wanna talk on german.

probier hier auch du hast hier ein mehrsprachiges sub-forum wenn du auf deutsch reden willst.

[adminless]

hehe that's another good spot there, yes, currently the criteria for the best and worst games are just absolute kills and death values. yes, probably a kill:death value would be more correct but it's also somehow a bit tricky (ex. one can join a game in the middle, or a small game, and have a nice streak, 10:3, and that would rank at the best game instead of a massive 70:30 game that could be more meaningful, so it's even arguable) so I'll see. anyways that's some of the things (among others too) scheduled to review on future revisions of the rank sites.

ok, just addressed this (as well as several many other things) now. since now the best/worst games top 10 list properly display "meaningful" results sorted appropriately by "efficiency" (kill per death ratio) as it has always been intended since the beginning (just like the best/worst enemies). in addition this time I also went "all the way" to complete this year revision so you should definitively check these "new" virtually finished now rank sites out as there plenty of new stuff to check out like new (full) metrics/awards, several patches/fixes (to me there's really no longer any known "bug" of significance with the sites, anymore), a improved/updated/custom web server as well as a new non-trivial (finally) "adequate" for this gametype skill formula. to be honest so much stuff that it's not even possible to drop it all here as usual so you should probably just do better by drop at the sites and check it out for yourselves and then reply to this for any clarification you may want to ask.

so well that's it, unfortunately I got messed up with this rank thing as well as previously with the web server and I missed the always needed new client documentation update as well as its proper announcement but well I believe that these were critical things that consequently had a higher priority. I'll still oversee this for a couple of days and then I hope I can finally proceed with documenting the new release a bit and properly publish it within some days/a week.

• UnFreeZe stats: https://unfreeze.fpsclasico.de
• "FreeFUn" (aka UnFreeZe:27961/II) stats: https://freefun.fpsclasico.de