Small update on the ladder site

As suggested here, maps can be downloaded from ladder website in one zip-archive. By the way, there are 631 maps for now. Of course, you can upload your (or others) maps, which are not yet in the archive.

Moreover, ladder now tracks not only personal stats, but whole game reports. It's not visible for now, but soon will be added:

• games list by server (how much people played, who they was, match results)
  
• own games records (maybe, one day, graphical representation too)
  
• number of wins (when player was in winning team, wasn't has negative ratio or captured at least one flag) will be counted in player ratio calculation
  
• clan section, clan rating and CMs history will be opened (match will be considered as CM and counted, if teams was equal, and each of teams was consisted of one clan players)
  

In other words, that's one small step and one giant leap

By the way, I'm accepting ideas for improvements. Don't know when I'll do it all, though

Love, keta.

Just a quick question, every how often does the server send info to the site, and exactly which variables does it report?

Game server sends report at the end of battle. Sent data looks like this (comments are mine):

Code:


<bv2gamereport>
   <game>
      <gameinfo>
         <servername>Babo Violent 2 Server</servername>
         <gamemode>CTF</gamemode>
         <map>CTF-Slant</map>
         <matchcode></matchcode>
         <matchmode>0</matchmode>
      </gameinfo>
      <teams>
         <team>
            <teamid>1</teamid>
            <score>0</score>
         </team>
         <team>
            <teamid>0</teamid>
            <score>0</score>
         </team>
      </teams>
      <players>
         <player>
            <playerid>1</playerid> <!-- Ladder ID -->
            <name>Unnamed Babo</name> <!-- In-game nickname, with colors -->
            <teamid>0</teamid> <!-- Ladder ID -->
            <time>32.43333</time> <!-- Time on field, sec -->
            <kills>22</kills> <!-- ... -->
            <deaths>11</deaths> <!-- ... -->
            <damage>27.55</damage> <!-- ... -->
            <caps>0</caps> <!-- Captured flags -->
            <attempts>0</attempts> <!-- Taken, but lost flags -->
            <returns>0</returns> <!-- Returned flags -->
         </player>
         <player>
         ...
         </player>
      </players>
   </game>
</bv2gamereport>


Well, one statistic I was always curious about is damage per second, but is the time you stay dead counted into time on field? If yes, then formula would be: dps = damage/(time-deaths*5)

It could also take a screenshot at the end of every CM and post it on the clans' forums. It cannot be hard to do, just something like

Code:

<when cm over>
      <"p"take screenshot>
        <upload on <preset clan#1 forum link<http://forum.com/CM section>
              <if>
                    <team won >> post in <win subsection>
                    <team lost >> post in <loss subsection>
                    <team tied >> post in <tie subsection>
  <///>

Cinc wrote:is the time you stay dead counted into time on field?

No. Only active time is counted. Spec time isn't counted too.

About screenshots... It actually IS hard to do. I have no access to game sources.

ketamine wrote:....... I have no access to game sources.

but we still love you !!

Nice to know, SettoreX I love you all too (NOT GAY)

Lol Meme xD

gj keta, thank you!

Another thing I was curious about, is it possible that instead of current formula for rankings which is based solely on your total stats, you get points based on the result of the match (add points for winning and lose points for losing) and your stats?

For example, you're in the winning team with awesome stats, you get more points than being in the winning team but with some -50 score, and if you are in the losing team but with 50-0 and 3 caps you get almost no points deducted etc. Everything seems easy to do except for checking the winning team (maybe that's already implemented?). Your stats are then calculated directly after the match. Also, the amount of points could also be dependent on the ranking of people in the enemy team. That's more like a real definition of the ladder

aarrgh... imagine a team balance at the end of the mach .. lol !

maybe...
in a DM server kills/time counts as positive and death/time count as negative, otherwise in a CTF server what count are caps/time (for attackers), returns/time(for defenders), damage/time (mostly for snipers) as positive and attempts/time count as negative (for nabs or heroes).

that can work for a single-babo ranking....for teams... well thats's another story.

Cinc, formula will be something like I've already described once, but buch more complicated (you may use google translate to read whole thread). Since I startet collecting more stats than before, more aspects will be counted. Everything is already invented, I just need some time.

And ladder will not be "ladder" anymore. It will become something more than just penis measurement tool.

And no, ranting will not be "calculated directly after the match", because formula depends on stats of all people in the ladder. Too much load on webserver just to show how much longer your penis became (non-empty reports arriving every 2-3 mins).

If you're not satisfyed with my formula, show me yours approach. Not "it will be better if... and if... and then...", but actual X & Y's, with results of calculations in a nice spreadsheet. I can give you actual data to play with, just ask for it.

BTW, only CTF matches are have effect on current rating. DM's are not, but will. One day.

UPDATE. And, yes, team (clan) rating formula will be totally different.

Code:

Rating = ( (Kills / Deaths) + (Damage / Deaths) + (Damage / Kills) + (Deaths / Kills * -1) + (Kills / Time * 20) + (Captures / Attempts * (Captures / Total_Captures) * 8000 ) + (Damage / Total_Damage * 200 ) + (Kills / Total_Kills * 2000 ) + (Deaths / Total_Deaths * -1 * 200 ) + ((Time / total_time * 120 ) ^ 2 * -1) + (Returns / Total_Returns * 100 ) + ((Attempts - Caps) / (Total_Attempts / Total_Caps) * -1 * 8000 ) ) * 100

I see multiple problems with division by zero, and too much bias to successful flag attempts ^^

Unless I misunderstood, and Kills is the total kills one player has, and Total Kills is the kills all players have together etc

power=K-De+Dm
rage=(K-time/60)+power
hero=((caps+returns)/time/60)*10
smart=(hero-(attempt/time/60))
Ladder = power+rage+hero+smart

i'm lost

:edited... by time i meant minutes

Cinc wrote:I see multiple problems with division by zero, and too much bias to successful flag attempts ^^

No division by zero. For now, you'll start to be rated when you die 50 times. Enough to fill all the values. And, this is simplified formula, as I noted before. It shouldn't be calculated in one single step, and it's easy to add division by zero checks. Meh.

"Too much bias to flag attempts" can be lowered by tweaking coefficients. I know, pr0's more aware about Kills and Damage, but, in my opinion (which will be decisive, haha) this game is not about only Kills and Damage. It's about team success. «Team success» is complex result, so kills, captures, returns, etc should be taken into account. Look closer at the formula, final rating is based on the sum of sub-ratings (you may not capture flags at all, but have many returns), and some of sub-ratings may even be negative (if you're good at doing damage, but have much lower kills count - you'll be far from top).

Settorex, your formula isn't taking into account «total players skill» and & «time spent», again. Rating calculated via my formula is degrading. You cannot sit scared in some corner with your snipe dealing 15k-5d score, having nice 3:1 ratio, you should kill more and more, or you'll never reach top, because other players kill more than you. And you cannot make 100500 kills and sit on top forever. While people play, your rating gets lower and lower, and you'll fall from your sky pretty soon if you'll relax.

ketamine wrote:.......
Settorex, your formula isn't taking into account «total players skill» and & «time spent», again. Rating calculated via my formula is degrading. You cannot sit scared in some corner with your snipe dealing 15k-5d score, having nice 3:1 ratio, you should kill more and more, or you'll never reach top, because other players kill more than you. And you cannot make 100500 kills and sit on top forever. While people play, your rating gets lower and lower, and you'll fall from your sky pretty soon if you'll relax.

the rage value will drop your stat down if you spend your time fooling around.
and the hero value will not make you grow that fast if you play DM instead of CTF.
if you are awesome the smart factor can double your hero beeing

but, ye this miss the fact that is far unbalanced if you join a server with lots of noobs.
got to consider other players score too.

a = Skill rating: 1000*(1 - 2^-((kills*damage + (time/30)^2)/deaths^2))

b =Teamplay rating: 1000*(1 - 2^-((4*returns*caps + (time/30)^2)/attempts^2))


Total rating: Geometric mean of a and b: R = sqrt(a*b), theoretically 1000 is the max rating but I expect the players to have maximum somewhere around 900.

What this means: Someone who focuses entirely on being self centered and has a 1000 skill rating but teamplay rating 100, will have less total rating than someone with skill rating 550 and teamplay rating 550 (AM-GM inequality). Rating distribution should be even, but I'd like those stats in some .xls file if possible so I can test the formula.

Also, time spent on the field doesn't have much influence (if i got it right, somewhere around 1/4 of the exponent, about 60 to 30 of total rating points) but it's still enough to make a difference.

Meh. Please-please-please, read thread on Russian community forum I linked before. Use Google Translator. I was tired of explaining mathematical statistics past time and can't do it again.

Other notable thing is rating should grow infinitely, without «max rating»: the more you fighting, the higher it gets, it's psychological thing.

Actual stats in Excel format are here, tune your formula to make ID 60 to be #1, ha ha.

Please, try to be abstract from your side of view, be objective, ladder is not only for people you calling pr0, stats are counted not only on RS', even if you think it is. This world is for everyone. There will never be '$∪℘∀-ρℜ∅-rating', at least while I'm alive here.

Lol you're totally missing my point. Don't underestimate my knowledge of statistics and mathematics, and besides, I went over the thread on the russian forums and nobody posted a similar formula to mine.

My formula has a more even distribution than the current one, and I don't think that stats should go on to infinity, simply to make more variations happen, instead of cementing top 100 after a month.

Anyway back to the formula, first i tried to stick close to the gaussian normal distribution, but what i came up with is derived from the equation for voltage in a DC circuit with a capacitor, but with the factor 2 instead of e to give more variety at higher ranks. Also, geometric mean ensures what you said, that "pro players" don't get all the high ranks, because you can't spend all time staying in def having 100-1 score while getting no flags at all. Also, it requires you to maintain a ratio for a long period of time or you'll fall in ranking. This will make the list more dynamic instead of being just "play more and you'll be ranked higher!".

Anyway, tomorrow morning I'll post a spreadsheet with all the calculations for current top [random number] players.

=======================================================================
EDIT: Calculated new ranking for current top 20 players, with some interesting results, probably because time playing made too little influence, which means the formula needs to be tweaked a bit more. I'll write like this: Previous place: player id, new place, reason why

20: id 16227, new rank 87, Average playing time, average ratio, moderately unbalanced skill and teamplay
19: id 6346, new rank 344, Average playing time, average ratio, many failed attempts, unbalanced skill and teamplay
18: id 403, new rank 12, Average playing time, average ratio, very balanced skill and teamplay
17: id 69, new rank 55, Average playing time, average ratio, moderately unbalanced skill and teamplay
16: id 24, new rank 167, Average playing time, average-low ratio, moderately unbalanced skill and teamplay
15: id 532, new rank 20, Average playing time, average ratio, moderately balanced skill and teamplay
14: id 199, new rank 26, Average playing time, average ratio, moderately balanced skill and teamplay
13: id 6803, new rank 161, Average playing time, average-low ratio, moderately unbalanced skill and teamplay
12: id 20010, new rank 504, Average playing time, low ratio, many failed attempts, very unbalanced skill and teamplay
11: id 13179, new rank 949, Average-high playing time, low ratio, many failed attempts, very unbalanced skill and teamplay
10: id 142, new rank 318, Average-high playing time, average-low ratio, many failed attempts, moderately balanced skill and teamplay
9: id 744, new rank 82, Average playing time, high ratio, very unbalanced skill and teamplay
8: id 5552, new rank 188, Average-high playing time, average ratio, many failed attempts, unbalanced skill and teamplay
7: id 4154, new rank 146, Average-high playing time, average ratio, many failed attempts, unbalaced skill and teamplay
6: id 15, new rank 47, High playing time, average-high ratio, unbalanced skill and teamplay
5: id 118, new rank 25, Average-high playing time, average-high ratio, moderately unbalanced skill and teamplay
4: id 11099, new rank 143, High playing time, average-high ratio, very unbalanced skill and teamplay
3: id 88, new rank 308, High playing time, average ratio, many failed attempts, very unbalanced skill and teamplay
2: id 919, new rank 58, High playing time, average-high ratio, unbalanced skill and teamplay
1: id 63, new rank 18, High playing time, average-high ratio, moderately balanced skill and teamplay


Now, to not make it that I'm biased, my ranking:

id 60, old rank 583, new rank 286, low playing time, low ratio, very balanced skill and teamplay

And the new top 3 would be:

3: id 11895, old rank 194, Low playing time, high ratio, very balanced skill and teamplay
2: id 3683, old rank 213, Low playing time, average-high ratio, extremely balanced skill and teamplay
1: id 8876, old rank 124, Low playing time, high ratio, very balanced skill and teamplay

Although, around 80 percent of the current top 250 remained top 250, but in different order. A bit of more tweaking and my formula is done. Gonna post it tomorrow morning like I said I would.

Just final note on rating calculation. It should avoid use of exponentiation or square roots, especially at the final steps. First, misuse of such calculations, like your "R = sqrt(a*b)", make changes on the bottom of ladder very slow, turning it into a swamp and making attempts to step higher boring; while the top of the rating changes fast as hell, making virtually impossible to gain a foothold at the top. Basically, top 25 rows will be just remixed with each update. Second, after player will step over some critical line, he will start climb up faster, much faster, which will give him a false sense of growth in the skills, and this will have negative impact on real look at things.

Finally, I am pleased to continue the discussion by e-mail (ketamine@warlabs.ru), because, apparently, quite few people are interested in this subject (and far fewer of them can say something sensible), and, frankly, we have fallen into offtopic. Also, I'm not happy to make a choice of formula democratic. Game design is serious shit, and ranking systems is most difficult and cryptic part of it. It's not as simple as changing textures or switching config variables, heh.

Check the edit on my previous post

I am locking this.