SGRstats evolved

Originally Posted by realazthat

What would offset this is you playing a class that is more difficult to win with, such as soldier. People who are Scientist would have a higher W/L ratio. The solution is to have an adjusted W/L ratio, that is rated per class, relative to other players. This takes away any advantage one class would have. You surely have a higher adjusted W/L ratio (ie. you are a better _soldier_ than others, and that would count much more heavily since you are mostly soldier).

Maybe I'm confused, but I was under the impression that the W/L ratio was a team thing and had nothing to do with the individual's performance in the round.

If I come in with 10 seconds to go and my team wins, does that mean I'm a better team player then the other guy who came in with 10 seconds to go on the losing team?

Rather then W/L ratio, it should be based on the players score at the end of the round regardless of Wins or Losses. There have been many times where I've been at the top or near the top of the server, but still been on the losing team. There have been many times where I've been near the bottom too, but on the winning team.

I for one would like the win/loss to be slightly factored... not becuase I am near the top of course...

apprently I need 5 minutes of play to upgrade my sci. Love that feature.

Awesome job Realz!

Originally Posted by Arcadenut

Maybe I'm confused, but I was under the impression that the W/L ratio was a team thing and had nothing to do with the individual's performance in the round.

You influence whether your team wins; if you are a good healer for example, you might end up with a horrible K/D, but having you on a team will make it X more likely to win. X is W/L.

If I come in with 10 seconds to go and my team wins, does that mean I'm a better team player then the other guy who came in with 10 seconds to go on the losing team?

Yes there are individual cases that are wrong, but over a large number of games they become insignificant. Also, other people experience the same phenomenon, and their W/L Ratio should reflect that.

Rather then W/L ratio, it should be based on the players score at the end of the round regardless of Wins or Losses. There have been many times where I've been at the top or near the top of the server, but still been on the losing team. There have been many times where I've been near the bottom too, but on the winning team.

Yes, obviously it might be skewed in that someone might have a high W/L Ratio and be a "Healer Moron" etc. ie. not a team player. However, in the absence of a good K/D, W/L would indicate that the player helps the team to win, and is a good player, but sacrifices kills to win the game.

Yeah, you can't be picky when it comes to stats. Averaging everything is the best way. Or some form of statistical analysis. Run t-tests, or z-tests, or Correlations, Multiple Regression analysis. W/e you do. In the long run, there will always be outliers, and things that don't fit nicely within your data. In a situation like this where there are so many factors involved. I would suggest a Multiple regression analysis. I don't know if browsers can run this though, lol.

Multiple regression analysis (in a very brief nutshell) is like equations in math class.

Variables:
a=scientist, b=goa'uld, c=soldier, d=jaffa, e=commando, f=ashrak

equation: .34a+2b+4.32c-2.2d+1e+1f= (ranking stat)

After taking all your data and throwing it into a program like stat crunch for example (there are free online statistical programs, i have a free link to stat crunch if needed). You will want to pull the statistics from players in SGR, you don't want it to be too skewed one way or the other (either, the people that play SGR like they breathe, or the people that played it once and quit).

Best way to do this in statistics? Same way in every thing, random sample (best to be from 30-100, above 100 you're results won't change much but it will become a mess, some will say this number is 200, its just preference). So, taking this sample you want to run some tests on it to find where your mean is, hopefully you have a normal sample (bell shaped curve). So, (skipping some steps) you find out how much one variable effects another, negatively or not. Which would probably result in no correlation as these factors are all independent. If you read this, good job. But I know that all these mumbo jumbo things with K/D is more important, Shots fired is more important, XP/time is best! isn't the way to go.

If someone plays all sci, their shots fired and accuracy will be too high compared to the rest. Scrap that

K/D, sometimes you have good team players who aren't afraid to die, other times you have people that can't live, and other times you have people that don't care about team work, all they care about is their K/D.

XP, some say this punishes new guys or those who have a life. which can be true. So you throw in the time factor. Well if you earn 1000 XP for every 1 hour you play, but are clearly a better player than someone that earns 1000 XP in 1/2 an hour (perhaps they only play KOTH and rack up the XP).

So the dilemma arises. But, wait perhaps W/L is best. But then you run into the problem that this is an individual ranking, not team ranking, especially because your team is never the same.

Then people say "Put them all together". But which do you give the most weight to? Or maybe all the same weight? But thing is that they all have extreme pitfalls. Which really means you should throw them out the window as your confidence interval is sooo low because you can't confidently say that any one of these represents the true best player. Let alone throwing them all together.

So what I was saying about multiple reg analysis earlier may be true for these factors as they all do effect one another. If you are a better team player your K/D may not be as high. So you give more weight to K/D than W/L then you look at XP. And you think how much does XP weigh in on the skill of a player? And time, does someones time devoted have anything to do with how good they are?

So, take this, run your Multiple reg analysis with your factors and it will tell you how much weight to give to any one variable. Plug that in your PHP and you're set.

But, don't accept any equation given here. As they will just be something that is simple division or multiplication and doesn't give weight to any one variable over the other, and if it does, it doesn't give the correct amount. Saying K/D is 2x more important is like guessing who will win the NBA finals, you just don't know. Best way to do it is through a study to figure out which equation (or team) is the best fit.

And if you are not up to devoting the time, or finding someone who is, to having a proper formula, then just go with XP.

i wanna be tracked too please

My ID 13577

Thx, that's a really really nice tool indeed !!

hey, nice work dude what about that realz's secret sauce lol ? what's the meaning ???

It has even improved! I love you for that! It's great! Could you please half my playing time? Then I'd absolutely marry u :*

Originally Posted by roulious

i wanna be tracked too please

My ID 13577

Thx, that's a really really nice tool indeed !!

Your already on there man .

Originally Posted by roulious

hey, nice work dude what about that realz's secret sauce lol ? what's the meaning ???

Its a ranking system that I am experimenting with. It involves KDR and XP, and combining them to make them fair accross classes, so that someone playing Sci heavily won't weigh more than someone playing Soldier heavily. If you want to see the math:

http://forums.stargateworlds.com/sho...5954&page=3#63

Originally Posted by Siba

It has even improved! I love you for that! It's great! Could you please half my playing time? Then I'd absolutely marry u :*

I could halve your time, but it would just be on my site lol. Why are you trying to halve ur time? There is another cool feature coming to let you see only a section of time for your stats; that might help. No marriage please .

Originally Posted by realazthat

Its a ranking system that I am experimenting with. It involves KDR and XP, and combining them to make them fair accross classes, so that someone playing Sci heavily won't weigh more than someone playing Soldier heavily. If you want to see the math:

http://forums.stargateworlds.com/sho...5954&page=3#63

multiple linear regression analysis is what you want to accurately portray that

Originally Posted by zainea13

Yeah, you can't be picky when it comes to stats. Averaging everything is the best way. Or some form of statistical analysis. Run t-tests, or z-tests, or Correlations, Multiple Regression analysis. W/e you do. In the long run, there will always be outliers, and things that don't fit nicely within your data. In a situation like this where there are so many factors involved. I would suggest a Multiple regression analysis. I don't know if browsers can run this though, lol.

Multiple regression analysis (in a very brief nutshell) is like equations in math class.

Variables:
a=scientist, b=goa'uld, c=soldier, d=jaffa, e=commando, f=ashrak

equation: .34a+2b+4.32c-2.2d+1e+1f= (ranking stat)

After taking all your data and throwing it into a program like stat crunch for example (there are free online statistical programs, i have a free link to stat crunch if needed). You will want to pull the statistics from players in SGR, you don't want it to be too skewed one way or the other (either, the people that play SGR like they breathe, or the people that played it once and quit).

Best way to do this in statistics? Same way in every thing, random sample (best to be from 30-100, above 100 you're results won't change much but it will become a mess, some will say this number is 200, its just preference). So, taking this sample you want to run some tests on it to find where your mean is, hopefully you have a normal sample (bell shaped curve). So, (skipping some steps) you find out how much one variable effects another, negatively or not. Which would probably result in no correlation as these factors are all independent. If you read this, good job. But I know that all these mumbo jumbo things with K/D is more important, Shots fired is more important, XP/time is best! isn't the way to go.

If someone plays all sci, their shots fired and accuracy will be too high compared to the rest. Scrap that

K/D, sometimes you have good team players who aren't afraid to die, other times you have people that can't live, and other times you have people that don't care about team work, all they care about is their K/D.

XP, some say this punishes new guys or those who have a life. which can be true. So you throw in the time factor. Well if you earn 1000 XP for every 1 hour you play, but are clearly a better player than someone that earns 1000 XP in 1/2 an hour (perhaps they only play KOTH and rack up the XP).

So the dilemma arises. But, wait perhaps W/L is best. But then you run into the problem that this is an individual ranking, not team ranking, especially because your team is never the same.

Then people say "Put them all together". But which do you give the most weight to? Or maybe all the same weight? But thing is that they all have extreme pitfalls. Which really means you should throw them out the window as your confidence interval is sooo low because you can't confidently say that any one of these represents the true best player. Let alone throwing them all together.

So what I was saying about multiple reg analysis earlier may be true for these factors as they all do effect one another. If you are a better team player your K/D may not be as high. So you give more weight to K/D than W/L then you look at XP. And you think how much does XP weigh in on the skill of a player? And time, does someones time devoted have anything to do with how good they are?

So, take this, run your Multiple reg analysis with your factors and it will tell you how much weight to give to any one variable. Plug that in your PHP and you're set.

But, don't accept any equation given here. As they will just be something that is simple division or multiplication and doesn't give weight to any one variable over the other, and if it does, it doesn't give the correct amount. Saying K/D is 2x more important is like guessing who will win the NBA finals, you just don't know. Best way to do it is through a study to figure out which equation (or team) is the best fit.

And if you are not up to devoting the time, or finding someone who is, to having a proper formula, then just go with XP.

That math is above my head lol. I absolutely love math in general, but I find that statisticians write really truly horrible math books. I mean despicable. Every one I've come across. They quickly go off into math without defining everything, expecting you to already "just know" stuff. The math I'm used to, is where there is a theorem for darn assumption layed out in order that the book teaches u stuff. Hence, I hate statistics when taught. If you can come up with some equation, I'll put your secret-sauce on the rankings.

Now to address your concerns: Between classes, I've pretty much found a method to even out the weight, see my secret sauce description in the 0.3 release post above. How to combine that adjusted KDR, with adjusted XP, with adjusted WLR (Win/Loss Ratio), is another matter. I was thinking of simple multiplication, as it is simply subjective of what weight each should have. Is there any real definition of "who's better"? It's up to us to decide right now. Thus the entire equation is really subjective, and the final results will only reflect the "professional" opinion of this thread. People will simply be competing to win the equation, not to be better. Hopefully the two are goals are somewhat the same.

As for higher KDR for someone who plays an easy map for instance, or some other factor, this data is not available. The kills/map are not available to me, so I cannot factor that in. I can factor in time, but time is horribly bugged: Court has time jumps, and Amarna is frozen in time (and the W/L for Amarna are frozen at yet another time).

Originally Posted by realazthat

No marriage please .

You dont know what u miss by declining this chance buddy..

I just noticed that if I mouse over any of the underlined numbers in the rankings page it says blevok followed by the number no matter who the associated person on the list is.

that's because all the numbers were taken out to 64 decimal points in honor of Blevok's half-"parent" Tuvok from Star Trek. Vulcans... sheesh

Originally Posted by Daniel_Jackson_

that's because all the numbers were taken out to 64 decimal points in honor of Blevok's half-"parent" Tuvok from Star Trek. Vulcans... sheesh

Sure that's on the personal stats, on the rankings though it would be handy to actually show the persons name who's stat it is since the chart is so big.

On the Player profile section. Would it be easy enough to allow users to choose a custom profile pic. And a small 500 character about me? It might be nice, idk.

Originally Posted by Arcadenut

Maybe I'm confused, but I was under the impression that the W/L ratio was a team thing and had nothing to do with the individual's performance in the round.

If I come in with 10 seconds to go and my team wins, does that mean I'm a better team player then the other guy who came in with 10 seconds to go on the losing team?

Rather then W/L ratio, it should be based on the players score at the end of the round regardless of Wins or Losses. There have been many times where I've been at the top or near the top of the server, but still been on the losing team. There have been many times where I've been near the bottom too, but on the winning team.

As Realz said in the long run, the times you join a game 10 sec before it ends are somewhat insignificant. However when you join a match right before the end you have a (near) 50-50 chance to Win/Lose the match averaging it out to (near) 1-1, this however does; pull your W/L down if your W/L is positive, or push your W/L up if it is negative (as i'm certain that with most players their W/L is not exactly 1-1), ultimately this results in players Wins to Losses ratio being closer to 1 then it should be.

The main problem i have with W/L is Winner-Team-Joiners, there are specifically 3 players in the top 25 on Realz's stats in which this is the case and their actual skill level should be lower then what it is. Some players tough it out if they are on the losing team and some players don't, this is much more influential on W/L then it is on K/D, and EXP/Time-Played.

I do however like the idea of the default sorting stat being based off of something other then EXP which is primarily based on how much time you have played, (for no controversy have it set to Alphabetical order!).

Can u please add Altair? ID: 49911

V 0.4

The next version of evolved is out. Some really cool new features .

Changes Summary:

  History (new):
    [feature]History graphs (graphs over time of all sorts of crazy things)
    [feature]History per class filter (see graph metrics for each faction
      or class)
    [feature](easter egg) SG themed random busy animations (must have slow 
      internet to see lol)
  Rankings:
    [feature]Fixed table headers for rankings (the column headers move
      with page)
    [feature]Rankings show players now playing (they are colored gold)
    [feature]Rankings per class filter (you can see rankings per faction
      or class)
    [feature]Optional extra rankingfields (for crazy stats nut jobs ;))
    [feature]Legend (Gold == now playing)
    [feature]Highlighted rows on hover (as the mouse goes over the row)
    [fixed]Player column is sortable
    [feature]Tooltips of decimals in the rankings tab say the player name
      associated with the value instead of blevok since the rows are large
  General:
    [feature]Tooltips are much cooler ;)
    [feature]Name & rank in navbar
    [fixed]Fixed css with small res (everything is always centered instead
      of overflowing to the right)
    [feature]Legend (Gold == Now playing, Raised == Most used)
    [feature]Added crosspage bar in navbar (just cool visuals)

SVN Change Log:

  Moved out graph equations to a common include file stats.sqlfields.inc.php
  Added History tab
  Added flot graph on history tab
  Added busy animations 
  Added class filter on history tab
  Added class filter on rankings tab
  Added optional extra fields to rankings tab 
  Added some weapons pics; not yet functional
  Moved classes array for the class table to common.inc.php
  Centralized class pic link generation to common.inc.php
  Player info retrieval is now done in index.php, before loading the 
    included php page 
  Centralized the get*Total() functions to common.inc.php
  Added skeleton code for best weapon pics on main stat page
  Added legend
  Moved legend code to centralized include at stats.legend.inc.php
  Centralized weapons array to common.inc.php
  Added name to navbar
  Added rank to navbar
  Added cross-page bar to navbar
  Added submit button to "Get stats" form in navbar
  Moved get*TotalSQL() functions to common.inc.php
  Added Fixed Header to the rankings
  Tooltips are now jquerytools tooltips
  Player column is now sortable
  Prevented division by zero in Time to Advancement in 
    stats.timetoachievements.inc.php 
  Fixed CSS centering issue with overflowing table
  Fixed URI encoding issue with the custom  variable
  Eliminated the base tag
  Eliminated center tag in index.php
  Moved index.php javascript file additions to  to bottom; the 
    tooltip functionality should probably be pushed on to the document ready 
    queue last, due to FixedHeader's magic 
  Sorting now recognizes default sorted column
  SQL middleware now uses hashes for labels
  Exposed all fields to SQL middleware
  Using transactions for update.php

I love ot Thank you very much

YEAH REALZ REALY COOL MAN....love it love it u have taken the stats to a new level

14454 <- much appreciated if you could add me!

I like the new rankings page, good stuff!