Galcon Karaoke Party! Saturday, August 23th
:: Community ::
ForumsMessagesGroupsChat (1)Friends
 

Forums :: iPhone :: Strategy :: Official Stat Warriors Leader Board

You must sign in to post.

Page 10 / 22 1   2   3   4   5   6   7   8   9   10   11   12   13   14   15   16   17   18   19   20   21   22 

Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 12:18am

Ensign jazz

Joined: Feb 20, 2009
Posts: 534
Location: aka The_Musician

Running some of the numbers...  this would hurt a lot of the currently higher ranked people...

This is listing your PARTNERS rank at the time of the game, the AVG of how many games you played with a partner of that rank, and the Total games played with a partner at that rank, and then finally, how many partners you had at this rank.

Steve.Jobs Partners
RankNum AvgGameCount TotalGameCount PartnerCount
9 19 152 8
7 10 1354 133
6 8 1238 144
5 6 1006 160
8 5 196 35
10 5 11 2
3 3 270 81
4 3 338 90
2 2 139 60
1 1 53 44

IonizedFire Partners
RankNum AvgGameCount TotalGameCount PartnerCount
9 26 239 9
7 9 524 58
6 7 479 68
8 6 104 17
5 4 329 70
3 3 54 15
4 2 68 25
2 2 18 9
1 1 10 8
10 1 1 1

Wallplant Partners
RankNum AvgGameCount TotalGameCount PartnerCount
9 31 189 6
10 19 19 1
8 9 119 13
7 8 382 47
5 4 180 45
6 4 231 54
2 2 23 10
3 2 20 10
4 1 28 16
1 1 14 12


And these people would NOT be hurt by this change...

no.yes Partners
RankNum AvgGameCount TotalGameCount PartnerCount
7 9 272 28
8 9 88 9
6 7 180 25
4 4 49 10
5 4 106 24
3 3 31 8
2 2 17 6
1 1 5 3

DiebkStabber Partners
RankNum AvgGameCount TotalGameCount PartnerCount
9 7 47 6
7 5 490 89
8 5 72 14
5 3 391 101
6 3 396 100
4 2 131 58
1 1 14 11
2 1 52 27
3 1 60 35
10 1 1 1

Zmar01 Partners
RankNum AvgGameCount TotalGameCount PartnerCount
5 5 234 43
6 5 246 43
7 5 253 50
8 5 42 8
9 5 44 8
3 2 30 12
4 2 43 16
1 1 7 7
2 1 16 10


Please note: this doesn't necessarily indicate team switching...  it could also look like this if the players stayed with a single partner for many games.
post updated on Mar 28, 2011 @ 12:46am
Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 12:56am

Ensign jazz

Joined: Feb 20, 2009
Posts: 534
Location: aka The_Musician

I think maybe what I would have to do is this:

Take the "Average Partner Count Per Rank" and figure out if that Partner Count is above or below the global average for that rank.  THEN, I could take the players "Average Game Count" and compare it to the global average as well.

If the Partner Count is higher but the Game Count is lower, this could indicate team switching.  And the reverse could indicate staying as a team for longer runs of games.

However....  I would have to make sure to somehow NOT ding players who just simply play lots of games because in that case, both player count and game count would be higher then their overall averages.
post updated on Mar 28, 2011 @ 1:02am
Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 1:18am

Cabin Boy no.yes

Joined: Mar 7, 2010
Posts: 119

WOW, its really amazing how productive you are with this project and how quickly you try to implement new ideas.

I know there is possible solution - a fair solution. Its just a matter what exact formula we need to use. I must admit that i did not have time (and maybe skill) to analyse all formulas that you invented so far but i just see that they work. 

The project started with many flaws but as it shapes it seems that this will be revolutionary for galcon. I was sceptic at first, but now I see that using your exceptional analytic skills and enthusiasm for project you will create much better (more comprehensive) ranking system that was before. I Wish I had time and ability to help you out more.

Keep up this exceptional job Jazz! This will make you famous for next generations of galconians :) They will ALL use JAZZ's SYSTEM soon. Soon, flags shown in game will be flags from JAZZ's SYSTEM :))) (exagerrated in purpose to keep enthusiasm around this project - but we never know how this will end up...)
post updated on Mar 28, 2011 @ 3:09pm
Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 1:30am

Cabin Boy no.yes

Joined: Mar 7, 2010
Posts: 119

its amazing what can be extracted from stats data - even recognize team switching patterns.
post updated on Mar 28, 2011 @ 2:14am
Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 2:28am

Cabin Boy no.yes

Joined: Mar 7, 2010
Posts: 119

ok so how could be calculated AVERAGE PARTNER RANK? Lets call it APR

For no.yes this would be:
[(7*9)+(8*9)+(6*7)+(4*4)+(5*4)+(3*3)+(2*2)+(1*1)]
divided by 
number of all games played (39)
that gives: 5.8

5 is for red, 6 is for blue flag so it will mean that my average partner in all games was right below blue.

In similar way we can also calculate AVERAGE OPPONENT RANK Lets call it AOR.

Then we can conclude something by comparing APR with AOR for each player.

We must remember all the time that we try to isolate/extract skill of single player irrespective of his partners and opponents.

Lets say I played with APR 5.8 and AOR 5. That means that on average my partner was stronger than my opponents and my score should be adjusted down to reflect it.

In other case if somebody has APR low and AOR very high that means that he plays a lot games with weak partner against very strong opponents. His score should be adjusted up.
post updated on Mar 28, 2011 @ 3:10pm
Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 3:29am

Cabin Boy no.yes

Joined: Mar 7, 2010
Posts: 119

APR divided by AOR gives us number that shows everything very clear. Lets call it SWITCHING INDICATOR (SI).

Lowest possible SI will be 1/10 and highest 10 (because there are 10 ranks)

SI below 1 means that you are a victim of switchers. SI above 1 means that you are a switcher. (to be confirmed..)

The higher the SI the more you can be called switching whore :))

I think steve.jobs will be having very high SI.

edit: forget APR/AOR, see more useful AOR minus APR for score correction in posts below
post updated on Mar 28, 2011 @ 3:16pm
Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 4:54am

Cabin Boy esparano

Joined: Nov 5, 2009
Posts: 2471

Ok hold on there. It's very difficult to indicate team switching in stats, but if you're going to do it then I'd use aor/apr, then multiply by score. this way, having a high aor and low apr (you play higher-ranked opponents) will boost your score, but having a low aor and high apr (you teamswitch) will lower your score. aor/apr is just more useful than apr/aor :) I like this idea a lot, but Just be careful with this because it could affect players who just prefer to play with higher ranks or lower ranks.
post updated on Mar 28, 2011 @ 5:00am
Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 5:38am

Cabin Boy no.yes

Joined: Mar 7, 2010
Posts: 119

Actually i proposed APR/AOR above only as informative number (not multiplier of score). its true that fraction could affect players who just prefer to play with higher or lower ranks. E.g. 2 ranks difference (between aor and apr) in lower rank would have more impact on multiplier than in highest ranks. 

the difference AOR minus APR might be more useful to make correction of score.

difference AOR minus APR will be in range -9 to +9 (most cases falling probably in between -3 to +3) this difference might be converted to some adjustment to score. For example +2 means that on average you have two ranks lower partner than average rank of your opponents. Your score needs to be adjusted up

we just need to decide what percentage to adjust for each point so that this works

For example 40% for each point:
+3    add 120% to score
+2    add 80% to score
+1    add 40% to score
+0.5  add 20% to score
-1    subtract 40% from score 
or something like that
post updated on Mar 28, 2011 @ 3:53pm
Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 7:58am

Cabin Boy zmar01

Joined: Jun 6, 2009
Posts: 655

If you have to implement some sort of score that either raises or lowers your score why not do something that i was thinking of doing with my excell version of the leaderboard.

(But never got round to, i think jazzes is quite good enough)

What you do make a score.
Like this one
http://toomanywheres.com/galcon/LeaderBoard.aspx 

say me 1406287
and ionized fire 2706465
vs
gogo1234 24300 
and cognuts 24280 

say the total of the 2 sides
4112752
48580

then do a ratio 84.6:1
then for every game won me and ion get 1 point, the other team get 84 points.
then divide by the number of games at the end
then to make it effect the leader board times this extra score by 1000 and add it to the original score.

people who play with the worst people on the team will get more of these extra points to help them with their rank. 
Or something like that.


The problem is that you have to do it for every game played
Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 9:29am

1 Stripe Admiral diebkstabber

Joined: Dec 2, 2009
Posts: 700
Location: under my bed

Well we all know that ionizedfire is not a team switcher, he just plays in games with only high ranking players. Is there anyway to make it so, if all players in game where of similar ranking, that game would be scored normally? Of the three players that would be affected, I know Steve.jobs is the only one who cannot win unless he team switches. Wall.plant occasionally team switches, but is fully capable of winning against anyone, and ionizedfire is not a team switcher, and is basically one of the best ever. So what do you think you could come up with that would be fair to all three of these completely different playing styles and different scenarios? The type of player we would like to be penalized for ranking above his potential of course is Steve.jobs. But how to do that without affecting the other two who are clearly within their range of capabilities.
Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 10:05am

Cabin Boy no.yes

Joined: Mar 7, 2010
Posts: 119

diebkstabber:

comparing AVERAGE PARTNER RANK with AVERAGE OPPONENT RANK for these players and using it to adjust score for these players would work (see above). We just need to find how much we need to adjust score for differences in AOR minus APR for each player
post updated on Mar 28, 2011 @ 3:12pm
Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 10:11am

1 Stripe Admiral diebkstabber

Joined: Dec 2, 2009
Posts: 700
Location: under my bed

Ohh, sorry I didn't understand, opponent rank is considered I wasn't reading properly, I thought only partner rank was calculated. Thanks no.yes.
Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 10:47am

Ensign jazz

Joined: Feb 20, 2009
Posts: 534
Location: aka The_Musician

Cool ideas everyone!!  I'm impressed!

I'll try to get some numbers behind the APR/AOR to see if it will work correctly.
Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 12:55pm

1 Stripe Admiral quantum

Joined: May 19, 2010
Posts: 454
Location: Every(No)where @ once.

Wow, I was just mentioning the whole teamswitch thing as a way of bi tching about Steve.jobs (u ****er), but never expected you guys to actually account for switchers. This could be a whole revolution on the etiquette of galcon. If we eliminate teamswitcherz desire to up rank their partner, we are talking major revolution in the game! Awesome!!

I can see it now, goozok and thomass become the most sought after teammates lol people start joining games with a noob partner just to rank up! This will make the game better for noobs and pros I think!

Good stuff! Thanks Jazz!!
Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 4:23pm

Cabin Boy no.yes

Joined: Mar 7, 2010
Posts: 119

To reiterate all ideas above and change naming for easier understanding:

Now that we solved "goozok problem" (exceptional number of games played) we need to solve "Steve.jobs problem" (exceptional  teamswitching) and we will have perfect leader board.
 
It is possible to recognize team switching patterns from stats data

We have ten ranks and each has number from 1 to 10 (1 is pink, 10 is GA). The ranks data for partners and opponents played is available in stats

We can calculate AVERAGE RANK OF PARTNER and AVERAGE RANK OF OPPONENTS for each player. Both values will be in range from 1 to 10. By comparing these two values we can recognize switching

Lets call AVERAGE PARTNER RANK just PARTNERS. How we can calculate it? 

For no.yes this would be (based on data provided by jazz): 

[(7*9)+(8*9)+(6*7)+(4*4)+(5*4)+(3*3)+(2*2)+(1*1)] 
divided by  
number of all games played (39) 
that gives: 5.8 

5 is for red, 6 is for blue flag so it will mean that my average partner in all games was right below blue. 

In similar way we can also calculate AVERAGE OPPONENT RANK. Lets call it OPPONENTS

Then we compare PARTNERS and OPPONENTS of each player. 

We must remember all the time that we try to isolate/extract skill of single player irrespective of his partners and opponents.

Lets say I played with PARTNERS = 6 (blue) and OPPONENTS = 5 (red). That means that on average my partner was blue (admiral) and my opponents were red (captain). Then my score should be adjusted down to reflect it.  

The difference OPPONENTS minus PARTNERS might be useful to make correction of score. 

For example above we have 5 - 6 = -1. My score needs to be adjusted down indicated by minus sign

we just need to decide what percentage to adjust for each point of difference so that this works

For example 40% for each point: 
+3    add 120% to score 
+2    add 80% to score 
+1    add 40% to score 
+0.5  add 20% to score 
-1    subtract 40% from score 
 
As quantum mentioned: 'this could be a whole revolution on the etiquette of galcon... people will start joining games with a noob partner just to rank up! This will make the game better for noobs and pros'
.
.
.
post updated on Mar 29, 2011 @ 1:05am
Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 4:39pm

1 Stripe Admiral quantum

Joined: May 19, 2010
Posts: 454
Location: Every(No)where @ once.

I hereby nominate Jazz for the greatest community contributing member award of all time!

Thanks for saving galcon! This makes things so much more logical. That's the biggest problem in the world IMHO, illogical behavior!

Who else agrees that Jazz should win some type of award for his brave contributions to our fav game?
Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 5:00pm

Cabin Boy LjuboM

Joined: Jun 25, 2009
Posts: 4088
Location: 1v1 & 2v2

i don't understand a thing out of this =)
but i think it is awesome since everyone says it is!
Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 5:45pm

Ensign jazz

Joined: Feb 20, 2009
Posts: 534
Location: aka The_Musician

Well it seems to work....  Mostly....  but I'm not sure its the best solution yet...  cuz several of you wont like these results.  ;)


SI = Switching Index (good name btw)


Steve.Jobs
PartnerRankNum AvgOpponentRank AOR APR SI
1 5 5 1 4
2 5 5 2 3
3 5 5 3 2
4 5 5 4 1
5 5 5 5 0
6 5 5 6 -1
7 5 5 7 -2
8 5 5 8 -3
9 5 5 9 -4
10 4 4 10 -6


IonizedFire
PartnerRankNum AvgOpponentRank AOR APR SI
1 5 5 1 4
2 5 5 2 3
3 5 5 3 2
4 5 5 4 1
5 5 5 5 0
6 5 5 6 -1
7 6 6 7 -1
8 5 5 8 -3
9 5 5 9 -4
10 4 4 10 -6


Wallplant
PartnerRankNum AvgOpponentRank AOR APR SI
1 5 5 1 4
2 6 6 2 4
3 5 5 3 2
4 6 6 4 2
5 5 5 5 0
6 5 5 6 -1
7 5 5 7 -2
8 5 5 8 -3
9 5 5 9 -4
10 7 7 10 -3


no.yes
PartnerRankNum AvgOpponentRank AOR APR SI
1 5 5 1 4
2 5 5 2 3
3 4 4 3 1
4 4 4 4 0
5 5 5 5 0
6 5 5 6 -1
7 4 4 7 -3
8 5 5 8 -3


DiebkStabber
PartnerRankNum AvgOpponentRank AOR APR SI
1 4 4 1 3
2 4 4 2 2
3 4 4 3 1
4 4 4 4 0
5 4 4 5 -1
6 5 5 6 -1
7 5 5 7 -2
8 4 4 8 -4
9 6 6 9 -3
10 7 7 10 -3


Zmar01
PartnerRankNum AvgOpponentRank AOR APR SI
1 5 5 1 4
2 4 4 2 2
3 5 5 3 2
4 5 5 4 1
5 5 5 5 0
6 5 5 6 -1
7 5 5 7 -2
8 4 4 8 -4
9 5 5 9 -4
Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 5:49pm

Ensign jazz

Joined: Feb 20, 2009
Posts: 534
Location: aka The_Musician

Full details for Steve.Jobs:

PartnerRankNum AvgOpponentRank AvgGameCount PartnerWeight OpponentWeight AOR APR SI
1 5 10 10 50 5 1 4
2 5 28 56 140 5 2 3
3 5 56 168 280 5 3 2
4 5 74 296 370 5 4 1
5 5 219 1095 1095 5 5 0
6 5 258 1548 1290 5 6 -1
7 5 297 2079 1485 5 7 -2
8 5 43 344 215 5 8 -3
9 5 30 270 150 5 9 -4
10 4 4 40 16 4 10 -6
Re: Official Stat Warriors Leader Board :: Mar 28, 2011 @ 5:56pm

Ensign jazz

Joined: Feb 20, 2009
Posts: 534
Location: aka The_Musician

I'm going to try these numbers again, but this time, taking decimal places into consideration.  All the above nubmers are dealing with whole numbers.  

Don't get discouraged. I'm probably doing something wrong in themath somewhere.
post updated on Mar 28, 2011 @ 6:16pm

Page 10 / 22 1   2   3   4   5   6   7   8   9   10   11   12   13   14   15   16   17   18   19   20   21   22 

You must sign in to post.