Combat Arms KDR - Printable Version

Combat Arms KDR - Printable Version

+- Southperry.net (https://www.southperry.net)
+-- Forum: Social (https://www.southperry.net/forumdisplay.php?fid=14)
+--- Forum: Rubik's Cube (https://www.southperry.net/forumdisplay.php?fid=58)
+--- Thread: Combat Arms KDR (/showthread.php?tid=9465)

Combat Arms KDR - Hazzy - 2009-03-20

I want to create a confidence interval for my KDR in Combat Arms.
If I gather information about my Kills and Deaths per match, I can get a mean and StdDev for kills or deaths per game, but how can I transform that information into a ratio of kills per death, given only x-bar kills and deaths and S kills and deaths?

I can just divide the means, but what do I do for the StdDevs?

Combat Arms KDR - JoeTang - 2009-03-20

The website calculates your mean kills and deaths per match for your entire account record, just so you know.

Combat Arms KDR - Russt - 2009-03-20

I think this is for some statistics thing.

I don't understand what you're asking, though. Rephrase?

Combat Arms KDR - Hazzy - 2009-03-20

Entire record can span some time. If I want to know a more recent account, it doesn't help.
Doesn't give me a StdDev either. ;[

Combat Arms KDR - Nikkey - 2009-03-20

Use
[Image: equation.gif]

Combat Arms KDR - Hazzy - 2009-03-20

Russt Wrote:I think this is for some statistics thing.

I don't understand what you're asking, though. Rephrase?

Given two means and StdDev of two independent samples, how do I find the StdDev of the ratio of the two samples?

If I go collect data from games, I can get an average kills/game, and an average deaths/game.
If I want the average KDR per game, can I get that without knowing every point of data in both samples?

Combat Arms KDR - Nikkey - 2009-03-20

Hazzy Wrote:Given two means and StdDev of two independent samples, how do I find the StdDev of the ratio of the two samples?

Well, we know that SD(X) is the square-root of VAR(X).

First of all, take the SD of both samples and square them. Now you got the Variance of both samples. Let's say we're going to find the Variance for the combination. Using the variance-formula (Though, it was expressed more detailed than needed, sorry about that.), we find the variance for both of the samples. Now take the root of the variance to find the StdDev:

[Image: 38617040.gif]

Hope that helps you. It's kinda like Pytagoras actually.

Combat Arms KDR - Hazzy - 2009-03-20

But does Var(x1)/Var(x2) = Var (x1/x2)?
I'm just not sure what to do with the StdDev of both samples when I divide the means.
Mean1/Mean2 = average KDR
Does Var1/Var2 = average varriance of the ratio? With the Data I have right now, that comes out to a StdDev 2.9 for the ratio, which is huge.

Combat Arms KDR - Nikkey - 2009-03-20

Wait, ugh, what's the KDR?

Combat Arms KDR - Hazzy - 2009-03-20

KDR - Kill Death Ratio. Kills/Deaths

Combat Arms KDR - Cyadd - 2009-03-20

[COLOR="Red"]Long post here:

I'm guessing you are going to be using a two sample mean test (Kills, deaths). First off you are going to assume that your kills-deaths (I'm guessing you kill more than die) equals zero. This is your mu1-mu2=0. Next, I'm going to guess you will use a 95% confidence interval. Now is where the fun starts.

Get around 30-ish samples for kills and deaths per game. Or you can use less, if you will dotplot it, and see no skewness or outliers. To create your confidence interval, use the formula:

x1(kills mean)-x2(deaths mean) +/- t"star"(sqrt((s1^2/n1[s.d. of kills^2 divided by the sample]+(s2^2/n2))[same as kill, but for deaths]

For t"star", go on the internets, or this site http://davidmlane.com/hyperstat/t_table.html , and go to your d.f. (degrees of freedom=sample size-1 [so choose the same amount for each sample]) that is on the leftmost column, and your % will be .025, as this is 2 tailed.

Substitute everything in, and this should give you an interval of the true difference between your kills and death.

If this seems difficult to read, gimmie the numbers, and I'll give you your interval, for only 3 bucks![/COLOR]

Combat Arms KDR - Hazzy - 2009-03-20

I know how to run a hypothesis test. My issue lies with obtaining a StdDev for the ratio of kills to deaths. ;[
I want to create a confidence interval for my KDR of a random game. I need a StdDev of the ratio before I can do that. Dividing means works fine. Dividing StdDev of both samples doesn't work out right...
My current data has:

	Kills	Deaths
Mean	22	8
StdDev	6.3	4.12

This comes out to a Mean KDR of 2.67.
Sqrt(6.3^2/4.12^2)=1.53

Mean of 2.67, StdDev of 1.53... That strikes me as wrong, since almost a third of my games will have me with under 1.00 KDR. (Quick mental calculation that might be wrong)

Combat Arms KDR - Cyadd - 2009-03-20

[COLOR="Red"]take the sqrt of your kill s.d. and square it/ number of sample + death s.d. square it/ number of samples.

Or, take all your kills and deaths, take the difference, and find the s.d. of that.

give me all your numbers[/COLOR]

Combat Arms KDR - Russt - 2009-03-20

Hazzy Wrote:I know how to run a hypothesis test. My issue lies with obtaining a StdDev for the ratio of kills to deaths. ;[
I want to create a confidence interval for my KDR of a random game. I need a StdDev of the ratio before I can do that. Dividing means works fine. Dividing StdDev of both samples doesn't work out right...
My current data has:

Kills Deaths
Mean 22 8
StdDev 6.3 4.12

This comes out to a Mean KDR of 2.67.
Sqrt(6.3^2/4.12^2)=1.53

Mean of 2.67, StdDev of 1.53... That strikes me as wrong, since almost a third of my games will have me with under 1.00 KDR. (Quick mental calculation that might be wrong)

Oh. Yeah, you couldn't just do that. The problem with dividing that way is that (kills - mean kills)/(deaths - mean deaths) doesn't equal (kills/deaths) - (mean kills/mean deaths).

I'm not entirely sure on what is valid though. Brb.

Combat Arms KDR - Hazzy - 2009-03-20

Cyadd Wrote:[COLOR="Red"]take the sqrt of your kill s.d. and square it/ number of sample + death s.d. square it/ number of samples.

Or, take all your kills and deaths, take the difference, and find the s.d. of that.

give me all your numbers[/COLOR]

Kills	Deaths
25	5
16	5
19	7
15	8
20	7
29	9
31	17

	Kills	Deaths
Mean	22	8
StdDev	6.3	4.12
n	7	7

Combat Arms KDR - Nikkey - 2009-03-21

hazzy Wrote:

kills deaths k/d
25 5 25/5
16 5 16/5
19 7 19/7
15 8 15/8
20 7 20/7
29 9 29/9
31 17 31/17

kills deaths k/d
mean 22 8 2.96*
stddev 6.3 4.12 0.99**
n 7 7 7

* 2.95602574...
** 0.98760861...

?

Combat Arms KDR - Hazzy - 2009-03-21

o.o
Excel lied to me? D:

Combat Arms KDR - Nikkey - 2009-03-21

Hazzy Wrote:o.o
Excel lied to me? D:

I dunno. I took kills and divided by deaths to get a k/d-rate for all the different matches. Then I just took those numbers on my calc and got stddev, mean, sum etc.

Combat Arms KDR - Russt - 2009-03-21

Excel's STDEV() divides by n-1. It's the sample stdev. STDEVP() divides by n.

Combat Arms KDR - Flaxative - 2009-04-03

Sorry, I thought this was a different thread. Extremely sorry. Won't do it again.