Calculate SVM based on math

[swisscheese]

Ok so lets say you have some quantaties that you believe are important to your SVM (height, face, etc) that you believe to be *STATISTICALLY INDEPENDENT* and you have some relative factors of importance (maybe you believe face is 2 times as important as height etc). Now let's say you estimate your percentile (top 10% = 90 percentile) for each of those quantities. Remember to condition on the environment you are competing in (top 10% of all people? of men? of men in a club?). How do you calculate your SMV?

STEP 1
Go from percentile p to z score for each quantity using the probit function
STEP 2
Calculate weighted sum of z scores
STEP 3
Divide the result by the square mean of the weights (the L2 norm of the weight vector)
STEP 4
Plug the resulting z score into the Gaussian CDF

The result is the percentage of people you mog. Here is a toy example:
Bob has top 15% height, top 5% dick size. Bob has a below average face, mogging just 20% of people. Lets say height is half as important as face but double as important as dick size.
STEP 1
probit(0.85) = 1.0364
probit(0.95) = 1.6449
probit(0.2) = -0.8416
STEP 2
2 * 1.0364 +
1 * 1.6449 +
4 * -0.8416 = 0.3513
STEP 3
sqrt(1**2 + 2**2 + 4**2) = sqrt(1 + 4 + 16) = sqrt(21) = 4.58257569496
0.3513 / 4.58257569496 = 0.07665994483
STEP 4
CDF(0.07665994483) = 0.53

This guy mogs 53% of people. He is average.

You can also use this kind of calculation to figure out how rich you have to be to be top 10% SVM or whatever else interests you.

________________________________________________________________________________________________________________
for nerds who want proof consider that the sum of independent gaussians is a gaussian with variance equal to the sum of the individual variances and that a factor times a gaussian is just another gaussian whose standard deviation is also multiplied by that factor.

 

[swisscheese]

If someone is interested i might make a little webapp like calcsd where you can calc this easier.

 

[SexuallyAbusive]

you need to go outside more often

 

[swisscheese]

you need to go outside more often

Duly Noted
Here is some python code:

p = []
w = []
while(True):
  inp = input()
  if "." in inp:
    break
  v1, v2 = [float(i) for i in inp.split(",")]
  if v1 > 1:
    v1 /= 100
  p.append(v1)
  w.append(v2)

from scipy.stats import norm

z = norm.ppf(p)
m = sum([zz * ww for zz,ww in zip(z,w)])

import numpy as np
n = np.linalg.norm(w)
print(norm.cdf(m/n))

It takes input in this form:
0.5,1
0.99,5
.

The period is important

 

[swisscheese]

Duly Noted
Here is some python code:

p = []
w = []
while(True):
  inp = input()
  if "." in inp:
    break
  v1, v2 = [float(i) for i in inp.split(",")]
  if v1 > 1:
    v1 /= 100
  p.append(v1)
  w.append(v2)

from scipy.stats import norm

z = norm.ppf(p)
m = sum([zz * ww for zz,ww in zip(z,w)])

import numpy as np
n = np.linalg.norm(w)
print(norm.cdf(m/n))

It takes input in this form:
0.5,1
0.99,5
.

The period is important

p = []
w = []
while(True):
  inp = input()
  if ";" in inp:
    break
  v1, v2 = [float(i) for i in inp.split(",")]
  if v1 > 1:
    v1 /= 100
  p.append(v1)
  w.append(v2)

from scipy.stats import norm

z = norm.ppf(p)
m = sum([zz * ww for zz,ww in zip(z,w)])

import numpy as np
n = np.linalg.norm(w)
print(norm.cdf(m/n))

EDIT: use this code and semicolon instead of period to end input

 

[Danish_Retard]

How to actually calcuate SMV: Download Tinder or touch grass and test your SMV outside.

Both will give you a much better intuitive understanding of your own SMV or someone else's. Imagine being worried a guy is mogging you at a party and whipping out your calculator.

 

[Chadullahlite]

If someone is interested i might make a little webapp like calcsd where you can calc this easier.

Do it

 

[swisscheese]

How to actually calcuate SMV: Download Tinder or touch grass and test your SMV outside.

Both will give you a much better intuitive understanding of your own SMV or someone else's. Imagine being worried a guy is mogging you at a party and whipping out your calculator.

I completely agree with this (and I have) but I made this thread because I saw people just doing a weighted average of their height and face to determine their SMV which gives an estimate that is closer to the middle of the spectrum than the 'correct' calculation.

Also I would like to note that on tinder you can fuck up easily by posting shit photos and in irl you need many samples for your estimate to converge because real life is complex.