Modeled PF Trophies

  • Level: Undergrad 
  • Thread starter AlexB23
  • Start date
14 replies · 3K views · last reply
AlexB23
Messages
337
Reaction score
343
Hey guys. I modeled the distribution of PF post trophies using Wolframalpha.com. The formula is 58757.3/x^0.669229 with a fit of 0.999988. It slightly breaks down when the numbers get high.

Forum post milestone badge reading 40,000 Post Mark (70 points) with subtitle Congrats on posting 40K messages, above a timeline of post mark tiers from First Post to 40,000 Post Mark.


Log-log scatter plot with connected points titled Number of Members vs. Number of Posts; members decline from about 58760 at 1 post to 5 at 40000 posts, beside a table of the same values.


( data from: https://www.physicsforums.com/members/?key=most_messages )
 
Last edited by a moderator:
AlexB23 said:
The formula is 58757.3/x^0.669229

Interesting, so the number of members (call that M) with x posts is reflected by that fit?
$$M = \frac{58757}{x^{0.669}}$$
 
berkeman said:
Interesting, so the number of members (call that M) with x posts is reflected by that fit?
$$M = \frac{58757}{x^{0.669}}$$
Yep, roughly. It goes to show how real life can be modeled somewhat with math, including social media.
 
AlexB23 said:
Yep, roughly. It goes to show how real life can be modeled somewhat with math, including social media.
I have a problem with this formula. ##M(x)## is convex, and your graph is concave.
 
fresh_42 said:
I have a problem with this formula. ##M(x)## is convex, and your graph is concave.
The graph is log plotted.
 
The plot seems correct by comparison with your list. I fed https://elsenaju.eu/Calculator/online-curve-fit.htm with your data to ##\log_{10}## on both scales. It also produces a power law with the wrong curvature: $$M=5.583 x^{−1.09}\quad\text{red}$$
But the polynomial (degree ##4##) has a good fit:
$$
M= -0.046x^4+0.418x^3-1.276x^2+0.597x+4.769 \quad\text{blue}
$$

Graph with x-axis from 0 to 5 and y-axis from 0 to 7. A blue curve rises slightly then falls steeply, and a red curve decreases from upper left to lower right. Red dots labelled p1 to p8 mark points on the blue curve.


Here is the file that I used:
0;
4.77;
2;
3.446;
2.4;
3.146;
3;
2.67;
3.7;
2.05;
4;
1.785;
4.3;
1.43;
4.6;
0.7;
 
fresh_42 said:
I have a problem with this formula. ##M(x)## is convex, and your graph is concave.
Hmm, I'm not seeing that. The formula for ##M(x)## is a simple power law and is thus linear on a log-log graph:
Log-log scatter plot of N_mems against N_posts with red data points and a blue straight fit line, labelled N_mems = 58761 / (N_posts)^0.66924, with both axes running from 1 to 10^5.

And the fit can be significantly improved by introducing a term quadratic in ##\text{ln}(x)##:
Log-log plot of N_mems against N_posts with red data points and a blue fitted curve, with the formula N_mems = 54141 e^(-0.037154[ln(N_posts)]^2) / (N_posts)^0.43060 printed on the graph.
 
renormalize said:
Hmm, I'm not seeing that.
The plot in post #1 is clockwise bent, and ##y=x^{-2/3}## anticlockwise. If we take the log then ##\log y=-\dfrac{2}{3}\log x## which is linear, still not curved like the plot is. I first considered only the original function, and then mistakenly took only the log on one scale.
 
"With four parameters I can fit an elephant, and with five I can make him wiggle his trunk." - John von Neumann

Is there something significant here? What does it mean?
 
I expect the 58757.3 coefficient would be a function of the number of members in the population, that will change with time.
But where does the coefficient 0.669 arise from. Could it theoretically be 2/3 = 0.6667 ? Maybe there is a set of PDEs that would reveal how the distribution developed?
 
DaveE said:
"With four parameters I can fit an elephant, and with five I can make him wiggle his trunk." - John von Neumann
https://arxiv.org/abs/2407.07909

Fitting an Elephant with Four non-Zero Parameters​


? Why non-zero ? Maybe they meant to say "non-Zeno".