Statistical Probability

[robluce1]

I said this during Nov 5th, and I'm sure there are other people that see the same thing too.

My guess is that today, the donation graphs parallel between the 5th and the 16th. From a statistical probability viewpoint, this is absolutely impossible in the real world.

There is absolutely no way that the graphs could be accurate without a artificial limiter somewhere within the system.

In other words, the only way for those graphs to be correct (and I believe they are correct), something in the system is restricting or queueing the total number of donations per minute (not $ per minute, donations per minute).

This will bug me -all- day long. Same as the 5th.

Rob

[RonFan1776]

The graphs aren't linear. You can tell this easily if you look at the $/hr graphs. There is actually quite a bit of fluctuation. I agree that it is weird that it is as linear as it is, but that's why this world is interesting. Most things don't work out as you would first expect.

[dvictr]

you bring up a good point which is disscussed in graduate theoretical statistics classes but there are things such as elliot wave theory that believe humans behave in rythmic patterns.. some questions as to why and when prime numbers appear in strings of large numbers is unknown. or why people choose to donate a 11:05 as opposed to 10:55/ morning or afternoon.. its weird

[FreeTraveler]

We saw over $600,000/hr rate just after midnight, and everything was keeping up just fine. We're currently under $300,000/hr rate, so I'll have to call this theory wrong.

[wisconsinite]

http://paulcash.slact.net/today-paulflow.png

[greves]

Yeah this is going to bother me again all day too. Can ANYONE explain it, with PROOF (not guessing, like last time around)???

[wisconsinite]

Are you saying that it is statistically improbable for most Americans to be asleep at 4:30am on both days?

[greves]

wisconsinite: Look at the graphs starting at, say, noon. From noon to midnight, the "total donations" graph is practically linear. How and why is this possible?

[wisconsinite]

The limiting factor in the graph below is that you are always adding on to the total, so the line can never curve downward it can only even out, as it did around 4am. While in the graph below that you see the hour on hour donation rate, not the total, and so the fluctuation is quite visible.

http://ronpaul.xnamatrix.com/comparison.png

http://paulcash.slact.net/today-paulflow.png

[robluce1]

300k vs 600k an hour is an issue with donation amount, not donations per hour.

1. Humans are predictable.
2. Queueing theory is an established science.

I do this for a living. I'm not going to say trust me, I'm going to say, someone give me a realistic model which explains this so I don't go nuts.

Donation spike at midnight... simple, people logged on at midnight and donated some amount greater than the typical donation size (constant donations per minute, higher amounts per donation). 2300 clubbers, that's simple.

After that, everyone goes to sleep. Starting this morning, we start to get a constant rate of donations coming in from the time at which people wake up until midnight. Where's Lunch? Where's dinner? There should be some variation on a -hour by hour- basis, not constant.

Minute by minute, there may be some variation. Total number of people donating $25 vs total number of people donating $100 in any given minute. But to average out over an hour by hour basis, based on a 4 hour window (EST - PST), that's just impossible.

There -must- be something queueing the donations. That's the only thing that could explain the graphs.

Rob

[RonFan1776]

The limiting factor in the graph below is that you are always adding on to the total, so the line can never curve downward it can only even out, as it did around 4am. While in the graph below that you see the hour on hour donation rate, not the total, and so the fluctuation is quite visible.

QFT.

[wisconsinite]

Oh, and incidently, it's still climbing, so we're about to tear the roof off this sucker.

[dvictr]

THE graph is in a liner not log scale!!!!

100K - 200K is a 100% increase but 200K - 300K is only a 50% increase although both are a 100K increase

the log scale would be more like this 1--------2-------3------4-----5----6---7--8-9 and the properties of this curve would be different

[FreeTraveler]

Oh, and incidently, it's still climbing, so we're about to tear the roof off this sucker.

Yeah, if it's locked in, how come the two lines are starting to diverge?

[werdd]

do you have a graph comparing the donation / hour to the 5th?

[Stealth4]

300k vs 600k an hour is an issue with donation amount, not donations per hour.

1. Humans are predictable.
2. Queueing theory is an established science.

I do this for a living. I'm not going to say trust me, I'm going to say, someone give me a realistic model which explains this so I don't go nuts.

Donation spike at midnight... simple, people logged on at midnight and donated some amount greater than the typical donation size (constant donations per minute, higher amounts per donation). 2300 clubbers, that's simple.

After that, everyone goes to sleep. Starting this morning, we start to get a constant rate of donations coming in from the time at which people wake up until midnight. Where's Lunch? Where's dinner? There should be some variation on a -hour by hour- basis, not constant.

Minute by minute, there may be some variation. Total number of people donating $25 vs total number of people donating $100 in any given minute. But to average out over an hour by hour basis, based on a 4 hour window (EST - PST), that's just impossible.

There -must- be something queueing the donations. That's the only thing that could explain the graphs.

Rob

The graphs are not linear, the donations are not queing. Take the first derivative of the donations graph and this is quite obvious. Hell just looking at Nov 5th you can see small fluctuations in the donation total graph. This theory of yours is bunk.

[dvictr]

THE graph is in a liner not log scale!!!!

100K - 200K is a 100% increase but 200K - 300K is only a 50% increase although both are a 100K increase

the log scale would be more like this 1--------2-------3------4-----5----6---7--8-9 and the properties of this curve would be different

close thread :D

[wisconsinite]

do you have a graph comparing the donation / hour to the 5th?
:) But notice the scales are different. We're at $300,000 per hour now, which was never reached on the 5th. :D
http://paulcash.slact.net/today-paulflow.png

http://paulcash.slact.net/November-5th/today-paulflow.png

[Ronin]

Guys, just want to point this out. If it is queueing then you should be looking at the number of donors/time period. The payment gateway doesn't really care how much is being donated.

Edit: Also, 11/5 was a work day. Most people sleep in a little on the weekend.

[voytechs]

I said this during Nov 5th, and I'm sure there are other people that see the same thing too.

My guess is that today, the donation graphs parallel between the 5th and the 16th. From a statistical probability viewpoint, this is absolutely impossible in the real world.

There is absolutely no way that the graphs could be accurate without a artificial limiter somewhere within the system.

In other words, the only way for those graphs to be correct (and I believe they are correct), something in the system is restricting or queueing the total number of donations per minute (not $ per minute, donations per minute).

This will bug me -all- day long. Same as the 5th.

Rob

You don't see too many fluctuations and outlies because of the scale of the graphs. If you looked at them minute by minute or second by second, you'd a lot more what you call fluctuations. The graphs reflects the curves of the total very accurately based on the rate graph. Its just a huge number of people donating all at the same time.

I have both today an Nov. 5th graphs printed out in my hand and comparing. Its startling how much more support we have this time around.