Moderated Coin Flipper

[jsfisher]

[ qimg]http://www.internationalskeptics.com/forums/imagehosting/thum_512824ef5162a06a0d.jpg[ /qimg]​

Careful. If you clutch those pearls too hard, they may shatter.

 

[Leumas]

Careful. If you clutch those pearls too hard, they may shatter.

Thanks for your concerns... I appreciate them.

Have you tried the Coin Flipper Game in the new version of Coin Flipper V5... did you win some $$$ or did you lose the $1000 dollars in a few rounds of GAMBLING on "determining" the number of heads "approximately"?

How do you like the ability to plot all the data (averages of the runs and the running averages) all in the app instead of having to copy it and take to another app... you can still do that too if you wish.

The plots can be zoomed in and out dynamically and panned and you can zoom in on a section ... all to help you see how arrantly you are wrong.

 

[jsfisher]

The plots can be zoomed in and out dynamically and panned and you can zoom in on a section ... all to help you see how arrantly you are wrong.

Arrantly wrong? Really, how so?

Your plots are useless because you application is not providing any sort of empirical data...just model data, and the model is defective.

 

[Leumas]

Here are some plots of data... generated in Coin Flipper V5 by the click of a mouse.... and the zoomed in stuff also by a click and a drag and then another click to save the images.

1. Deceptively tricks the not careful observer that it indeed looks like converging.
2. Shows that it is not... when one zooms in on the details... much like a terrain from 20,000 feet in the air might look flat but when one gets down to the level of the terrain one can see that it is not.
3. And lest one say AHA... but more runs will ... blah blah blah... this plot nips that AHA right in the bud

[IMGW=700]http://godisadeadbeatdad.com/CoinFlipperImages/NotConverging(1).png[/IMGW]


[IMGW=700]http://godisadeadbeatdad.com/CoinFlipperImages/NotConverging(2).png[/IMGW]


[IMGW=700]http://godisadeadbeatdad.com/CoinFlipperImages/NotConverging(3).png[/IMGW]
​

 

[Thermal]

Leumas, I don't think anyone expects a smooth, even convergence ever closer to exactly 50/50. What is expected is what is seen. Wild fluctuations away from 50% early on, then settling to closer to even odds, but never staying there precisely or converging smoothly. Just like, the days of seeing 70% heads are long in the rear view mirror after the first few runs. Then the game is averaging damn close to the 50 yard line with no more wide deviations.

 

[Leumas]

Leumas, I don't think anyone expects a smooth, even convergence ever closer to exactly 50/50. What is expected is what is seen. Wild fluctuations away from 50% early on, then settling to closer to even odds, but never staying there precisely or converging smoothly. Just like, the days of seeing 70% heads are long in the rear view mirror after the first few runs. Then the game is averaging damn close to the 50 yard line with no more wide deviations.

Yes... and no.

The wild fluctuations are still there... it is just that in the AVERAGING OUT over a large numbers of the +/- fluctuations adding up you get a less fluctuating AVERAGE.

See this image

[IMGW=700]http://godisadeadbeatdad.com/CoinFlipperImages/NotConverging(4).pmg[/IMGW]​

Leumas, I don't think anyone expects a smooth, even convergence ever closer to exactly 50/50.

Some here did... e.g.

Over the series, the oscillations damp down, staying closer and closer to the 50/50 line. The oscillations converge on 50/50. Q.E.D.

.... but never staying there precisely or converging smoothly.

EXACTLY!!!

In other words...

...
It will forever oscillate like this although you are correct it is closer to the 50%... BUT NEVER CONVERGES on it....

 

[Leumas]

Then the game is averaging damn close to the 50 yard line with no more wide deviations.

Ok... why don't you try the Coin Flipper Game.

You can run it from Coin Flipper V5.

See how good your guessing is and how long you can last without losing the $1000 purse the game gives you to start playing.

Say you bet $100 at a time with say 10,000,000 flips a run and setting δ to say 500... that is an allowance for 500 +/- of fluctuations from your guess and pays out nice odds (3:1) if you win the run.

See how many runs you would last.

[IMGW=350]http://godisadeadbeatdad.com/CoinFlipperImages/CoinFlipper5.png[/IMGW]

[IMGW=350]http://godisadeadbeatdad.com/CoinFlipperImages/CoinFlipper5_Game2.png[/IMGW]

[IMGW=350]http://godisadeadbeatdad.com/CoinFlipperImages/CoinFlipper5_Game3.png[/IMGW]​

 

[Thermal]

Ok... why don't you try the Coin Flipper Game.

For the same reasons I have been saying since the start of this thread. I am no more interested in flipping coins then I am playing tic tac toe for 500 billion rounds to see if the odds of winning change.

You can run it from Coin Flipper V5.

See how good your guessing is and how long you can last without losing the $1000 purse the game gives you to start playing.

Say you bet $100 at a time with say 10,000,000 flips a run and setting δ to say 500... that is an allowance for 500 +/- of fluctuations from your guess and pays out nice odds (3:1) if you win the run.

See how many runs you would last.

[IMGW=350]http://godisadeadbeatdad.com/CoinFlipperImages/CoinFlipper5.png[/IMGW]

[IMGW=350]http://godisadeadbeatdad.com/CoinFlipperImages/CoinFlipper5_Game2.png[/IMGW]​

Tell you what: you put up some real money and I'll run the books. Well.make a killing. But I'd sooner get a root canal than flip coins as a means of entertainment, even less so for watching a computer pretend to flip them.

 

[Leumas]

For the same reasons I have been saying since the start of this thread. I am no more interested in flipping coins then I am playing tic tac toe for 500 billion rounds to see if the odds of winning change.

Tell you what: you put up some real money and I'll run the books. Well.make a killing. But I'd sooner get a root canal than flip coins as a means of entertainment, even less so for watching a computer pretend to flip them.

In a Casino in Western Australia they had an area where people (many) gambled on two coins flipped by a human and landing on the floor. And people really gambled on this real money... it is called Two Up.

In many Casinos in Las Vegas they have machines after machines where people gamble real money on pushing a button and watching the SIMULATION of wheels rolling to get three images that are alike or not.

Some people would gamble on two flies on a wall betting which one will fly off first.

But this is not the point I am making with the Coin Flipper Game.

 

[Thermal]

Two crackheads might bet on which of their fromt teeth fall out first too. But that's not the point that I'm making either.

 

[Leumas]

Empirical Data Belies Bare Assertions

Two crackheads might bet on which of their fromt teeth fall out first too. But that's not the point that I'm making either.

What I am talking about is this

In my reckoning Empirical Data will always trump any hand waving bare assertions.

Having faith in mathematical 300 years old theorems being what one CONSTRUES them to mean when they do not mean anything of the sort is fine and dandy.

...

Well... SCIENCE to the rescue.... EMPIRICAL EXPERIMENTATION is what is needed.
So... get 10,000 coins and toss them up and start tallying.

And for a good experiment to be valid it has to be repeated many times to assure repeatability and reliability... so you will have to collect the coins again and do the whole thing again and again say for 10 or 100 or even 1,000 times to be assured of the results' validity or even 1,000,000 for even better confidence.

And what are the parameters of the experiment here

• 10,000 coins tossed
• Guessing a number for the heads H_g (tails will of course be 10,000 - H_g)
• Specifying what margin of error is allowed (e.g. H_g +/- δ) where δ has to be specified.
• If after tallying the results, the number of heads (H_r) is H_g-δ <= H_r <= H_g+δ then SUCCESSFUL GUESS (S_g).... if not then FAILED GUESS (F_g).

Repeat the above say 10 or 100 or 1,000 times and tally up S_g.

...

So

come on... get the 10,000 coins and start tossing.​

Or

get cracking on calculating n= f(p=1,ε=0)... or get a math priest to divine it for you.​

Verify for yourself using 300 years old math that no one here can specify so far (let's hope one day soon)... or use EMPIRICAL EXPERIMENTATION to show that it is indeed random (i.e. probability < 100%) that guessing the H_g for 10,000 coin tosses will be determinable with no random error.
What is that you say... doing all that is prohibitive in time and cost and physical efforts?

Ok... you can use Coin Flipper V5 instead to do all that tossing and tallying and displaying of the results for you... you can now easily obtain EMPIRICAL DATA for yourself.
What is that you say.... you are concerned about PRNG and TRNG and whatnot and whatchamacallit... then go do it physically or figure out the 300 years old math formula above....

Failing to do any of the above 3 options but then handwaving this or that is nothing but an Ipse dixit fallacy.

[IMGW=700]http://godisadeadbeatdad.com/CoinFlipperImages/CRNG_Oscillating_1.png[/IMGW]
[IMGW=700]http://godisadeadbeatdad.com/CoinFlipperImages/CRNG_Oscillating_2.png[/IMGW]
[IMGW=700]http://godisadeadbeatdad.com/CoinFlipperImages/CRNG_Oscillating_3.png[/IMGW]
[IMGW=700]http://godisadeadbeatdad.com/CoinFlipperImages/CRNG_Oscillating_4.png[/IMGW]
[IMGW=700]http://godisadeadbeatdad.com/CoinFlipperImages/CRNG_Oscillating_5.png[/IMGW]
[IMGW=700]http://godisadeadbeatdad.com/CoinFlipperImages/CRNG_Oscillating_6.png[/IMGW]​

 

[Wudang]

Have you finished reading the book you linked to... or was it "gigo gallop"?




When are you going to

Sigh. I have already said that one test of an rng is whether it converges or not. You have written a unit test for your rng then are getting excited because your rng fails.

 

[Foster Zygote]

Oh! I see. Leumas has such a poor understanding of probability and statistics that he thinks that it predicts that as more and more coins are tossed, they will somehow magically begin to non-randomly alternate between heads and tails. Like after a billion tosses the results will converge to ...HTHTHTHTHT... for eternity. He thinks that the probability curve generated by a large sample of events, or the tendency for total results of coin flips to converge very close to 50/50 with larger and larger samples, means that the discrete events are somehow no longer random. He apparently thinks that statisticians are claiming that the number of events somehow alters the behavior of the coins, which strikes me as very similar to a gamblers fallacy.

I wonder which derogatory image I'll get?

 

[Dr.Sid]

Made mine in VB6. Not sure what implementation of RND VB6 uses .. but it repeats after roughly 16M calls.

Here's how 100M runs look like (vertical axis is 0.5 +- 0.0001):

This is how 1 billion of runs looks like (same vertical range):

Now does it oscillate, or converge ? And does it say anything about how coin flips ?

 

[Myriad]

A simple question to see if anyone has learned anything in this thread. Here is an integer in the range of 1-20:

6

Is it random?

a. Yes
b. No
c. It depends on how it was chosen or generated
d. It depends on what someone intends to use it for

 

[jt512]

Made mine in VB6. Not sure what implementation of RND VB6 uses .. but it repeats after roughly 16M calls.

Here's how 100M runs look like (vertical axis is 0.5 +- 0.0001):

[qimg]https://i.ibb.co/SKYCfr4/100M.png[/qimg]


This is how 1 billion of runs looks like (same vertical range):

[qimg]https://i.ibb.co/sKB6fjW/1B.png[/qimg]

Now does it oscillate, or converge ? And does it say anything about how coin flips ?

n.m.

 

[Leumas]

Oh! I see. Leumas has such a poor understanding of probability and statistics that he thinks that it predicts that as more and more coins are tossed, they will somehow magically begin to non-randomly alternate between heads and tails. Like after a billion tosses the results will converge to ...HTHTHTHTHT... for eternity. He thinks that the probability curve generated by a large sample of events, or the tendency for total results of coin flips to converge very close to 50/50 with larger and larger samples, means that the discrete events are somehow no longer random. He apparently thinks that statisticians are claiming that the number of events somehow alters the behavior of the coins, which strikes me as very similar to a gamblers fallacy.

Thanks.... so so sooooo much.... QED!!!

I wonder which derogatory image I'll get?

Not derogatory whatsoever... just so very very thankful for your explanation above.

 

[Foster Zygote]

Thanks.... so so sooooo much.... QED!!!

Glad I could help. You see, in statistical analysis, when statisticians talk about the frequencies of certain events converging on a certain ratio of results, they don't mean that it affects the outcomes of individual events. Nothing about the number of previous coin tosses has any effect on the next coin toss. That's why it's a fallacy when a gambler thinks that a number of same results in succession means that a different result is "due" on the next turn. Even though the distribution of heads and tails gets statistically closer to 50/50 with more tosses, if you "zoom in" on any part of the recorded results to look at, say, ten successive results, it won't be distinguishable from any other sample of ten tosses in the same overall series. The statistical convergence is simply a result of the sample size, not any change in the probability of coin tosses.

Not derogatory whatsoever...

You're right. I should have said, "passive aggressive".

 

[Thermal]

What I am talking about is this

Yeah I get that. You've been saying it for pages. You really really don't need to keep cluttering up the thread with the repetition.

There are two themes here: the probability of coin flipping, and the code of your app.

The probability has shown to be precisely what anyone could have told you in the last few centuries.

The coding of your app is not EMPIRICAL EXPEREMENTATION^TM. It's a simulation. A model you've made. Microscrutinizing the code is, I'm sure, interesting to code writers (I haven't done much with programming since the DOS prompt days), because readers might learn a new trick or something. But your actual analysis has nothing much to do with your apparent belief that you are turning the science of probability on its ear. You're really really not, man.

 

[Thermal]

A simple question to see if anyone has learned anything in this thread. Here is an integer in the range of 1-20:

6

Is it random?

a. Yes
b. No
c. It depends on how it was chosen or generated
d. It depends on what someone intends to use it for

No Planet X option. And come on, that's an inexcusable omission ITT.