Human Clairvoyance Project

[DNK]

I have come up with a very simple idea, and would like input.

Basically, there almost certainly exists evidence of whether humans are, on average, clairvoyant. Statistical evidence gathered over countless instance of chance over many, many years, involving humans: casino payout records.

As I understand the casino business, every roulette table will have certain payout ratios. 33:1 for hitting a single number (or 35/36 if they're not cheap), while the actual odds are 37:1 (1/38 (36+0+00)). Likewise, it will be 2:1 for black/red (actual odds 13/38). Unfortunately it is difficult to determine how often people play single numbers versus "board bets" like black or 1/3s, but an average within a certain statistical certainty could be devised. Another problem is gamblers are likelier to put big bets on numbers, I think, rather than board bets so weighting is in order.

Still, with enough observation of actual roulette play, one could find an appropriate weighting. Perhaps it will not be appropriate for play in, say, 1990, though, which brings up another great experiment: given a set structure of gambling, do people respond to different payout rates (34:1 instead of 33:1 really change betting away from board bets?) or is there some set psychological tendency to make certain bets, and does this change over time with culture or IQ scores (assuming they matter)?

Anyway, with these weightings you can then go through years of roulette table payouts and intakes and see if there is any statistical aberration in the long-run, after analyzing for possible discrepancies in the weightings.



Also, human telekinesis could be far more easily studied on slot machine payouts, which would not require the weightings. Just see if they make as much as they should in the long-run. If they pay more than they should, maybe people are affecting the outcomes. Then again, it might not be so simple. Maybe people's negative expectations are having effects as well as positive desires, muddying it all up. Still, any significant move from the expected average should signal something if the sample size is big enough (which it most certainly would be).

Too simple? The last one is simply perfect, imo...

Edit: already issues with the first one: telekinesis could also affect roulette results. Still, again, any significant deviation would signal something is amiss.

 

[Perfection]

Clairvoyance has been done to death. There's no reason to believe it exists.

Waste of resources.

And really, you 've got gambling machine defects, cheats, and systematic recording errors to fight.

 

[hossam]

im clairvoyant.

ask me anything.

 

[Perfection]

Are you 14?

 

[Leonel]

TL;DR




10char

 

[Perfection]

TL;DR

hUH?!

 

[Leonel]

Too Long; Didn't Read.

 

[DNK]

Clairvoyance has been done to death. There's no reason to believe it exists.

Waste of resources.

And really, you 've got gambling machine defects, cheats, and systematic recording errors to fight.

Still, it would put the nail in the coffin.

hUH?!

nubroflmaolollerskates

 

[Perfection]

Still, it would put the nail in the coffin.

Not a particularly good one, plus you have the issue of false positives. Then you have to deal with all the whiny explanations by idiots why it didn't work.

It's a elaborate expensive test that is prone to false results for something that has already been sufficiently shown to be bunk to the scientific community. Just not worth it IMO.

 

[thescaryworker]

To do a study such as this (let's go with the slot machines), there are a few problems that you would need to overcome:


1. Getting a large enough sample size of people (Grouping and increasing sample size as necessary to eliminate variation)
2. Getting a large enough sample size of identical slot machines
3. Proving a statistically significant difference between true randomness and what occurs

The first two are relatively easy and can be done with simple calculations (shown below) and a sizable quantity of money.

(nerdy statistics stuff)

Spoiler :

For simplicity of calculations and planning, let us assume that there is only one group for now (we will compensate for this later). To find out what our sample size needs to be, we need to know the return rate. (Let's go with 95% return, a commonly chosen setting)

But before we find the number of participants needed, let's find the number of trials per person ...

To simulate a 95% return rate AND allow for a net gain (represented by p>1), we'll make a very simple version of slots:
- The slot has two wheels, numbered 0 to 9.
-The given number is produced by using the numbers as digits "5" on the left and "3" on the right would be "53"
- A "win" is from 0 to 18 (and a 19 to 99 is a "loss")
- For a "win", the participant gets $5 back (net gain of $4, but for a loss the participant looses the $1)*

So, Hnull: p = .95 and HA: p != .95
(!= is does not equal)
To find the number of trials for a 99% certainty (and a margin of error of .01) of our experiment, we will need:
2.576 * 2.576 * .05 * .95 / (.01 * .01) = 3151.9936 -> 3152 trials per person (one trial consists of pulling the handle once)

However, this only gives us the probability for one person (the probability that they are able to somehow influence the outcome). Note that there is a 1/100 chance that the test could give a false positive or negative, which is why we want more than one person. Because of this, we expect that each person has a 99% chance of being within 2.576 standard deviations of p=.95.

2.576 * 2.576 * .01 * .99 / (.01 * .01) = 656.94 -> 657 participants

As you introduced the possibility that one's predisposition of the outcome could determine it, we'll randomly assign participants into two equal-sized groups, one with positive, and the other with negative feelings towards the outcome. Since we want to test both possibilities, this becomes 1314 participants total.

* Yes, the 1 Prop Z-test can be used even though there is the possibility of a multi-return trial because it is possible for p to be greater than 1 in this situation

* Note that the "statistically significant" condition is satisfied because we are using a confidence interval

Basically, to be 99% certain of the validity of this claim, we would need a study of at least 1314 participants who each pull the slot 3152 times.

I do not believe that there would be any difference between what should happen and what happens in the experiment, but I have no problem with admitting that I am wrong if a study such as the one described above is done.

 

[hossam]

Are you 14?

no, i am not.

see, clairvoyance.

 

[MobBoss]

I just knew a thread like this would happen.

 

[holy king]

i'd like to post a "thread derailed" picture, but then, it never was on any kind of track anyway.

 

[_random_]

im clairvoyant.

ask me anything.

Where do I live?

 

[Gogf]

Also, if you're using casino payout records as your evidence, you have to ask yourself what would happen if clairvoyance didn't exist. Statistically speaking, there are going to be some people who are just lucky.

 

[MobBoss]

Where do I live?

On the planet Earth.

Next question.

 

[Ecofarm]

A million monkeys and a million typewriters would produce enough character combinations for a Shakespeare tragedy. You could probably do a comedy with one monkey.

 

[Narz]

This reminds me of that random number generator thingy they built that bugs out right before large world events (9/11, the tsunami, etc.), if someone link me to that thread I can't think of.

 

[DNK]

Also, if you're using casino payout records as your evidence, you have to ask yourself what would happen if clairvoyance didn't exist. Statistically speaking, there are going to be some people who are just lucky.

Over time it all evens out, assuming no telepathic abilities. Yes, in the short run a few people get lucky, but in the long run only the house makes money.

Not much of a gambler, are you?

Thanks, thescaryworker!

Now, I'm not sure why this would be "expensive". The casinos likely already have full records of each slot machine and table returns. I would expect such a business to keep track of these things. The only "expensive" part would be monitoring people, but like I said you can perform two experiments with one monitoring, since you can also look into the psychological experiment (probably the more useful of the two) with the same observations. In the grand scheme of things, paying a few undergrads to sit and take notes on betting patterns for a few weeks isn't that costly... if they get "course credit", it's basically free.

 

[thescaryworker]

Thanks, thescaryworker!

No problem. My pleasure.

Though, (and I can't believe I forgot this) we need a control group. This increases our sample size to nearly 2000 people.

Now, I'm not sure why this would be "expensive". The casinos likely already have full records of each slot machine and table returns. I would expect such a business to keep track of these things.

While they do keep track of these things, you have to take into account the variable introduced that a person's predisposition of an outcome could change it. This would require an experiment. Also for the experiment to work, we need a controlled environment. Casinos are not by any means a controlled environment.

The only "expensive" part would be monitoring people, but like I said you can perform two experiments with one monitoring, since you can also look into the psychological experiment (probably the more useful of the two) with the same observations. In the grand scheme of things, paying a few undergrads to sit and take notes on betting patterns for a few weeks isn't that costly... if they get "course credit", it's basically free.

The problem with running a true experiment is that it MUST be random. We have to do a random sample of nearly 2000 people. We cannot allow volunteers, as this could change the results of the experiment. In order to actually test the given claim conclusively, we need to do a random sample of a very diverse range of people. In simpler terms, we need a sample that represents the population. (college grads are not representative of any population other than college grads)

To do this, we need to take these people from, say, the adult population of the U.S. (and the selection process must be random) If we do not do this, any information garnered from the experiment is completely useless.

The expensive part comes with:
- Getting a sufficient sample size to participate (many will refuse to participate for various reasons)
- Creating an environment that is the same (or at least very similar) for each trial for every person
- Travel for the 2000 people
- A place for the 2000 people to stay during the experiment
- Compensation for the time that the 2000 people spend (if we don't compensate them for their time, we will lose a large percentage of our possible sample, making the outcome of the experiment meaningless)

Just with travel, at least $2 million will be spent. The experiment itself will require assembly, maintenance, rent, housing, and many other costs.