• Quick note - the problem with Youtube videos not embedding on the forum appears to have been fixed, thanks to ZiprHead. If you do still see problems let me know.

JREF too picky? :THOMAS BOWE

deathduck

New Blood
D
Joined
Sep 20, 2007
Messages
12
I think in some cases the JREF is being a little too picky about the protocol. A very good example of this is THOMAS BOWE, Astrologer,

forums.randi.org/showthread.php?t=85841

He said he could tell by facial structure the difference between a fake birthdate and a real one if they were both on the back of a photo.

The JREF protocol insisted that 10 dates be put on the back of each photograph, and if he could not comply (and he said 10 is too many) it would be 'untestable'.

This is lame because you could easily give him a larger number (100) of photographs with 2 dates on them. I don't see why this is a problem, the odds of guessing all 100 are 2^100 = 1:1267650600228229401496703205376. Maybe help him out a little and say he needs 95-98:100 to pass the test.
 
He would still have to do 200 horoscopes for your test, which he says is too many. One hundred photos is excessive of course; a more reasonable test would be 17 out of 20 with two names each. That would be 40 horoscopes. But three out of three with ten names would suffice with only 30 horoscopes. It is not clear whether he thought he could get three of three though.

If he went for three of four with 10 names each that would be 40 horoscopes. Assuming his goal is to do as few horoscopes as possible I think the best he could do would be 10 of 10 with two names each for 20 horoscopes. All these are based on 1 in 1000 odds.

More names on each photo means fewer trials, but more horoscopes overall. We'd have to know his claimed success rate to calculate the absolute minimum number of horoscopes he could do.

The point is that he has to give the JREF that info in his proposal. It sounds like he just didn't want to do enough tests to establish a better than chance record.
 
Christ. For a million dollars I would do 200 horoscopes. I don't know how much time each one takes but that's gotta be worth the investment.

If he really thinks he has this ability I'm not sure how 'it'll take too long' is a credible objection. The only sensible response to that is 'well you can't be too serious then, can you?'.
 
I'm having trouble understanding the issue.

If he uses 20 pictures, the odds of getting all of them correct is over 1:1,000,000. In the past JREF has accepted 1:10,000 for the test, as the odds of winning both the preliminary and the formal test with those odds is 1:100,000,000. So you could go as low as 14 tests if you run 2 trials.

Alternatively, provide only 1 date, but several pictures. The date is only correct for one of the pictures. If there are 5 pictures per date, then 10 trials yields odds of ~1:10,000,000. I am aware that this is slightly different than the claim, and the claimant would need to concur that this is a fair test.

Nonetheless I have a lot of sympathy for Teek's position: when some one starts claiming "too much work" it's clear they don't really believe they have the ability. That doesn't excuse JREF from designing the most concise, fair test that they can.
 
200 horoscopes for a million is $5,000 per horoscope. I wonder how much he gets paid per horoscope in the real world?
 
The original proposal was for using pictures with two or more dates on the back, one of which is correct. If you had 20 pictures each with two names the odds of getting all 20 is 1 in 2^20, which is I believe what you calculated. It's more than sufficient to pass the preliminary test.

The odds for both the preliminary and final test are supposed to be about 1 in 1000, for total odds of about 1:1,000,000. You could actually go with only 10 pictures--2^10=1024.

The other numbers come up when you don't require the claimant to get all ten right. For example, if a claimant feels he can reliably get at least 8 of 10, you can do twenty trials and require 17 right. It is harder to calculate the odds than you might think, because while there is only one way to get 20 of 20 or 10 of 10 there are several ways to get 19 of 20 and even more to get 18 or 17 of 20.

If you have the version where you have 5 names on each picture instead of just two, then there are even more ways to get one wrong. For example, if you have ten pictures and each has five dates on the back, and you get nine of ten correct, there are ten choices for the single picture you could have gotten wrong, and 4 ways to get any one picture wrong. So while there are 5^10 possible ways to pick the names, there is but one way to get all ten correct and 40 ways to get nine of ten, for odds of 41:9,765,625 against getting nine of ten or ten of ten.

It's pretty confusing when someone throws it at you on a message board, but once you get the basic idea you can see how it works.

It is the same odds if you have ten dates each with five pictures of course.
 
Thomas Bowe has only claimed a 58% success rate. The number of photographs would have to be higher as he does not claim to be able to get 3 of 3 correct.
 
Okay then, JREF is definitely not being too picky with the test protocol.
 
He'd have to do on the order of 1000 trials with two names then. I haven't worked out the exact numbers but it's between 500 and 1000--between 1000 and 2000 horoscopes. Even for 10 dates it's still something like 300 trials and 3000 horoscopes. Note that fewer dates always means more trials but ultimately fewer horoscopes. Or to put it another way as you increase the number of dates and each trial gets harder, horoscopes per trial goes up faster than number of trials goes down.

It sounds like this guy has no concept of statistical significance. Add in a few cases where he threw out his results after the fact ("lousy picture") and a few cases where he had the sitter there to give him hints, and you've got 58% easy.
 
58%?? That is so irritating, sorry I ever stuck up for this jerk:rolleyes:. I flip 'up' when playing tennis (to determine who goes first) much more than 58% of the time (I think)!
 
ChristineR said:
The odds for both the preliminary and final test are supposed to be about 1 in 1000, for total odds of about 1:1,000,000. You could actually go with only 10 pictures--2^10=1024.
Click to expand...

Is that accurate? I thought the odds were 1000 to 1 and 1,000,000 to 1 respectively. I can't find a detailed description in the official challenge.
 
Suppose a billion psychics applied for the test. Wouldn't sheer coincidence imply a winner?
 
Is it logical to assume that, with 6 billion people, there will be a daily occurence of an event with a one-in-six billion odds of occurring?
Such a fluke could win the prize, though impossible to do.
And even if it was possible, for instance, to test every person on the planet in an hour, and one gave correct answers, it would not be proof of woo-hoo. In fact, it would disprove wu fairly conclusively, if only one of 6 billion pass the test.

paradoxes give me gas
 
quarky said:
Is it logical to assume that, with 6 billion people, there will be a daily occurence of an event with a one-in-six billion odds of occurring?
Such a fluke could win the prize, though impossible to do.
And even if it was possible, for instance, to test every person on the planet in an hour, and one gave correct answers, it would not be proof of woo-hoo. In fact, it would disprove wu fairly conclusively, if only one of 6 billion pass the test.

paradoxes give me gas
Click to expand...

Perhaps Cuddles referred to the mere statistical chance of someone from a number of 1,000,000,000 applicants to win the MDC.

Dowsers, telepaths, mediums or people like Mr. Bowe could simply get lucky, given a huge number of tries.

However, that would not necessarily infer a paranormal, supernatural or occult event.



Damn you, maatorc.
 
quarky said:
Is it logical to assume that, with 6 billion people, there will be a daily occurence of an event with a one-in-six billion odds of occurring?
Click to expand...

Not really. It is logical to assume that there is a good chance of this happening, but you can't guarantee it will. For example, roll a die 6 times. Can you guarantee you will get a 6? No, but there is a decent chance you will, and the more times you roll, the more likely it becomes.

Such a fluke could win the prize, though impossible to do.
And even if it was possible, for instance, to test every person on the planet in an hour, and one gave correct answers, it would not be proof of woo-hoo. In fact, it would disprove wu fairly conclusively, if only one of 6 billion pass the test.

paradoxes give me gas
Click to expand...

There is no paradox. The challenge sets the bar at odds of 1:1,000,000. If you test a similar number of people, at least one is very likely to pass. This is not a problem because nowhere near that many people will ever be tested, although there is of course still a small chance that someone could win through chance anyway. If it were actually practical to test that many people, all that would need to be done would be to change the odds. If you're going to test 6 billion people, you'd want to set the odds at something like 1:100 billion.
 
FWIW, my quick math shows that even if you tested 1 million people at 1 million:1 odds, there's a better than 36% chance that none of them would win.
 
You must log in or register to reply here.

Back
Top Bottom