04 Mar 2011, 03:18

[quote="G. ADAM"]Great! While we kill 24 hours, I just finished programming a website that does the batch protocol!

Only difference is the 4 words are entered up front. I got 2/4 first go, but going to do a proper set tomorrow.[/url]

Let us know the results. I'd be interested to know how many times you get 0/4, 1/4, 2/4, 4/4. Why don't you do 50 trials and give us the results?

Only difference is the 4 words are entered up front. I got 2/4 first go, but going to do a proper set tomorrow.[/url]

Let us know the results. I'd be interested to know how many times you get 0/4, 1/4, 2/4, 4/4. Why don't you do 50 trials and give us the results?

04 Mar 2011, 10:37

I'll do MIN 5 batches MAX 20 depending if it's working or not.

I'll use a screen recorder and put the first 10 mins on Youtube if any good.

I'll use a screen recorder and put the first 10 mins on Youtube if any good.

04 Mar 2011, 10:56

04 Mar 2011, 12:03

Nobody will take my word for it for one! It's just to find a protocol worth testing by others.

80 words will take about 2 hours already.

At 50% guesses correct, 40/80 it's about 16,000:1 ODDS.

I applied a compensation factor to the binomial, this should be reasonably accurate in the 50% hit rate range, since replacement drops the odds from 25% to 29% in one batch, (>=2 right).

I.e WITH replacement, each word is P=0.25.

ODDS broken are 1/X the last row

80 words will take about 2 hours already.

At 50% guesses correct, 40/80 it's about 16,000:1 ODDS.

I applied a compensation factor to the binomial, this should be reasonably accurate in the 50% hit rate range, since replacement drops the odds from 25% to 29% in one batch, (>=2 right).

I.e WITH replacement, each word is P=0.25.

ODDS broken are 1/X the last row

04 Mar 2011, 12:22

04 Mar 2011, 13:53

I'm not an expert, but it still seems to me that your math is all mixed up on the ordering protocol. You will guess 2/4 25% of the time at chance.

Anyhow, doesn't really matter, we can look at the stats once you've done the trials. If you're getting 2/4 every time over a proper sample, than certainly something interesting will be going on. Course we haven't seen anything like that so far from you. But I'll await the results eagerly.

Doesn't matter if people trust you. Your own trials are preliminary anyway, as you've said, if it looks like something interesting is going on we'll have to do it more controlled. So there really is no benefit for you cheating other than to waste people's time. But we'll be able to see pretty quickly doing the more controlled test what's up. If your private results differ from the controlled results, which do you think people will believe?

So do your trials, and let us know. But please break it down by how many you get of 0/4, 1/4, 2/4 and 4/4

Anyhow, doesn't really matter, we can look at the stats once you've done the trials. If you're getting 2/4 every time over a proper sample, than certainly something interesting will be going on. Course we haven't seen anything like that so far from you. But I'll await the results eagerly.

Doesn't matter if people trust you. Your own trials are preliminary anyway, as you've said, if it looks like something interesting is going on we'll have to do it more controlled. So there really is no benefit for you cheating other than to waste people's time. But we'll be able to see pretty quickly doing the more controlled test what's up. If your private results differ from the controlled results, which do you think people will believe?

So do your trials, and let us know. But please break it down by how many you get of 0/4, 1/4, 2/4 and 4/4

04 Mar 2011, 15:00

Let me know which line you first disagree with.

1 there's 6 ways to score 2/4

2 there's also 1 way to score 4/4

3 the probability of scoring 2 or higher is 7/24. 29.17%

4 with replacement, each word guess is 0.25 chance of a hit

5 by the binomial chart above, the probability of scoring 2 or higher with replacement is 26.17%

6 so, if your score is close to 50%, the difference between a replacement and a no replacement test is 3.00%

7 so all you have to do for a good approximation in the 50% range, is adjust the TotalHits/TotalGuesses binomial calculation - probability for a single hit by 29.17/26.17.

Say I get 35/80.

Trials = 80

Successes = 35

P(success for single guess) = 0.25 * 29.17 / 26.17 = 0.279

I used 0.29 but it's actually very, very close to the basic binomial odds with replacement, 28% Vs 25% for each hit.

This only works in the 40% to 60% hit ratio range.

Personally I don't see the difference between 10,000:1 or 1,000,000:1, and neither will Randi.

Did a couple batches with the screen recorder on, will do a bigger run tonight.

1 there's 6 ways to score 2/4

2 there's also 1 way to score 4/4

3 the probability of scoring 2 or higher is 7/24. 29.17%

4 with replacement, each word guess is 0.25 chance of a hit

5 by the binomial chart above, the probability of scoring 2 or higher with replacement is 26.17%

6 so, if your score is close to 50%, the difference between a replacement and a no replacement test is 3.00%

7 so all you have to do for a good approximation in the 50% range, is adjust the TotalHits/TotalGuesses binomial calculation - probability for a single hit by 29.17/26.17.

Say I get 35/80.

Trials = 80

Successes = 35

P(success for single guess) = 0.25 * 29.17 / 26.17 = 0.279

I used 0.29 but it's actually very, very close to the basic binomial odds with replacement, 28% Vs 25% for each hit.

This only works in the 40% to 60% hit ratio range.

Personally I don't see the difference between 10,000:1 or 1,000,000:1, and neither will Randi.

Did a couple batches with the screen recorder on, will do a bigger run tonight.

04 Mar 2011, 22:30

05 Mar 2011, 09:36

Well in that case I should state I'm just going to start with however many batches I fit into 10 minutes so I can upload to youtube.

Proof of concept more than big odds. You have to assume I didn't take 100 takes.

You can still terminate a test at whim, but the unkown endpoint skews the stats slightly.

Minor hiccup with the screen capture

Will have some data tomorrow! Youtube vid. of me using paranormaltest.com if it works!

Proof of concept more than big odds. You have to assume I didn't take 100 takes.

You can still terminate a test at whim, but the unkown endpoint skews the stats slightly.

Minor hiccup with the screen capture

Will have some data tomorrow! Youtube vid. of me using paranormaltest.com if it works!

06 Mar 2011, 00:30

Hey G.Adam! Can you post your preliminary results from your self test? Curious as to how you're doing.

06 Mar 2011, 13:34

I got 0/4 0/4 but I see some promise!

haha! I'm just making a small change, the first batch when I typed in the 4 random words then did the test I seemed to get some matches / but STILL in the wrong order.

Then I moved onto the second batch "Here's one I prepared earlier".

The channels were too weak so I aborted, then when I had a second look a couple weak channels lined up, "he asked him" <-> "professor".

So I set up 5 batches and waited 2 days... and accidently closed Explorer Window!

Then the screen capture was crippleware when I tested it out, ...

It's my 40th tomorrow so I'll rearrange paranormaltest.com so the SUBMIT WORDS button is under the 4 words so I can do the test without seeing them.

Anyway, I figured out how to get it working.

I did the 1Q at a time batch test with BruceS and AB.

I did GREAT with BruceS, 11 questions, 4 guesses per question, expected was 4/11 and I got 7/11.

I could have done 9/11 too, but let's not go there.

Then with AB we went online at a set time, same test but the channels were randomized and not lined up.

SO, after 2 months I finally figured it out!

You have to

1/ Do 1 question per day. (BruceS took a week to get all 11 done)

2/ Guess all 'words' from all options after that.

That will work I'm sure of it! It fits all the experiments so far.

Anyway, I'll be back in a few days, then we should do the 5 day test. W1, W2, W3, W4, ABCD

haha! I'm just making a small change, the first batch when I typed in the 4 random words then did the test I seemed to get some matches / but STILL in the wrong order.

Then I moved onto the second batch "Here's one I prepared earlier".

The channels were too weak so I aborted, then when I had a second look a couple weak channels lined up, "he asked him" <-> "professor".

So I set up 5 batches and waited 2 days... and accidently closed Explorer Window!

Then the screen capture was crippleware when I tested it out, ...

It's my 40th tomorrow so I'll rearrange paranormaltest.com so the SUBMIT WORDS button is under the 4 words so I can do the test without seeing them.

Anyway, I figured out how to get it working.

I did the 1Q at a time batch test with BruceS and AB.

I did GREAT with BruceS, 11 questions, 4 guesses per question, expected was 4/11 and I got 7/11.

I could have done 9/11 too, but let's not go there.

Then with AB we went online at a set time, same test but the channels were randomized and not lined up.

SO, after 2 months I finally figured it out!

You have to

1/ Do 1 question per day. (BruceS took a week to get all 11 done)

2/ Guess all 'words' from all options after that.

That will work I'm sure of it! It fits all the experiments so far.

Anyway, I'll be back in a few days, then we should do the 5 day test. W1, W2, W3, W4, ABCD

06 Mar 2011, 13:40

Here's the successful test with BruceS, it's just me posting using his alias, the real test in on google groups.

http://hercshome.com/paranormaltest-com ... um.php?f=8

Just substitute QUESTION with WORD, and 4 guesses from 11 options.

http://hercshome.com/paranormaltest-com ... um.php?f=8

Just substitute QUESTION with WORD, and 4 guesses from 11 options.

06 Mar 2011, 13:55

06 Mar 2011, 16:21

Happy Birthday G.Adam!

06 Mar 2011, 17:37