did I open a can of worms?
I knew this was going to be controversial, but perhaps I was not quite prepared for it. I think the problem is that to summarize my 15 years of research in 18 minutes ends up leaving my technical oriented audience needing more science and thinking that it is a bunch of fluff. If I get technical, then I alienate the average person. When I was asked to speak at TEDx in Toronto in November, I was asked to talk about what I am passionate about. I have learned so much about the fascinating power of conscious intenion as a result of my years of study, that I have always felt that this was something that needed to be shared with the rest of the world. I want to share this because I think you should know about it. I have NOTHING to sell by the way. The demonstration video:
and resulting questionnaire::
http://alturl.com/6c3qvWere both sent out to my blog subscribers and FaceBook friends. So far, according to the questionnaire, it was been fairly well recieved:Please rate your level of interest:
43% = "I liked it"
21% = "it was ok"
13% = "it blew me away!"
3% = "snore! boring."but some of the comments are making me realize that I am not providing enough technical science to back up my claims. When I am accused of being "unscientific" about my approach, it really irks me. Anyone who knows me, knows how anally technical I can be. My time limit on the TED presentation is 18 minutes. I can't launch into all of the controls, and statistical analyasis that I do, because I don't have time, and like I said, I would end up alienating 80% of my audience. I'm not so sure I want to do this 18 minute presentation now. I don't think there is a middle ground without turning it into a 2 hour seminar / workshop, or - even better, a book.For those members of my blog audience who need a little more, I totally understand. I would be exactly the same. I have prepared the following (below) outlining the details on the protocol that I use, the controls that are in place to prevent information leakage, and the statistical evidence that there is indeed something going on here. So for those of you interested, that information is below. And for those who have not yet viewed the demo video, I invite you to please watch it, and then read the details below. If you have questions, or comments or just want to discuss, email me adventuresofgreg@gmail.com ---------------
More information about the Princeton Pear program:
http://www.princeton.edu/~pear/
http://en.wikipedia.org/wiki/Princeton_Engineering_Anomalies_Research_Lab Detailed protocol:1. Selection of a market to trade, trade entry date & time and trade exit date & timeThe selection of a futures market that will be the subject of the prediction is done randomly by a computer application which I wrote to manage this entire process. It uses the pseudo random generator to select one futures market from a basket of various futures contracts such as: Gold, Oil, S&P 500, 10 year T bills, Japanese Yen, Wheat, Canadian Dollar, and Swiss Franc. The application selects the contract month with the highest average trading volume from the most recent trading day (active contract), and calculates the number of contracts to trade based on recent volatility to equalize risk/reward between all markets in the basket. Almost all trades are entered at the market open, and exited on the same day at the market close. Each market and each exchange has it's own open and close time, and these exact times are used to place the entry and exit orders. In most cases, the actual identity of the futures market chosen by my computer program is kept hidden from me. There have been times where I needed to preview the market due to ensuring that the current contract month selected was still a valid month for the trading date, and other trading function concerns. However, knowledge of the market being traded isn't required as a control in stopping any information leakage, as the trade direction (buy or sell), is also kept hidden from me. A future date is chosen by myself. This date is any date in the near future that is an actual trading date (ie: the exchange is open for business on that date), and allows me enough time to complete all of the trials. Typically, the trading date is in 7 to 10 days from when I start this whole process. 2. Set up of the trialsI have the computer program automatically set up 20 to 100 trials. Each 'trial' is 2 photographs that are randomly chossen by the program from a database of photo URLs. These images are from a number of different libraries and are generated by another program that I wrote that takes random words from a dictionary and collects images from the internet using search engines from google, Flickr, and various stock photography web sites. The computer application then randomly assigns "BUY" or "SELL" to each image in each set for each trial. The computer application does not allow me to see what the "BUY" or "SELL" association is - that is always kept HIDDEN. This is an important control because if I did know which image was "BUY" or "SELL" then I would be able to have some conscious input on my trading decision as I compared my sketch to both images (Even though, most times, I don't even know what market I am trading). The number of trials that I create for each trade depends on how accurate I want to be with my prediction, and how much time I have available between the set up date and the trade date. Recently, the number of trials has been from 50 to 80. 3. The random thinking process.I will take from 3 to 10 days to work my way through all 50 to 80 trials. Typically for each trial the process goes like this: I sit in a comfortable chair with noise blocking headphones on, and a note pad in hand. I close my eyes, and try to imagine myself sitting in front of my computer screen on the day of the trade, at a time after the market has closed and my trade has either been profitable or not. I imagine looking at my computer screen, and imagine what the photograph for "Trial # 1" would look like. I draw, write, sketch what ever comes to mind on one sheet of my note pad with the title "Trial # 1". As I said in the video, I'm just a random idea generator. I repeat this process multiple times per day for 3 to 10 days until I have completed generating random thoughts for all trials in the trade.4. The sketch comparison processWhen I have completed 'random thinking' from 6 to 10 trials, I will compare my sketch for each trial to the two images that were set up for that trial. I use my computer program to do this. I see 2 images, side by side, and compare my sketch to them. I select either the image on the right, or the image on the left depending on which one I thing more closely resembles my sketch. The program randomly displays one image on the right and one on the left and this changes each time I open the image set to avoid any right/left subconscious preference. The "BUY" or "SELL" associations are KEPT hidden, and are also randomly assigned. There is no way I could know which image is connected to which association, and that control is VERY important because if I happened to know what market was being traded, and also knew which image was associated to up, and down, then I would be able to override the sketch comparison, and use my knowledge of fundamental information for that market, and have a predisposition to choose the image that was associated to that opinion. I make the selection by clicking a radio button, then assign that selection a confidence score from 1 to 4. I assign a score of 1 if neither image resembles my sketch. I assign a 4 if I am confident that the similarities between one of the images and my drawing is not due to chance (rare). 5. determining consensusAfter all the trials have been mediated on, and the sketch comparison process has been completed, my computer app ADDS all of the scores that were assigned to the "Buy" association and "Sell" association, then considers the difference. If the difference in the sums is greater than a predetermined threshold value (about 54% of the trials should confirm one prediction), then I initiate the trade in the direction of the association with the highest sum. 6. Placing the order for the tradeAs part of my application, I wrote an API that connects to my brokers order entry system and submits my order. So basically, my program will decide if there is a strong enough consensus, and automatically place the order entry, and order exit for me without my knowledge of the market or direction. Most of the time, the first time I learn about what market I just traded and what direction I went (long or short), is AFTER the trade has been closed. This is not always the case due to technical reasons. 7. Viewing feedbackTo close the loop, the last step is to view each image from each set that was associated to the ACTUAL outcome. This is the point where, when looking at an image on my screen, I would have the opportunity to influence the outdone of my random thoughts from a week or two ago to resemble the image that I am now looking at. 8. StatisticsMy application also will calculate the statistics for me including export plots, histograms, generate Monet Carlo simulations, etc. I have learned a lot over the years about statistical analysis, and have had a few very highly educated and experienced mentors over the years. I'm currently sitting on the deck of our hotel room in Santa Cruz enroute to LA from Sanfrancisco with Helen by bicycle, and I don't have access to my database to generate some current stats for you. But to summarize, as I indicated in the video, my success rate for trades which are a consensus of individual trials is about 75% over 100's of trades, my Z score is somewhere around 4 standard deviations from chance expectation. The individual trials are 54% correct, and that is a Z score of 5 standard deviations from chance. as I said in the video, that converts to odds against chance of 3.5 million to 1. Since I am not aware of the market being traded, nor am I aware of the direct of the trade (buy or sell), according to commonly accepted physics models, I should be at 50% success rate for both the trades and the individual trials because my predictions (buy or sell) would be just random (assuming that the mentation process is nonsense). This is so far from being chance (4 to 5 deviations), that this should cause you to consider that perhaps there is really something that current physics understanding can't explain. Unless you think I'm lying - which, if you do, then there is really nothing more I can say. 9. External factorsThere are correlations between my effect size and solar wind speed, lunar phase, and other external causal factors - I won't get into that unless you are specifically interested. Again, if you would like to see some plots, let me know, and I'll generate them for you when I get back. Questions and conversation welcome!Cheers,
Greg Kolodziejzyk
and resulting questionnaire::
http://alturl.com/6c3qvWere both sent out to my blog subscribers and FaceBook friends. So far, according to the questionnaire, it was been fairly well recieved:Please rate your level of interest:
43% = "I liked it"
21% = "it was ok"
13% = "it blew me away!"
3% = "snore! boring."but some of the comments are making me realize that I am not providing enough technical science to back up my claims. When I am accused of being "unscientific" about my approach, it really irks me. Anyone who knows me, knows how anally technical I can be. My time limit on the TED presentation is 18 minutes. I can't launch into all of the controls, and statistical analyasis that I do, because I don't have time, and like I said, I would end up alienating 80% of my audience. I'm not so sure I want to do this 18 minute presentation now. I don't think there is a middle ground without turning it into a 2 hour seminar / workshop, or - even better, a book.For those members of my blog audience who need a little more, I totally understand. I would be exactly the same. I have prepared the following (below) outlining the details on the protocol that I use, the controls that are in place to prevent information leakage, and the statistical evidence that there is indeed something going on here. So for those of you interested, that information is below. And for those who have not yet viewed the demo video, I invite you to please watch it, and then read the details below. If you have questions, or comments or just want to discuss, email me adventuresofgreg@gmail.com ---------------
More information about the Princeton Pear program:
http://www.princeton.edu/~pear/
http://en.wikipedia.org/wiki/Princeton_Engineering_Anomalies_Research_Lab Detailed protocol:1. Selection of a market to trade, trade entry date & time and trade exit date & timeThe selection of a futures market that will be the subject of the prediction is done randomly by a computer application which I wrote to manage this entire process. It uses the pseudo random generator to select one futures market from a basket of various futures contracts such as: Gold, Oil, S&P 500, 10 year T bills, Japanese Yen, Wheat, Canadian Dollar, and Swiss Franc. The application selects the contract month with the highest average trading volume from the most recent trading day (active contract), and calculates the number of contracts to trade based on recent volatility to equalize risk/reward between all markets in the basket. Almost all trades are entered at the market open, and exited on the same day at the market close. Each market and each exchange has it's own open and close time, and these exact times are used to place the entry and exit orders. In most cases, the actual identity of the futures market chosen by my computer program is kept hidden from me. There have been times where I needed to preview the market due to ensuring that the current contract month selected was still a valid month for the trading date, and other trading function concerns. However, knowledge of the market being traded isn't required as a control in stopping any information leakage, as the trade direction (buy or sell), is also kept hidden from me. A future date is chosen by myself. This date is any date in the near future that is an actual trading date (ie: the exchange is open for business on that date), and allows me enough time to complete all of the trials. Typically, the trading date is in 7 to 10 days from when I start this whole process. 2. Set up of the trialsI have the computer program automatically set up 20 to 100 trials. Each 'trial' is 2 photographs that are randomly chossen by the program from a database of photo URLs. These images are from a number of different libraries and are generated by another program that I wrote that takes random words from a dictionary and collects images from the internet using search engines from google, Flickr, and various stock photography web sites. The computer application then randomly assigns "BUY" or "SELL" to each image in each set for each trial. The computer application does not allow me to see what the "BUY" or "SELL" association is - that is always kept HIDDEN. This is an important control because if I did know which image was "BUY" or "SELL" then I would be able to have some conscious input on my trading decision as I compared my sketch to both images (Even though, most times, I don't even know what market I am trading). The number of trials that I create for each trade depends on how accurate I want to be with my prediction, and how much time I have available between the set up date and the trade date. Recently, the number of trials has been from 50 to 80. 3. The random thinking process.I will take from 3 to 10 days to work my way through all 50 to 80 trials. Typically for each trial the process goes like this: I sit in a comfortable chair with noise blocking headphones on, and a note pad in hand. I close my eyes, and try to imagine myself sitting in front of my computer screen on the day of the trade, at a time after the market has closed and my trade has either been profitable or not. I imagine looking at my computer screen, and imagine what the photograph for "Trial # 1" would look like. I draw, write, sketch what ever comes to mind on one sheet of my note pad with the title "Trial # 1". As I said in the video, I'm just a random idea generator. I repeat this process multiple times per day for 3 to 10 days until I have completed generating random thoughts for all trials in the trade.4. The sketch comparison processWhen I have completed 'random thinking' from 6 to 10 trials, I will compare my sketch for each trial to the two images that were set up for that trial. I use my computer program to do this. I see 2 images, side by side, and compare my sketch to them. I select either the image on the right, or the image on the left depending on which one I thing more closely resembles my sketch. The program randomly displays one image on the right and one on the left and this changes each time I open the image set to avoid any right/left subconscious preference. The "BUY" or "SELL" associations are KEPT hidden, and are also randomly assigned. There is no way I could know which image is connected to which association, and that control is VERY important because if I happened to know what market was being traded, and also knew which image was associated to up, and down, then I would be able to override the sketch comparison, and use my knowledge of fundamental information for that market, and have a predisposition to choose the image that was associated to that opinion. I make the selection by clicking a radio button, then assign that selection a confidence score from 1 to 4. I assign a score of 1 if neither image resembles my sketch. I assign a 4 if I am confident that the similarities between one of the images and my drawing is not due to chance (rare). 5. determining consensusAfter all the trials have been mediated on, and the sketch comparison process has been completed, my computer app ADDS all of the scores that were assigned to the "Buy" association and "Sell" association, then considers the difference. If the difference in the sums is greater than a predetermined threshold value (about 54% of the trials should confirm one prediction), then I initiate the trade in the direction of the association with the highest sum. 6. Placing the order for the tradeAs part of my application, I wrote an API that connects to my brokers order entry system and submits my order. So basically, my program will decide if there is a strong enough consensus, and automatically place the order entry, and order exit for me without my knowledge of the market or direction. Most of the time, the first time I learn about what market I just traded and what direction I went (long or short), is AFTER the trade has been closed. This is not always the case due to technical reasons. 7. Viewing feedbackTo close the loop, the last step is to view each image from each set that was associated to the ACTUAL outcome. This is the point where, when looking at an image on my screen, I would have the opportunity to influence the outdone of my random thoughts from a week or two ago to resemble the image that I am now looking at. 8. StatisticsMy application also will calculate the statistics for me including export plots, histograms, generate Monet Carlo simulations, etc. I have learned a lot over the years about statistical analysis, and have had a few very highly educated and experienced mentors over the years. I'm currently sitting on the deck of our hotel room in Santa Cruz enroute to LA from Sanfrancisco with Helen by bicycle, and I don't have access to my database to generate some current stats for you. But to summarize, as I indicated in the video, my success rate for trades which are a consensus of individual trials is about 75% over 100's of trades, my Z score is somewhere around 4 standard deviations from chance expectation. The individual trials are 54% correct, and that is a Z score of 5 standard deviations from chance. as I said in the video, that converts to odds against chance of 3.5 million to 1. Since I am not aware of the market being traded, nor am I aware of the direct of the trade (buy or sell), according to commonly accepted physics models, I should be at 50% success rate for both the trades and the individual trials because my predictions (buy or sell) would be just random (assuming that the mentation process is nonsense). This is so far from being chance (4 to 5 deviations), that this should cause you to consider that perhaps there is really something that current physics understanding can't explain. Unless you think I'm lying - which, if you do, then there is really nothing more I can say. 9. External factorsThere are correlations between my effect size and solar wind speed, lunar phase, and other external causal factors - I won't get into that unless you are specifically interested. Again, if you would like to see some plots, let me know, and I'll generate them for you when I get back. Questions and conversation welcome!Cheers,
Greg Kolodziejzyk
