This Story covers the entire time line of how I wrote and tested my stock market predicting computer program, as well as explaining why it is necessary to use so much computer processor power in order for it to be as accurate as it is. The Story also explains how the trading system you can subscribe to at this web site has evolved over time as a result of discovering more accurate ways of interpreting the data that comes from Cameron, as well as the accumulation of over 750 Cameron "Data Points" (each day's computer run) over the last three years. We HIGHLY recommend reading the ENTIRE STORY, otherwise you will not understand what is being discussed near The End.

Cameron and Merlin's history:

The original idea for the Cameron computer program was born out of the necessity to find the group-moves to solve a 5x5 Rubik's Professor Cube. From the time I was around 10 years old in 1980, I've known how to solve a 3x3 Rubik's Cube. I used to do it in under 45 seconds, but the 3x3 Rubik's Cube I own right now isn't lubricated with vaseline. So it usually takes me about a minute and a half to solve it.

In 2007 I purchased a 5x5 Rubik's Professor Cube, and there were a few moves I needed to know in order to solve it that weren't part of the solution for the old 3x3 cube. If you've never solved a Rubik's Cube, you should know that the way it is solved has to do with moving one piece on the cube to another location without disturbing certain other parts of the cube. So when I say Rubik's Cube "Move," I'm actually referring to turning several different sides of the cube around 10 times in order to accomplish one of these tasks of moving one piece on the cube to another location without disturbing certain other parts of the cube. And of course, solving the entire cube consists of doing several different types of these moves many times.

When I bought this 5x5 Professor Cube in 2007, I knew I could go online and easily find those moves on a web site somewhere. However, I was bored at that time in my life and not really doing anything creative. So I decided to write a computer program using the Basic Programming Language to "Randomly" find these moves. And at that time in 2007, I hadn't dabbled in computer programming since around 1985 when I was about 15 years old.

My FIRST computer in 1982, when I was 12 years old, was the Timex Sinclair 1000. I used it mostly for writing Basic computer programs. I later had an Atari 800XL (basically a clone of The Apple IIC), which I would also use to write Basic computer programs. These home computers from over 25 years ago wouldn't have been capable of running a program like Cameron, or even its predecessor Merlin. In fact, that first computer I owned, the Timex Sinclair 1000, would have taken over 4 years to crunch the numbers that my two Intel Dual-Core laptops can crunch in 16 hours. Completely impractical when you consider I have to enter the day-end market data after 4:00 pm, and I need an "Answer" by 9:00 am the next day.

Getting back to the 5x5 Rublik's Professor Cube; My idea for a Basic computer program was based on the fact that in 2007, computers were already fast enough to turn the sides of a 5x5 Rubik's Cube about 500 million times per hour as long as the program wasn't drawing the cube to the screen. The theory was akin to running a computer program that randomly picks lottery numbers. And due to the fact that computers were so fast by 2007, it would only take an hour or two to spit out a coincidental match to all six of a specific set of numbers that were entered into the beginning of the computer program.

So I went online and purchased a Windows program called PureBasic that runs a Basic Programming editor within Windows. This is sort of like having an Old computer simulator," but it runs the Basic Programming language on any of our newer PC's about 12,000 times faster than those old home computers did.

First I wrote a program that was a 5x5 cube simulator, where pressing certain keys would turn certain sides on the cube. (Right) This program actually drew a colored text version of the 5x5 cube to the screen so you could see the moves and see the pieces. Then I wrote a subroutine that automates this and makes 20 moves randomly, but I removed the "Drawing to the screen" part of the program, which drastically speeds up the process of making and analyzing the random set of moves. Next I added a part to the program where I assigned a marker to a specific piece on the cube to track where it goes, and then I wrote some code lines that run a loop of those 20 random moves over and over again.

So for each cycle of the loop; The cube is reset to solved, the marker is set on that one piece, the 20 random moves are made, and then the program checks to see if that piece was moved to a specific spot on the cube without disturbing certain OTHER parts of the cube. Then this process is repeated hundreds of millions of times per hour.

It worked. The first time I ran the program to move a piece from one place on the cube to another without disturbing certain other parts of the cube, it took about 45 minutes and 20 million sets of those 20 random moves before it just happened to find the right group of side-turns. And I thought that was the coolest thing since peanut butter because I was reflecting on the fact that my old Timex Sinclair 1000 would have taken weeks to find this one group of side-turns, and I needed SEVERAL of them.

So what does a 5x5 Rubik's Professor Cube have to do with candle charts and predicting moves in the stock market? I'm getting there.

Fast-forward six months. (Late 2007) I was trading stocks in the market, and not doing too well of course. It always seemed like I was taking two steps forward and two steps back. Or worse... two steps forward and THREE steps back. All the while, wishing I had a simple and reliable stock trading system in which at the very least, I could take two steps forward and ONE step back.

Then one day I was looking at a candlestick chart for a particular stock and I saw a 2-candle pattern that repeated itself 3 times over a 2 month period preceding a drop in the stock's price. And just so you know, I'm a very analytical person. And as was mentioned on the home page, I score in the "Analytical Detective" category on IQ tests.

(If you need a quick 101 on candlestick charts, here's a short Wikipedia article)

So looking at this candle pattern, I decided to read up on candle charts and patterns. The first thing I thought to myself was, "There's no way I could ever remember all these patterns, and I'll NEVER have the time to pour over candle charts looking for needles in hay stacks... not while working a full time job."

I also knew, and this is extremely important, that if there ever existed a piece of over-the-counter, simple candle chart analysis software that works too good at finding candle chart signals, hundreds or even thousands of day traders would likely be chasing the same trades in any one particular stock, and it would eventually cancel out the profits of at least 51% of those traders if the computer software became too popular. So looking to BUY a piece of software for this was never even a consideration of mine.

Now don't get me wrong; I'm not saying there isn't good candle chart analysis software out there that "Works well." I'm simply saying that I was thinking on a needle-in-a-haystack level, where candle patterns might exist that have nothing to do with the most well known candle chart signals, and that these patterns might sometimes exist on a micro-level that can't be seen by the human eye or even by the most analytical person.

So while I was reading up on the candle charts, I wished for a moment that I had a computer program that could find candle chart patterns in stocks similar to the 2-candle pattern I had recently found in that company. I already knew that this candle pattern only worked for that SPECIFIC COMPANY, because I did a quick check through several pages of previously known candle patterns and it wasn't there. For what it's worth, this candle pattern included having a certain sized lower shadow on one of the candles. So I knew that a piece of software would have to run with this theory. And I also knew that most software probably DIDN'T run with that "Out of the box" theory, and was using the basic candle chart patterns that are known to most people in the stock market... and using upper and lower shadow sizes that are generic. (From here on out, we'll refer to these generic candle patterns as existing in the "Candle Pattern Dictionary")

On another analytical level, which doesn't necessarily have to do with a candle's wick size; Who's to say that an engulfing candle pattern has to literally "Engulf" in Microsoft's (MSFT) candle chart? Maybe the lower portion of the body on the second candle can be .5% higher than the body of the candle behind it, but still have the same bearish or bullish meaning as what the Candle Pattern Dictionary says? And maybe this is SPECIFIC to MSFT's candles. I was thinking on a level that maybe patterns are appearing in particular stocks with certain size shadows and other combinations of candle placement, candle size, filled and hollow, that aren't in the Candle Pattern Dictionary.

So next I thought to myself, "Damn... a computer would have to analyze Open-High-Low-Close data for a stock MILLIONS of times and using MILLIONS of calculations for candle placements... and using MILLIONS of different random numbers that represent different candle placements, the sizes of candles and the sizes of upper and lower shadows to find this stuff." And again, I knew that the software being sold at the consumer level wasn't doing this, mostly because of the fact that the majority of a person's computer processing power would be needed for several hours at a time just to analyze the candles of ONE STOCK.

Then it hit me like a ton of bricks. I thought to myself, "I'll just WRITE a program like the Rubik's Cube program I wrote!" Within a few hours after that thought, the first version of my program, "Merlin," was born in late 2007.

As the idea for this program was taking shape in my mind, I was thinking on a Holy Grail level, with the goal of finding a way to make between 5% to 15% per month with almost 100% confidence in my computer program and requiring very little time and effort after the program was written, tested and running smoothly. Again; Requiring very little time and effort after the program was written, tested and running smoothly. (We'll get to the reality of this goal later) And don't be confused, when I refer to Cameron as "The Holy Grail of Candle Chart Analysis Technology," I simply mean a computer program and trading system that always makes more money than it loses, with a trade accuracy ratio of 4 to 1, 3 to 1 or even 2 to 1... as long as it takes more steps forward than the steps it takes back, and as long as the monthly-average for the last 12 months is ALWAYS positive / profitable.

Interestingly; If I already had $100,000 dollars at that point in my life (late 2007), I might have never even felt the need to write a program like this. If I already had that kind of money, I probably could have and would have just sat at home for 8 to 12 hours a day doing nothing but due diligence, home work and stock trading, and I probably would have been a millionaire within 20 years WITHOUT needing a computer to find "Holy Grail" candle patterns. However, I had to keep working a full time job at this point in my life and I didn't have much money. So doing that much stock and stock market homework every day wasn't possible. I needed something else to do it FOR me. So there is a certain irony about the fact that this computer program (the eventual "Cameron" program) wouldn't exist and I would have never written it unless I was in the lowest income bracket at this point in the timeline. (2007)

So as I wrote the "Merlin" program, I spent about a week looking at different candle charts and finding patterns on my own. I also spent that time dreaming up DOZENS of patterns. I thought of every possible way that a candle could be situated next to another one, and completely ignoring The Candle Pattern Dictionary. Remember, I knew I was dealing with processor power that in 1981 would have been considered like "Star Trek" technology. So I programmed Merlin with DOZENS of random candle rules... filled or hollow, filled or hollow being in a certain order, upper and lower shadow sizes, different candle sizes and different candle placements. And I did this without any regard to what was previously known about candle chart patterns, and without any regard to how long it would take a computer program to find patterns. Again; Without any regard to what was previously known about candle chart patterns. I NEVER did any in depth reading on the previously known candle patterns. So in a sense, my mind was never "Poisoned."

Some of those already-classified candle chart patterns might very well be FOUND by my program from time to time, because I programmed the software to literally analyze the candles 150 million ways from Sunday. However, I let the PROGRAM find them and let the PROGRAM determine if they are relevant.

This first version of my program, Merlin, used the daily Open, High, Low and Close data for individual stocks, and took too long to run tests and took too long to find stocks that worked well with the program. However, when a stock DID work well with the program, it worked EXTREMELY well. So I knew I was onto something. The way Merlin worked was it looked for the patterns, and if it found a match, it triggered a Buy On The Open. Then in the following 1 to 7 days, I would run each day's newest candle (and all the candles behind it) into a similar version of the program to see if a "Sell" was triggered. In other words, to see if the likelihood existed for the stock to start going down again.

This didn't work too well, mainly because if the markets in general went down, so did most stocks. So eventually I decided to have my program look for Buy clues that would ALWAYS produce a certain percentage within 2 or 3 trading days, and use that exact percentage as a Sell Limit rather than the method in the paragraph above. If you read that paragraph above, you will find a hint of "Greed" in it, which was one of the reasons why it didn't work as well. In other words, I was basically telling the program to buy Stock ABC and try to get as much as possible out of it before triggering a sell. A better theory was to find certain candle patterns that repeat over time and produce a certain percentage in every instance, and then get the hell out before something goes horribly wrong. This bold sentence was one of the keys to eventually making the final version of my computer program, Cameron, work so well.

But still, this Merlin version had its weaknesses, and not just because of the market going down leading all stocks down as well. The other weakness was that my program couldn't predict a company-specific news event, like, "Company ABC is being sued by Company XYZ." Or, "Company ABC lowered its Q2 earnings estimates." These examples are the reason why the Gordon Gecko's of the world try to get inside information.

Before retiring Merlin, I experimented with the program using the daily candle data from six different leveraged ETF's. They were the 2x Dow up and down (DDM / DXD), 2x S&P up and down (SSO / SDS), and the 2x Nasdaq-100 up and down (QLD / QID). And I didn't know it at the time, but Merlin was not being run for enough cycles to find the maximum number of candle patterns within any given set of data and based on my complex algorithm. I was using around one year's worth of data and running Merlin all weekend to come up with a set of candle patterns that would work for the following week. Regardless, this idea of using the forward and inverse ETF's was the beginning of creating a simple trading system that would eventually mirror the operations of a hedge fund.

Also of important note; When I was experimenting with the DOW, S&P and Nasdaq-100 ETF's mentioned above, Merlin seemed to be NOTICABLY more accurate at predicting the Nasdaq going up OR down, but wasn't as accurate at predicting both directions of the DOW or the S&P 500. This "Nasdaq secret" will be covered below later in the story.

(For a quick 101 on Short Selling and Forward & Inverse ETF's, here's a short eHow.com article)

Regardless of Merlin's limitations and not being run for enough "Gigahertz CPU Hours," the accuracy of the data coming from the computer program was still quite unusual in certain instances. Sometimes Merlin would spit out information that suggested a sharp drop in the markets. From watching financial news media every day, there were no perceived warnings of these drops. And yet, Merlin nailed most of them. The same thing would also happen sometimes with Merlin's data suggesting a rally, with the same "Out of the loop" clueless commentary on TV, and with the same success in predicting most of the rallies. This was actually the reason why I had the idea of using forward and reverse ETF's that follow an entire market rather than an individual stock.

That was when I came up with this theory that pressure might be building up in the market in one direction or the other, that can accurately predict if markets are about to go up or down in the short term. And if my computer program could get an accurate reading on that pressure number, THAT might be "The Holy Grail." So I thought about the possibility that maybe my program would work better for predicting an entire market rather than a single stock or ETF. And maybe this actual market data, like for the Nasdaq (as opposed to using market-following ETF data), contained a more accurate version of the information from the behavior of the major institutions and hedge funds... which are likely the individuals who give off this "Pressure Reading."

So I experimented with Merlin and different versions of it for about another year on and off. Then one day in early 2009 I decided to use the exact Nasdaq levels in my computer program due to the high likelihood that the Long and Short market-following ETF's probably weren't giving off an accurate market pressure reading in their daily movements. I actually woke up out of a sleep with this idea in my head and couldn't shake it. (This often happened in during the two year period of time in which I developed my computer program)

After waking up with that idea in my head, it was the morning when Cameron was born. And I decided to only use my computer program with the actual Nasdaq daily candle data. (Daily Open, High, Low and Close) I also decided to use a different way of analyzing the daily Nasdaq data within the guts of the program, and also using a different way of handling the data that comes OUT of the program. So for all intents and purposes, Cameron was a different computer program at this point and deserved a new name.

The reason the computer program is named Cameron is because it uses artificial intelligence in a different and more powerful way than Merlin did. The name Cameron comes from a female Triple-8 Terminator in a Terminator TV show called "The Sarah Connor Chronicles." And the way the artificial intelligence works in my computer program will be explained below when I get into the method I use to interpret the data that is now spit out by Cameron, and I will also explain how the program can be expected to smoothly transition from a Bull Market to a Bear Market without being tripped up. (Also as a result of this Artificial Intelligence)

Also in this new Cameron program, I planned on running it for equivalently more time than with Merlin. SIX TIMES longer. Basically using my two dual-core laptops for most of the 17 and a half hours between market trading days. So there was no time to run experiments using the S&P data or data from any other sector. And remember; For Cameron, I chose the broader Nasdaq daily candles (Nasdaq daily Open, High, Low and Close) because of the unusual aspect of Merlin in which that program seemed to be better at predicting BOTH directions of the Nasdaq... up OR down.

Here's what I will say about how Cameron works at this point in this story, without divulging too much proprietary information..

Cameron uses a certain period of rolling, daily Nasdaq candle data. (Can't say what this period is... maybe 100, 200, 400 or 800 trading days) The program runs in cycles... about 150 million cycles in total on both laptops. For one cycle, Cameron chooses 92 random "Candle rules" (the program chooses 92 random numbers that will either be a 0 or a 1), and Cameron also chooses another set of random numbers (won't say how many) that represent the sizes of certain candles, and the sizes of the upper and lower shadows of the candles being analyzed and compared to each other. After these random rules and random numbers are chosen, the program runs through that period of previous Nasdaq data and does something hypothetical. (Can't say what that is) If a certain type of candle pattern and group of chosen rules is found to repeat "Successfully" in comparison to the most recent 2, 3 or 4 trading days' candles (won't say how many recent candles), then this is considered a "Hit" and the program prints a hypothetical "Number of Trades" to the computer screen. (You'll see this below)

If during this one cycle, the patterns created by the random candle rules and random numbers related to those rules DON'T match the recent 2, 3 or 4 candles in any way, and if the "Successful" aspect of the program isn't hit (which I can't detail for you), the program loops back, resets all variables, and then chooses another set of random candle rules and random numbers relating to those candle rules... hence another one of the 150 million cycles has begun. (Just one of these cycles only takes .0007 of one second on our modern computers)

During this 16+ hour time period, as Cameron is running on two dual-core laptops, 500,000 random numbers are chosen PER SECOND and many more calculations are done in between, as the program runs through that loop of previous Nasdaq data over 5 thousand times per second. After 16 hours, over 20 BILLION random numbers will have been chosen and the program will have looped through the previous daily Nasdaq candle data over 150 million times to come up with "Tomorrow's Cameron numbers," which are analyzed by other computer programs that revolve around this original Cameron program.

As far as INITIATING Cameron, here's what I do at 4:01 pm on any given trading day; After a trading day has ended at 4:00 pm, I open the Cameron program and chop off / delete the oldest day of data. Then I add today's resulting Nasdaq data, like this...

ElseIf day=1488 : date$="13-Oct-09" : top=173460 : thi=184425 : tlo=171574 : tcl=184425 (Nasdaq open, high, low, close)

By the way; The day number "1488" above doesn't have anything to do with the number of days that are analyzed. It's actually just a number that always goes up one digit for every trading day. (Not calendar days) This number has been reset to 1,000 once or twice over the last couple of years.

After entering that one line of data and chopping off the oldest one, I run Cameron at around 4:05 pm and the program spends over 16 hours blindly coming up with the guess for the next day, and it does this before 9:00 am. Most importantly; I do not add or change anything within the guts of the computer program. Each day, I only add the code line you see above to run Cameron for the NEXT day. Nothing about the way the program works is changed once a testing period begins, and nothing in the guts of the program has EVER been changed since this 2009 version of Cameron was written. If I ever DID make changes to the guts of the computer program during that time or any other time, it would change any day's resulting data with regards to the way that data is relative to all PREVIOUS days that were run. This would mean that a testing period would have to start all over again.

So in layman's terms, the program analyzes that previous set of daily Nasdaq candle data, which actually just exists within the program as Open, High, Low and Close numeric values, and looks for as many reasons as it can find, AKA: "Repeating Patterns," to justify "Buying on the open" the next day. This "What to do the next day" was different than the way Merlin worked, and would be one of the important changes that resulted in the Cameron computer program being so successful.

At this point, around March of 2009, I ran the back-testing for Cameron. And I should let you know at this point that "Back Testing" isn't really a good term to use for this, as some people perceive this as meaning that something is known about the future and the program is initiated with knowledge of this information. Or that when something doesn't work right, the program's guts are changed and testing continues without "Starting Over." Nothing of the sort happens. The computer program is 100% analytical.

For example; I basically pretended it was 4:01 pm on April 15th, and initiated Cameron for a 16+ hour run using April 15th's data and all of the previous days of Nasdaq data before it, to make the prediction for April 16th. If the April 16th prediction was correct, based on what ended up happening on April 16th, it was logged as a hit. If it was wrong, it was logged as a miss. I then continued doing this day after day until at least a few months of data were complete. No tinkering with the guts of the program were done along the way. It was very time consuming because there was/is no short cut or method of "Simulating" the 16+ hours of number-crunching and random number-picking that would be done at any time in the future when Cameron would be used for actual trades or market predictions.

One of the things that the Cameron computer program does is it turns the Nasdaq data upside down into a flipped version, much like when you view two different charts of a Short and Long ETF. (As seen on the right) This also results in a filled candle becoming a hollow candle and vice versa. This will be better understood below, as I explain why the data is flipped/inverted.

For this new Cameron computer program, I also started saving to my computer and also printing out 15-minute S&P 7-day candle charts because I planned on analyzing the data to see if Cameron might be able to predict the market's movement the next day. (For in-and-out Day Trades) And you need intra-day data to see how far the market goes in the opposite direction of the Call before going in the PROFITABLE direction of the Call. (For setting Stop levels) This sort of performance analysis can't be done with DAILY candle charts.

I needed these charts so I could design the trading system with Limit Sell orders and Stop orders. And in design, the Stop would be closer to the buy-in level than the Sell Limit order. (Example: Stop at -.66% from buy-in and Sell Limit at +1.00% from buy-in) Using this method, and in a worst case scenario, Cameron could theoretically miss every other call and still stay ahead because each hit should profit more than each miss over the long term. (Losing 3 units for every 4 that are gained)

Around the middle of April of 2009, and after just 4 weeks of testing (about 5 trading weeks), it became obviously apparent to me that Cameron was EXTREMELY more accurate than Merlin. In fact, the data from Cameron could be analyzed and translated so that a trade could be carried out within one trading day. And due to this "Playing out the same day and before market close" aspect of Cameron, it would be much easier for the average trader to carry out the trades if I ever decided to provide the information through a subscription service.

At any rate, it turned out that Cameron was now correctly making 2 out of 3 Calls during most stock market conditions, and losing considerably less money during rough periods than what was gained during the extremely profitable times. And the average profit on a good Call was HIGHER than the average Loss from a BAD Call.

This was incomprehensible to me. You have to understand, after being at this for two years at this point in the timeline, this 2 out of 3 performance was akin to the moment when the first plane flew, the first light bulb lit up, or the first time a voice was heard coming over a metal wire. And ALL of it was being done in a way that was equivalent to being locked in a room with no outside contact, and only using the Nasdaq Open, High, Low and Close every day to make any decisions... all of which comes from a computer program. ("AI") No market volume or other data was being analyzed, and no market opinions were injected by ME.

But let there be NO ILLUSION; Cameron and the analytical technology that makes the program work is SUPPOSED to miss about 1 in 4 to 1 in 3 trades over a long term average. This is EXTREMELY important to understand because the daily predictive ability is NOT meant to be portrayed as "Perfect" or even "Near Perfect." The term "Holy Grail" only refers to making a consistent average MONTHLY percentage profit over a long period of time that is MANY times more than what other Long-only funds or the average bank would give you for simply holding your money. And it also refers to the high likelihood of AT LEAST making DOUBLE the performance of the S&P 500 or the average Hedge Fund during any rolling one year period. If there are ups and downs along the way, that is just fine and should be expected from time to time.

With all that said, now is a good time to explain something else about how Cameron works, explain the numbers that Cameron spits out, as well as explain how I analyze those results during testing so you can see that there is no data mining going on. In case you didn't know; Data Mining is a term used by The FDA to describe a process of discovering information within a given set of data over a past period of time, that doesn't necessarily have to do with the original experiment that was being carried out. And as a result, may not repeat further successful results in a period of time that follows.

But first, if there's still some doubt in your mind that I discovered a Holy Grail of sorts, let me show you a quote from the description of Renaissance Technologies Hedge Fund. This was pointed out to me by someone at a stock market message forum AFTER I wrote my computer program, and gave me goose bumps the first time I read it...

Renaissance Technologies is a hedge fund management company. Started in 1982 by James Simons, Renaissance currently has more than $15 billion in assets under management. Since 1989, the company's $5 billion Medallion Fund has averaged 35% annual returns, after fees, and the Medallion Fund is said to be one of the most successful hedge funds.

For over two decades, Simons' Renaissance Technologies hedge fund, which trades in markets around the world, has employed complex mathematical models to analyze and execute trades—many of them automated. Renaissance uses computer-based models to predict price changes in easily-traded financial instruments. These models are based on analyzing as much data as can be gathered, then looking for non-random movements to make predictions.

Renaissance employs many specialists with non-financial backgrounds, including mathematicians, physicists, astrophysicists and statisticians. About a third of the more than 200 employees at the East Setauket office have Ph .D's.

The words "analyzing as much data as can be gathered, then looking for non-random movements to make predictions" is EXACTLY what Cameron does. And the fact that I was able to write an algorithm without tainting it with "Previously known stock market behavior or candle behavior" is an example of how the company above hires mathematicians who have no financial background. MY experience with the stock market came in the time AFTER I wrote the Cameron algorithm that I use to analyze the daily Nasdaq candles. This is what enabled me to transition over to using market-following ETF's and running a subscription service that is easy for almost any trader to follow.

So another thing I'll be telling you about the way Cameron works is shown on the right. (Click to enlarge) That's a full screen shot of what it looks like on one of the dual-core laptops while Cameron is running. We'll ignore the OTHER laptop because it is actually doing the same exact thing as what I'm about to explain concerning the laptop shown in the image. The reason I use TWO dual-core laptops to achieve the maximum data numbers that I'm about to speak of is because using two computers greatly increases the accuracy of the system due to something proprietary about how the computer program works... something I can't talk about.

The laptop shown in the image is running the Long side of the equation (Left, with the green colored results), and the Short side of the equation. (Right, with the red colored results) These red and green numbers represent those "Maximum hypothetical trades" numbers I spoke of earlier in this story. As you can see, two instances of my computer program are running to come up with these 2 maximum data numbers... one from the Long side of the equation and one from the Short side of the equation.

By the way; Cameron works very well on dual-core processor computers and quad-core computers because Windows directs exactly 50% or 25% CPU power to each PureBasic program instance. This is EXTREMELY important for overly-analytical reasons because everything about the way the program functions has to do with finding a maximum data number based on a certain number of program cycles, and using a certain amount of previous daily Nasdaq candle data. In other words; It wouldn't be logical to run a certain exact "Number of cycles" because each day's run ends up running a little faster or a little slower due to what the program does or does not find in every cycle using a different day's data. Therefore, using an exact amount of "Time" each night for the program instances is the only way to achieve a constant, relevant data set that is based on the previous set of Nasdaq data being equally analyzed day after day. And all of THAT is organized into an equation based on the processor speed of a particular computer, which determines exactly how long Cameron is run on any particular computer... or even how long Cameron is run on a particular processor CORE on a computer. (One of the processor cores in a dual-core or quad-core computer sometimes runs a little slower than the other/others)

So the point I'm getting at by showing you the screen shot above is that the SAME PROGRAM is running on both sides in the graphic above... the "Cameron program." There is nothing in the guts of either program instance that is telling Cameron to act differently in bias of Short or Long. The only difference between the programs within these two instances are these two code lines existing in the Short-side program on the right...

top=(mult*2)-top : thi=(mult*2)-thi : tlo=(mult*2)-tlo : tcl=(mult*2)-tcl

tthhii=tlo : tlo=thi : thi=tthhii

There is also, of course, a color code change when the highest trade is printed on one instance compared to the other. (Red or Green) And an ID tag that says "Nasdaq - Long" or "Nasdaq Short."

So do you know what the two program code lines are above? Those are the calculations that flip the Nasdaq data for the program instance on the right in the image. So actually, BOTH instances... Short AND Long, are looking for reasons to "Buy" tomorrow. The only thing I need to know for my Cameron data analytics is that the red-numbered results from the program instance on the right represent the likelihood that the market will go DOWN because it's using flipped/inverted Nasdaq candle data.

The reason I'm stressing this "Same program on both sides" aspect of Cameron is to try and drive home the fact that this computer program almost literally has nothing to do with the stock market, except for the fact that previous Nasdaq data numbers are being analyzed, as well as the fact that I apply a few logical "Data Interpretation Rules" to the results. This is important to know because most other software isn't 100% analytical and 0% biased to Short or Long. My computer program literally doesn't even KNOW it is a stock market analysis program. :) I can be fairly certain that most other software is programmed to "Think too much" based on stock market history and stock performance history when dealing with candle data, and also calculating basic candle analysis within a few seconds using what is contained in the Candle Pattern Dictionary. Cameron runs for over 16 hours each day using four computer processor cores, using hundreds of millions of random candle patterns, and COMPLETELY IGNORES what people THINK they know about candle patterns.

The screen shot above of the computer program is a few hours into the 16 hour run, but for the purpose of now explaining the data analysis to you, we'll pretend that both laptops completed their tasks, and that the maximum number from all four instances on both laptops and for each side of the equation was 24 / 4, as shown in the image. In simple terms, and what this means is, Cameron was able to come up with 24 reasons to go Long in the stock market tomorrow and only 4 reasons to go Short. These two numbers are then "Corrected" to represent numbers that can only be 2 to 98. (I can't explain this any further for proprietary reasons... only to say that the process of correcting the numbers is logical and relative to all past Cameron runs)

So now we have two numbers that might be 46 / 8, and the smaller number is divided into the larger number to come up with a ratio. In this example the ratio is 5.75 and it is a Long Ratio. (Because the Long/Left number is higher) Remember, the number 5.75 has nothing to do with how much the market is expected to move the next day. It is actually a "Buying or Selling Pressure Reading."

So this is a 5.75 Long Call, which would be a "Called Trade" as long as the S&P 500 opens UP a certain small percentage amount on the morning of the Call. Otherwise the Call is canceled.

Now take a look at the graphic on the right. (Click to enlarge) These are some of the results from Cameron as the program was running Live in November of 2009. The only numbers that come out of Cameron are under the red and green dots at the top. Everything else is either numbers representing what happened in the Nasdaq on that date, the ratio number that is calculated from the two numbers under the red and green dot that are spit out each morning, as well as whether or not the Call was Long or Short. (The red and green dots in the column on the right) Remember; Cameron's result numbers and ratios to the RIGHT side of the date, are what is spit out early in the morning BEFORE the trading day of that date begins. The trade performance/profit is written on the right side of the page, and no losses were taken on during this time period.

The first few months of computer runs, July to September of 2009, are what were used to come up with the Data Interpretation Rules below. (Referred to here-on as "DI Rules") There were four DI Rules at this point in the timeline, and I will share two of them with you.

"Buy Call" Rule: Buy Ratio >1.15 and a Short number (under red top dot) <27. Market must open up between .18% and .81% from the previous day's Close

"Short Call" Rule: Short Ratio >1.15 and a Buy number (under green top dot) <27. Market must open down between .28% and .70% from the previous day's Close

As you can see, there is logic in all aspects of those rules, and nothing close to "Data Mining" or "Finding what you want to find " is involved with the creation of those rules. They are determined from "What worked" in the first few months of computer runs and using logical reasons related to what the program is supposedly trying to do... which is to come up with the likelihood of the market going up or down from the open to the close on the next day.

These DI Rules were used for all of the days that followed the first 3 month period and were adjusted from time to time as weeks went by and more market behavior data was gathered. However, if a new "Range" appeared to work and fit within the logical structure of the DI Rules, that first trade was not counted as a "Hit." Only the Calls triggered AFTER that, and in real time, would be considered a profitable trade for the system. Otherwise it would compromise the integrity of the program's performance. The same goes for a FAILED Call, and an opening range in the DI Rule needing to be adjusted as a result. The failure was accepted and reflected in Cameron's "Daily Call" Performance Data. But from that point on, the DI Rule will be slightly different.

The logic behind the reason that the market can't open up too much for a Long Call and can't open down too much for a Short Call is because some of the likely .66% or 1.0% predictable move in the market might already be "Baked in" to the pre-market activity and the level where the market opens at 9:30am.

The third and fourth DI Rules in Cameron's Daily Call trading system at this point in the timeline were proprietary to how the program worked, and were the result of what I considered to be hard work and creative analytical thinking on my part. (They must remain a secret) I can assure you however, that they WERE just as logical as the two rules above.

Now let there be no illusion, the daily predictive aspects of Cameron are NOT supposed to make 2% or 3% using the leveraged ETF's on every confirmed Call and trade. The REAL purpose of the program is to spit out a "high likelihood" of the stock market going up or down .66% or 1% from open to close, and an EXTREMELY high likelihood of the market at least closing in the direction of the confirmed Call with a profit from open-to-close. (Without hitting the Stop) Therefore, some of the profit results from daily trades triggered by Cameron could be in the range of .50% to 1.50% during some market conditions. The most POWERFUL aspect of Cameron is the fact that the daily Long or Short ratio number represents an EXTREMELY accurate "Buying Pressure" or "Selling Pressure" reading. This is almost akin to a barometer, or maybe "Buy-rometer" for the US stock market. :) This pressure reading accuracy and an unbelievable, as yet explained predictive ability of Cameron is explained in more detail later in this story.

So now that I've explained the logic of the DI/Data Interpretation Rules, here are some exaggerated examples of "Data Mining" with regards to the creation of DI Rules within a trading system that uses this kind of Artificial Intelligence. This is for those who aren't too analytically savvy...

"Buy Call" Rule: Buy Ratio >1.15 and a Short number (under red top dot) <27. Market must open up between .13% and .76%. Cancel Call if it is a Wednesday.

The day of the week should have NOTHING to do with and should be of NO concern to the market or a powerful technical analysis tool. Another example of a Data Mining tag line on a rule like the one above would be, "Buy Ratio between 1.30 to 1.80, but not any higher than 1.80." This would suggest that the numbers being spit out by Cameron, the ratio calculated by those numbers, and the relevance they have to the probability of any particular movement in the stock market are worthless.

Yet another example of a Data Mining tag line on a rule would be, "Cancel Call if any major companies are releasing Earnings that day." This one is not only an example of Data Mining, but would be very disillusioning with regards to the possibility that so many different events could trip up the program, and goes against the theory that due diligence on company earnings and world markets from major institutional trading firms is being represented in recent market candles.

Take a look at the chart to the right. (Click to enlarge) You'll see that this period of Nasdaq activity from the 20 calendar days of 11/18/09 to 12/08/09 was chaotic and unpredictable. The fact that Cameron spit out three trades within that period of chaotic market movement, and the fact that all three of the trades were accurate at guessing the market's move from open to close based on the logical rules established and extracted from the first 3 months of Cameron data, is absolutely mind boggling. Cameron can sense volatility.

Why is this happening? Well there are two reasons. One is simple, and the other has to do with Artificial Intelligence.

The first reason is the answer to the question someone asked me at a message forum... "How will Cameron handle a Bear Market?... or handle the TRANSITION from a Bull to a Bear Market?" The answer I gave was, "There is nothing that can happen in present and future market activity that hasn't already happened in the previous set of Nasdaq historical data that is run through 150 million times when Cameron is initiated each evening." In other words, the Street trader, Institutions and Short Sellers all behave the same way during every set of two different Bull Markets and two different Bear Markets, or during any other period of volatility. And similarly, when a volatile period of time rears its ugly head, Cameron can sense it based on previous market predictability.... or lack there of.

Now keep in mind, I never wrote any part of the program that was intended to specifically look for a Bull/Bear transition, or to recognize a volatile pattern like the one shown above. The decision to not spit out trades during a period of volatility or unpredictability is based SOLELY on past market activity in the previous set of candle data used in the program, and also based on the logical DI Rules that the trading system uses to determine the likelihood of the Nasdaq moving up or down the next day.

The other answer, on the subject of Artificial Intelligence, has to do with what I call "Micro Failures" that are happening almost every time Cameron is run and spits out the data. For example, the data for the morning of 11/30 was a Long Call with a 1.15 ratio, and it wasn't a confirmed Call. Literally THAT NIGHT, after the Nasdaq failed to go up a certain percentage on 11/30 (can't say what percentage), this micro failure is "Recognized" by Cameron as the program was coming up with the guess for 12/01... the next day. That's because the market didn't go up by the percentage that is programmed into the Cameron algorithm to search for in hypothetical trade situations as the program is running through the past Nasdaq candle data 150 million times. Unfortunately I can't say what percentage Cameron looks for in the cycles because it is something proprietary to how the program works. Regardless of that exact number, whatever patterns were used for the 11/30 guess will be ignored by the algorithm. They will also be ignored in the program runs over the following days and weeks until the data drops off the end. (The oldest part of the historical data being analyzed by Cameron on each run)

The same can be said for the Long Call that was spit out on 12/22/09, but still produced a .62% profit due to the market closing in the direction of the Call. Cameron also handles that as a failure because the Nasdaq didn't go up a certain percentage that day. So in all future runs from that point on, Cameron will disregard any patterns that would raise the final Buy Trade number in the left program instance based on these patterns that, on a technical level, actually "Failed."

In this manner, Cameron IMMEDIATELY learns from even the slightest mistakes and/or market changes, regardless of whether or not a failed trade was triggered. Remember, the program doesn't "Know" anything about what Call Ratios are required for or are being used to trade with. It only "Analyzes a certain amount of predictable percentage change based on previous candle patterns." (Can't detail any more) And the process of data analysis is completely SEPARATE from the actual Cameron computer program. In fact, the data analysis at this point in the timeline was done by hand... and with a pencil. :) The data analysis is NOW done by other computer programs that I wrote... programs that revolve around Cameron's pressure points and data, but aren't "Connected" to the Cameron computer program. The Cameron computer program hasn't changed at all since the Fall of 2009.

Also of note; In late December of 2009 Cameron nailed 6 Long Calls in a row and made 9.4% in 11 calendar days. The result in the stock market was 5 stacked and gapped white candlesticks. The fact that these 6 trades were triggered by artificial intelligence without having any forward looking information and only using logical data analysis, is completely mind boggling because the program is strictly analytical. For all intents and purposes, Cameron is NOT a "Stock Market Analysis Program." This 9.4% in 11 days is the flip side of how Cameron's daily predictive ability works, and is the reason why patience is so important for strictly following the technical analysis of ANY Artificial Intelligence-based trading system. Remember, if you gave $250,000 or a million dollars to a hedge fund and were able to peek in at the performance during certain 2 month periods, you would probably have heart failure. But over the longer term, 6 months to a year, the trading results of the more profitable hedge funds usually turn a tidy profit because of the massive gains that can be made during more easily predictable market conditions... like the period of time shown on the right, and because losses can be limited by using tight Stop orders for all of the trades.

So getting back to the subject of a market that may transition from Bull to Bear; Cameron is ALREADY handling that situation on a day to day basis, as WELL as having information in the historical candle data that has Bearish and Bullish patterns in it.

With all that said, why is Cameron so accurate? Was it my analytical choosing of different possible candle placement patterns? Was it due to the fact that I had very LITTLE stock market experience at the time I wrote the program and I simply thought like a COMPUTER would?... like the Ph.D. guys and gals at Renaissance Technologies? Does it have to do with using the NASDAQ levels?

I believe all THREE reasons mentioned above are why Cameron is so accurate. Most importantly though, it is because I'm using the NASDAQ levels for analysis. The Nasdaq represents an "Average" of over 2,500 companies. And even though most of them are technology-based companies, it still represents a more accurate trader sentiment compared to the Dow Jones Industrials average or the S&P 500 average because everything about the way our world works leads back to technology. Technology is EVERYWHERE. It's in a car, an office, a store, in the hands of youth (video games and iPods), and is even right in front of YOU. :) Whether it's Gold Mining, office cleaning, food cultivation or oil drilling, it ALL uses technological devices to complete the tasks. Never before has technology in so many forms been the most important driving force behind the US and World stock markets. And I now believe that this is the #1 reason why my computer program works so well. (Using Nasdaq data and an average of so many companies... over 2,500) As to whether or not there really are "clues" within the daily Nasdaq candle data from Institutional Traders and others, that will always be a debate that could go on forever. Regardless, Cameron works in more ways than one, and this is all that should matter to the average investor. Personally, I don't care WHERE the predictive ability is coming from.

So remember, Cameron is not necessarily a candle chart analysis program. And it is not necessarily a stock market-predicting program. It is an "Over-analytical, non-random event finder" that is being used through data ANALYSIS to predict what the entire stock market will do the next day... or in the COMING days as you'll read about shortly.

On the subject of the "Market must open a certain percentage up or down" aspect of the DI Rules for ANY Daily Call trading system; There is no trick to this. If you think it's possible that a simple equation and simple trading strategy can be extracted from "Did the market open up or Down" and was it "Within this range," go ahead and try it. I actually wrote a small computer program to try and see if that was possible... just to confirm that creating market Open-up & Open-down requirements with Cameron's Calls really was a logic-based strategy and not some kind of trick. And the answer was no. It is not possible to have a positive return during any one year period based only on the Nasdaq or S&P 500 opening up or down a certain percentage. It is only by using highly accurate Buying Pressure and Selling Pressure technical analysis COMBINED with an Open-Up or Open-Down requirement, that a positive rolling one year profit can be consistently made in any market condition and over a long period of time.

And how did the Cameron 'Daily Call' trading system PERFORM between October of 2009 and April of 2010? It made between 2% and 5% per month. But things were about to enter a whole new level of accuracy at this point in the timeline.

So Cameron's two data numbers from the analysis of the daily Nasdaq candles are actually a representation of Buying Pressure or Selling Pressure in the ENTIRE stock market. The ratio number created from these buying and selling pressure numbers was also used to accurately gauge trader sentiment over longer periods of time for our 2010 "Market Trend Call " subscription, which is explained and linked in "Part 2" of The Story of Cameron below.