As I mentioned in the overview I threw together a sample system based on the output from the five neural networks that comprise the Vantage Point Software.
First we'll take a look at the five neural networks and the output they produce. They are as follows:
1. Medium trend indicator. (PTM) This network predicts the 10 day moving average of closes 4 days in advance. In other words the 10 day moving average is expected to be at this price 4 days out. When Vantage Point says that their networks are 75 percent accurate or 80 percent accurate this is the network they are basing it on. They claim that around 80 percent of the time when the PTM (Predicted Trend Medium) says the market will be up or down or flat -- that this indicator is the one which is accurate that percentage of the time.
2. Short term trend indicator. (PTS) This neural network is similar to the PTM but predicts the 5 day MA of closes 2 days in advance.
3. Predicted High. This network predicts the price of the high within the next 24 hours.
4. Predicted Low. This is the lowest price the market is expected to hit within the next 24 hours.
5. Index. This indicator predicts whether market direction is expected to be up or down. It usually shows a value of 1.00 for up and "0.00" for down. It may also show numbers in between 1.00 and 0.00 but this is relatively rare. It usually delivers a very clear bias as to its predicted market direction over the next 24 hours.
The following graphic image is a screen shot of the actual daily output chart from the Vantage Point system.
http://www.forexfactory.com/fijireports/E_VP_output.jpg
The basis for the day trading system
Having been fascinated by the idea of a predictive system and having never quite given up on discovering something that could predict market behavior with notable accuracy I was tempted to see if a system that relied on VP's daily price predictions would be viable. After looking at the indicators and out-putting 600 trading days from VP's history of predictions and comparing it with EOD daily bar history in Excel -- I experimented with various kinds of day trades just to get a feel for the VP software and what I might be able to expect from it when placing day trades.
Having no easy way to back test a system designed in Vantage Point except a rather crude attempt using an Excel spreadsheet or with a rather painstaking manual process -- I opted instead to try something simple that would be likely to work if the predictions of this software were reliable and which would most likely not work if the predictions were too sporadic to be meaningful enough to win with a day trade system.
Using the output from Vantage Point and taking into account that I would not have time to position trade sufficiently to produce a meaningful review -- I decided that since the platform predicts the high and low for the next 24 hours -- if simply entered at the low and exited at the high on a long trade and did the reverse with a short trade -- the system should work given proper risk management and if the predictions were somewhat reliable. As I mentioned in the overview thread -- one of the criteria that I have for all systems that I review -- is how much daily time and focus is required to trade the system. Keeping that criteria in mind along with my week of experimenting I constructed a crude but simple set of rules. They are as follows:
1. Scan the output of Vantage Point for the predicted high and low for each day right after the market close when the new EOD data arrives. Plot each as a horizontal line. The next graphic shows the predictions that were plotted before the most recent day and the market activity that occurred that day.
http://www.forexfactory.com/fijirepo...ne30%20JPY.jpg
2. Scan the trend (Index indicator) direction for the next 24 hours and use it as a basis for anticipated market direction. If directional strength appears strong -- trade only in the direction of greatest implied strength. (See the INDEX column for a "1" or "0")
3. Place entry orders for any or all of the six markets -- a little inside the high or the low of the day based primarily on the directional bias of the Index indicator. If there did not seem to be a directional bias that was strong enough to be significant and if current market price was somewhere well inside of the predicted daily range then I place entry orders at both the high and low on an OCO basis.
4. Check the market periodically to see how the trades are doing. I permit myself the freedom to close any or all trades if things did not look right or if the aggregate of trades was up several percent.
5. Use stops about 1/3 the size of targets -- basing them on support/resistance or just max risk. Typically this ranges 30 to 45 pips.
The following graphic shows a tyipcal entry order that was in place. The dotted black line is the entry and the two red lines are exits for profit or loss. This one was obviously hit and is close to exiting for profit.
http://www.forexfactory.com/fijirepo...aug3%20CHF.jpg
6. Risk less than 1 percent per trade. Typical risk averages around .6 percent per trade. Most likely I could have risked more and done better but I experimented with a position building rule that stated if a position had been entered inside the predicted range that I'd add to the position up to 20 pips above the first entry with tight stops. As I blundered around I decided that this rule required too much analysis to properly build a position for my experimental purposes here. I was not really focusing much on money management due to my interest in testing the viability of the predictive system idea. Some would disagree with that approach and I may disagree with it myself under other conditions. However I'm happy with the amount of effort required to produce acceptable results pretty much right out of the gate.
Below is a log of trades from the month of july. I've continued trading the system since this log and the results have been pretty consistent. I generally risked an average of .6 percent per trade but some trades I risked as little as half that much. Still the profit for a little over a month of trading came in around 14 percent which I think is excellent considering the negligable drawdowns and ultra conservative risk level.
http://www.forexfactory.com/fijirep...ly_trades_3.htm
Also check out the graphics thread in this section for a number of shots of predicted highs and lows that occurred throughout the month.
FT
<!-- / message --><!-- sig -->
First we'll take a look at the five neural networks and the output they produce. They are as follows:
1. Medium trend indicator. (PTM) This network predicts the 10 day moving average of closes 4 days in advance. In other words the 10 day moving average is expected to be at this price 4 days out. When Vantage Point says that their networks are 75 percent accurate or 80 percent accurate this is the network they are basing it on. They claim that around 80 percent of the time when the PTM (Predicted Trend Medium) says the market will be up or down or flat -- that this indicator is the one which is accurate that percentage of the time.
2. Short term trend indicator. (PTS) This neural network is similar to the PTM but predicts the 5 day MA of closes 2 days in advance.
3. Predicted High. This network predicts the price of the high within the next 24 hours.
4. Predicted Low. This is the lowest price the market is expected to hit within the next 24 hours.
5. Index. This indicator predicts whether market direction is expected to be up or down. It usually shows a value of 1.00 for up and "0.00" for down. It may also show numbers in between 1.00 and 0.00 but this is relatively rare. It usually delivers a very clear bias as to its predicted market direction over the next 24 hours.
The following graphic image is a screen shot of the actual daily output chart from the Vantage Point system.
http://www.forexfactory.com/fijireports/E_VP_output.jpg
The basis for the day trading system
Having been fascinated by the idea of a predictive system and having never quite given up on discovering something that could predict market behavior with notable accuracy I was tempted to see if a system that relied on VP's daily price predictions would be viable. After looking at the indicators and out-putting 600 trading days from VP's history of predictions and comparing it with EOD daily bar history in Excel -- I experimented with various kinds of day trades just to get a feel for the VP software and what I might be able to expect from it when placing day trades.
Having no easy way to back test a system designed in Vantage Point except a rather crude attempt using an Excel spreadsheet or with a rather painstaking manual process -- I opted instead to try something simple that would be likely to work if the predictions of this software were reliable and which would most likely not work if the predictions were too sporadic to be meaningful enough to win with a day trade system.
Using the output from Vantage Point and taking into account that I would not have time to position trade sufficiently to produce a meaningful review -- I decided that since the platform predicts the high and low for the next 24 hours -- if simply entered at the low and exited at the high on a long trade and did the reverse with a short trade -- the system should work given proper risk management and if the predictions were somewhat reliable. As I mentioned in the overview thread -- one of the criteria that I have for all systems that I review -- is how much daily time and focus is required to trade the system. Keeping that criteria in mind along with my week of experimenting I constructed a crude but simple set of rules. They are as follows:
1. Scan the output of Vantage Point for the predicted high and low for each day right after the market close when the new EOD data arrives. Plot each as a horizontal line. The next graphic shows the predictions that were plotted before the most recent day and the market activity that occurred that day.
http://www.forexfactory.com/fijirepo...ne30%20JPY.jpg
2. Scan the trend (Index indicator) direction for the next 24 hours and use it as a basis for anticipated market direction. If directional strength appears strong -- trade only in the direction of greatest implied strength. (See the INDEX column for a "1" or "0")
3. Place entry orders for any or all of the six markets -- a little inside the high or the low of the day based primarily on the directional bias of the Index indicator. If there did not seem to be a directional bias that was strong enough to be significant and if current market price was somewhere well inside of the predicted daily range then I place entry orders at both the high and low on an OCO basis.
4. Check the market periodically to see how the trades are doing. I permit myself the freedom to close any or all trades if things did not look right or if the aggregate of trades was up several percent.
5. Use stops about 1/3 the size of targets -- basing them on support/resistance or just max risk. Typically this ranges 30 to 45 pips.
The following graphic shows a tyipcal entry order that was in place. The dotted black line is the entry and the two red lines are exits for profit or loss. This one was obviously hit and is close to exiting for profit.
http://www.forexfactory.com/fijirepo...aug3%20CHF.jpg
6. Risk less than 1 percent per trade. Typical risk averages around .6 percent per trade. Most likely I could have risked more and done better but I experimented with a position building rule that stated if a position had been entered inside the predicted range that I'd add to the position up to 20 pips above the first entry with tight stops. As I blundered around I decided that this rule required too much analysis to properly build a position for my experimental purposes here. I was not really focusing much on money management due to my interest in testing the viability of the predictive system idea. Some would disagree with that approach and I may disagree with it myself under other conditions. However I'm happy with the amount of effort required to produce acceptable results pretty much right out of the gate.
Below is a log of trades from the month of july. I've continued trading the system since this log and the results have been pretty consistent. I generally risked an average of .6 percent per trade but some trades I risked as little as half that much. Still the profit for a little over a month of trading came in around 14 percent which I think is excellent considering the negligable drawdowns and ultra conservative risk level.
http://www.forexfactory.com/fijirep...ly_trades_3.htm
Also check out the graphics thread in this section for a number of shots of predicted highs and lows that occurred throughout the month.
FT
<!-- / message --><!-- sig -->