Traders' Roundtable

Searched through the forum but couldnt find anything on this, so feel free to point me in the right direction.

I have developed 2 spot forex system based on 5 minute charts both - over the last 17280 5 minute bars, the following results have been generated

System 1 : 185 trades - 43% win rate, a risk adjusted pay off ratio of about 2:1 and it has a positive geometric mean

System 2 : 120 trades - 55% win rate, a risk adjusted pay off ratio about 3:7 to 1 and a positive geometric mean.

All entry signals are generated based on price action and time of day/ no other were indicators used. All testing has been forward testing on live action and I have seen the payoff ratios vary during the course of the trades. There has been no fitting or anything of the sort. All trades have been entered manually

My question is, how many data points/trials/trades would you want to see before you "think" you have a valid "system" or approach to the market. Had this been a system that used daily bars, the back testing period would have been (17000 days/260 trading days a year) about 65 years.

Thanks

Hi,
Doing a quick calculation suggests you have 60 days of 5 minute data. I recently re-signing up to eSignal which also only provides three months or so back of intra day data, but I then bought five and a half years of 5 minute data back to January 2004 from Forex Tick Data(http://www.forextickdata.com/index.html). I then just cut and pasted the Forex Tick Data (you specify the format you want it in), on to the eSignal data and I now have seamless data that I can update in real time but goes back to 2004.
You are right to say that you have enough tests for statistical significance, but those tests still only cover what has happened in the markets over the past three months. Your two systems have generated approximately 300 trades over three months which is an average of five trades a day which means you probably intend to sit in front of your computer all day which is going to make the psychological aspects of trading much more important for you. It will be much easier for you to have belief in the validity of your system and stick to it during any bad patches if you can tell yourself not to worry because you have tested it and it has worked over the variable markets of the past five odd years, instead of only knowing that it has worked for the past three months.
Just my thoughts but I hope they help.
Matthew

Much thanks for the input. will take a look.

You are right to say that these tests only cover what happened over these markets of that past 3 months, but my thoughts are that given the small scale of my trading time frame (a 5 minute chart) relative to the last three months means that I have sat thru a lot of screen time and have sat thru times when the average daily range (based on 5 days has gone from 300 pips or so right down to under 100)

Each system generated two to 3 trades a day, but often there would be a 4 days of no trades and sometimes 2 weeks of straight losses

But yes, the more data the better

Thanks

Matthew28 wrote:Hi,
Doing a quick calculation suggests you have 60 days of 5 minute data. I recently re-signing up to eSignal which also only provides three months or so back of intra day data, but I then bought five and a half years of 5 minute data back to January 2004 from Forex Tick Data(http://www.forextickdata.com/index.html). I then just cut and pasted the Forex Tick Data (you specify the format you want it in), on to the eSignal data and I now have seamless data that I can update in real time but goes back to 2004.
You are right to say that you have enough tests for statistical significance, but those tests still only cover what has happened in the markets over the past three months. Your two systems have generated approximately 300 trades over three months which is an average of five trades a day which means you probably intend to sit in front of your computer all day which is going to make the psychological aspects of trading much more important for you. It will be much easier for you to have belief in the validity of your system and stick to it during any bad patches if you can tell yourself not to worry because you have tested it and it has worked over the variable markets of the past five odd years, instead of only knowing that it has worked for the past three months.
Just my thoughts but I hope they help.
Matthew

Additionally, the correlation coefficent between the results of the two systems is about 0.04, which is pretty low.

Matthew28 wrote:Hi,
Doing a quick calculation suggests you have 60 days of 5 minute data. I recently re-signing up to eSignal which also only provides three months or so back of intra day data, but I then bought five and a half years of 5 minute data back to January 2004 from Forex Tick Data(http://www.forextickdata.com/index.html). I then just cut and pasted the Forex Tick Data (you specify the format you want it in), on to the eSignal data and I now have seamless data that I can update in real time but goes back to 2004.
You are right to say that you have enough tests for statistical significance, but those tests still only cover what has happened in the markets over the past three months. Your two systems have generated approximately 300 trades over three months which is an average of five trades a day which means you probably intend to sit in front of your computer all day which is going to make the psychological aspects of trading much more important for you. It will be much easier for you to have belief in the validity of your system and stick to it during any bad patches if you can tell yourself not to worry because you have tested it and it has worked over the variable markets of the past five odd years, instead of only knowing that it has worked for the past three months.
Just my thoughts but I hope they help.
Matthew