Menu
Sister Sites

Options

Bookmark Thread

First Page First Unread Last Page Last Post

Print Thread

Similar Threads

How to Correctly BackTest, Collect Data, and Analyze A System 13 replies

How much historical data needed to backtest day trading system 0 replies

Visual backtest result different from non-visual backtest 0 replies

data needed to backtest system 1 reply

How to backtest system with past data? 8 replies

Attachments: NFP System - Backtest data provided
Exit Attachments

NFP System - Backtest data provided

  • Post #1
  • Quote
  • First Post: Edited Apr 12, 2007 11:37am Mar 26, 2007 4:13pm | Edited Apr 12, 2007 11:37am
Whats up fellas? I have something that I would like to share with you all. Rob Booker www.robbooker.com recently held a contest to see who could come up with the best system to trade the NFP report. Here is the system (backtest data in Excel format), as I've agreed to make it freely available shortly after I won the contest. A couple of notes about it:

This system relies heavily on an indicator called the Sentinal Index. Origionally created by Joe Knight, I adopted the concept and gave it a use in my system. What it is is a "dollar index" created by taking the movement of the 7 major USD based pairs (GBP/USD, EUR/USD, AUD/USD, NZD/USD, USD/JPY, USD/CAD, USD/CHF) and averaging their daily moves together to come up with a meld of them all, and an overall direction of the dollar. I plot this in Excel, and lay my 7 day SMA indicator overtop, which I use for entry into the system. I don't have an automated means of charting and distributing the Sentinal Index yet (which I intend to distribute freely), but I do still update it by hand, and do offer services such as custom system development based on the index, my signals based on the Sentinal Index and a short description of each trade daily, and so forth. If you are interested in any of these, PM or email me, and we can make a system like this one together

Success rate: aprox. 80% over 4 years
Equity: $5,000 to $460,526.50 USD in 4 years
Pips: 10,467 in 4 years (or an average of 215 a month)


Kev
Attached File(s)
File Type: txt NFP.txt   3 KB | 4,844 downloads | Uploaded Apr 12, 2007 11:34am
File Type: txt Senital Index.txt   2 KB | 5,103 downloads | Uploaded Apr 12, 2007 11:35am
File Type: mq4 Sentinal Index_SMA.mq4   3 KB | 3,179 downloads | Uploaded Apr 12, 2007 11:35am
File Type: mq4 Sentinal Index_SMA_modA.mq4   3 KB | 2,697 downloads | Uploaded Apr 12, 2007 11:35am
File Type: mq4 Sentinal Index.mq4   3 KB | 2,865 downloads | Uploaded Apr 12, 2007 11:37am
This is really cool. Thanks for sharing the info. I'll try it in the near future and will get back to you ASAP. Thanks again.

Al
seems neat .. waiting for the indicator dude
Hi,

This looks great, I know rob well and I can see he has showed you how to backtest...

I will backtest this myself tonight

Thanks for sharing

-S-
  • Post #5
  • Quote
  • Edited Apr 21, 2007 7:52pm Mar 26, 2007 4:44pm | Edited Apr 21, 2007 7:52pm
Well, the Sentinal index isn't widely distributed as of yet. Joe Knight is the origional creator, and he let me borrow his idea, computerize and backtest it according to my NFP system. I am in the process of making a webpage that has real time java based charts that will provide the indicator (free of charge). I am on holiday in Canada, and will be returning this Wednesday, so expect something by this coming weekend. I don't have a completely automatic system for producing the indicator, though I am working on a program to do so, so I can't exactly re-do my work here It would take entirely too long. So I hope you can bare with me in the next few days.

-Kev


Edit: Below are the Sentinel MT4 files for IBFX Mini
Attached File(s)
File Type: zip NFP data.zip   18 KB | 2,166 downloads | Uploaded Apr 12, 2007 11:36am
File Type: mq4 Sentinal%20Index[1]IBFX Mini.mq4   3 KB | 2,938 downloads | Uploaded Apr 21, 2007 7:51pm
File Type: mq4 Sentinal%20Index_SMA[1]IBFX Mini.mq4   3 KB | 1,353 downloads | Uploaded Apr 21, 2007 7:51pm
File Type: mq4 Sentinal%20Index_SMA_modA[1]IBFX Mini.mq4   3 KB | 1,465 downloads | Uploaded Apr 21, 2007 7:51pm
take your time kevin
  • Post #7
  • Quote
  • Edited 5:07pm Mar 26, 2007 4:56pm | Edited 5:07pm
Thank you Kevin for bringing this to us. Definatly worth the test. Enjoy the holiday
Will the sentinal rules be published so we can incorporate them into our API? It sounds like it will end up being for sale. Freely available from the website, but how can we automate the trades our selves without the rules for the sentinal? We can't back-test your results...and we can't verify the signals.

Are you planning on making the rules freely available? or will the indicator be for sale?

Quoting Kevin_E
Well, the Sentinal index isn't widely distributed as of yet. Joe Knight is the origional creator, and he let me borrow his idea, computerize and backtest it according to my NFP system. I am in the process of making a webpage that has real time java based charts that will provide the indicator (free of charge). I am also starting a signal service based on this indicator where I will offer updated sentinal numbers, how to trade based on the indicator, signals based on the indicator, one on one help in trading with it, etc. Also, all the data that I have on the indicator is back home. I am on holiday in Canada, and will be returning this Wednesday, so expect something by this coming weekend. I don't have a completely automatic system for producing the indicator, though I am working on a program to do so, so I can't exactly re-do my work here It would take entirely too long. So I hope you can bare with me in the next few days.

-Kev
mrmikal,

I'm not withholding the Sentinal index from anybody. I'm going to make it freely available, but none of my work is here. If you want the rules to create the Sentinal index, something I worked hours a day for weeks to come up with in a useable format, then here you go Any time frame (30 minutes used in the system), take the close of the current candle and subtract the close from the previous candle from it. That gives you a number that shows which direction (up if the number is positive, down if the number is negative) the pair has moved in the last two 30 minute candles. Now, do that for all 7 USD based pairs for the same candle (if you subtract the close on candle 8:30 from the close on candle 8:00 for Jan 8th, 2006 GBP, then you must subtract the close on candle 8:30 from the close on candle 8:00 for Jan 8th, 2006 for the other 6 pairs). Add all the numbers together, divide by 7 (to come up with an average USD move) and that gives you a number....logistically the total average amount all the USD based pairs have moved.

Now chart it. Take a base number, doesn't matter what it is...I use 1000. For each 30 minute candle you do the above rules for (and come up with your sentinal number for every 30 minutes), add this to your base number. If your first calculation gives you +8.24 (the number you get after you add all the movements together and divide by 7), add it to your base number 1000 to get 1008.24. Plot that on your graph. The second calculation (next 30 minute candle) gives you a movement of -3.12...then add that to your previous sentinal number which becomes 1005.12. You see? Thats a sentinal number. If you'd like to wait a few days, I can definately get you my hand generated numbers. Until then, if somebody comes up with an automated way of producing this on a chart, then I'd love to see it. It would save me much work

-Kev
Savage Pip Fiend
  • Post #10
  • Quote
  • Edited 2:10pm Mar 27, 2007 1:03pm | Edited 2:10pm
Hi Kevin_E,

Thanks for posting this intriguing idea...

I have been playing with the math for this sentinal index... I have a few questions;

1. In your rules you say you overlay the sma on top of it... How are you normalizing the index to match the scale of the SMA? Or are you referring a 7day sma of the sentinal index and not a 7 day sma of the close? If that is the case, how are you determining a 7 day SMA of a 30min index? Are you using 336 periods of the 30 minute timeframe index? Or are you choosing some particular time of day for the index to denote a end of day.?

2. I see from your backtest, that you have performed the test on GBPUSD.
Have you tried anyother pairs or is this strategy for GBPUSD?

It is an interesting idea... I would like to look into it further... thanks in advance, Scott
Kevin,

Thanks for the info...question, though...

The USD/JPY cross...obviously that number will be considerably higher than the others...won't that screw with your average? For example:

The USD/JPY moves 7 pips but the the GBP moves 60 pips...if you don't normalize and just take the subtracted value, you will get (0.07 + 0.006) / 2 = 0.038, which obviously will be skewed...if you were to do the pip values, you would get (60 + 7) / 2 = 33.5 pips...so what is it?

Do you have to normalize the pip value or something?

Also, you say chart the 7 USD base pairs.

What pairs are those? If you're charting the majors, that should be 4 pairs. After that, we have the NZD, AUD and CAD, I would guess...but what about the HKD, the SKK, the Rand, etc...I'm taking it as you're not including those...just the 7 crosses I mentioned. Right?

Quoting Kevin_E
mrmikal,

I'm not withholding the Sentinal index from anybody. I'm going to make it freely available, but none of my work is here. If you want the rules to create the Sentinal index, something I worked hours a day for weeks to come up with in a useable format, then here you go Any time frame (30 minutes used in the system), take the close of the current candle and subtract the close from the previous candle from it. That gives you a number that shows which direction (up if the number is positive, down if the number is negative) the pair has moved in the last two 30 minute candles. Now, do that for all 7 USD based pairs for the same candle (if you subtract the close on candle 8:30 from the close on candle 8:00 for Jan 8th, 2006 GBP, then you must subtract the close on candle 8:30 from the close on candle 8:00 for Jan 8th, 2006 for the other 6 pairs). Add all the numbers together, divide by 7 (to come up with an average USD move) and that gives you a number....logistically the total average amount all the USD based pairs have moved.

Now chart it. Take a base number, doesn't matter what it is...I use 1000. For each 30 minute candle you do the above rules for (and come up with your sentinal number for every 30 minutes), add this to your base number. If your first calculation gives you +8.24 (the number you get after you add all the movements together and divide by 7), add it to your base number 1000 to get 1008.24. Plot that on your graph. The second calculation (next 30 minute candle) gives you a movement of -3.12...then add that to your previous sentinal number which becomes 1005.12. You see? Thats a sentinal number. If you'd like to wait a few days, I can definately get you my hand generated numbers. Until then, if somebody comes up with an automated way of producing this on a chart, then I'd love to see it. It would save me much work

-Kev
Savage Pip Fiend
I think he mentioned the 7 pairs in his first post

This system relies heavily on an indicator called the Sentinal Index. Origionally created by Joe Knight, I adopted the concept and gave it a use in my system. What it is is a "dollar index" created by taking the movement of the 7 major USD based pairs (GBP/USD, EUR/USD, AUD/USD, NZD/USD, USD/JPY, USD/CAD, USD/CHF)

I just ignore the decimal when I built my index. That way it normalizes itself... It would be good to here Kevin's take on this...
Wow, I must have totally been blind...the details were in the first post...I just couldn't get the gist of what was going on. OK...Now I totally get it. I assume it wouldn't be too difficult to create an indicator on this for MT. I've been looking at doing something similar but for another purpose.

On a side note, I would think a "real" dollar index would have to possibly weight each of these currencies based on average daily volume. Obviously trades in less liquid pairs (like the GBP/USD) can move the index disproportionately just because of the higher volatility (due to lack of liquidity). Again, I'm not saying that the Sentinal index should be modified, but I was thinking along the same lines.

Anyways...interesting stuff.


Quoting smjones
I think he mentioned the 7 pairs in his first post

This system relies heavily on an indicator called the Sentinal Index. Origionally created by Joe Knight, I adopted the concept and gave it a use in my system. What it is is a "dollar index" created by taking the movement of the 7 major USD based pairs (GBP/USD, EUR/USD, AUD/USD, NZD/USD, USD/JPY, USD/CAD, USD/CHF)

I just ignore the decimal when I built my index. That way it normalizes itself... It would be good to here Kevin's take on this...
Quoting mrmikal

On a side note, I would think a "real" dollar index would have to possibly weight each of these currencies based on average daily volume. Obviously trades in less liquid pairs (like the GBP/USD) can move the index disproportionately just because of the higher volatility (due to lack of liquidity). Again, I'm not saying that the Sentinal index should be modified, but I was thinking along the same lines.

Anyways...interesting stuff.
This is a good point.. I had not even thought of it... something else to play with... Thanks...
Yeah usually I just ignore the decimal to normalize. My first attempts at a sentinal index were to create a base day (Jan 3rd of 2000) and take the first close of that day as a base number (insead of 1000). Then, for each day, subtract the sentinal numbers from the base day like normal, totally taking the decimal out of the equation, giving me a normalized rounded number of the total pips moved. The number doesn't really matter, just the movements.

I have also taken into account the strength of each pair in relation to the USD, and its effect on the Sentinal index. Ideally, if you make the sentinal index without factoring this in, it would react differently to different pairs (seeing as how some pairs are more effected by the USD than others), instead of attempting to normalize to make it behave the same on each pair. I sort of counted on this, as I would test all the pairs versus the Sentinal index, and see which one it would work best on during different days and times (and months/years). I kind of stopped at GBP/USD because I entered into a contest, and the time constraints forced me to stick to one pair as opposed to manually backtesting on all the other ones. I'm working on putting this into an expert advisor in MQL4, which will let me backtest this against other pairs much more quickly, and let me share it more openly.

A "corrected" index would be me creating proportions on how far a pair has gained or slid against the USD (its weight on the USD currently versus the other pairs) and factoring that into the Sentinal number, giving me a different indicator for each pair. I'll probably work on that later on, but I think theres alot more for me to milk out of this as of now, so I'll put that idea on the back burner

Also, I've already thrown around a few ideas and charts about only charting the number in specific time frames. If you're only able to trade between the hours of 5pm and midnight, for example, make your number only deal with the times between 5pm and midnight. I've done one between 9am and midnight (completely ignoring the times of midnight to 9am), and the results are pretty nice, and still reflect a good indicator. When I get home in a couple of days, I'll definately share more on it. Until then.

-Kev
Savage Pip Fiend
Kevin, when you get a minute, could you answer my first question about the 7 days sma in this post ?
http://www.forexfactory.com/showpost...1&postcount=10

thanks
I simply factor.
30 minutes = 336 SMA

You could use one time of day, but you would only come up with one point per day, and a straight line between them...also meaning you could only take a trade at that one point during the day, because only then would you know if your sentinal number is above or below the SMA.

sorry about missing the question.

-Kev
Savage Pip Fiend
Quoting Kevin_E
I simply factor.
30 minutes = 336 SMA

You could use one time of day, but you would only come up with one point per day, and a straight line between them...also meaning you could only take a trade at that one point during the day, because only then would you know if your sentinal number is above or below the SMA.

sorry about missing the question.

-Kev
Savage Pip Fiend
Thanks, the 336 is what I figured it would be.
Another question. Where and how are you placing stops, so you can determine your risk % I see you haveused 5%, but what is that 5% based on? thanks
Well, I leverage 5% of my account.

Meaning if I have an account worth 50,000, I'll be leveraging 2,500 per position (or 2.5 lots for both positions that I have open, 5 lots total..which is 10% total leverage). Then I just put the stops 100 pips in the opposite direction, which mean a drawdown of aprox. 10% of my account if I lose, but has in the past gained back decently

-Kev
Savage Pip Fiend
0 traders viewing now