Page 1 of 6

Middle analysis - Taking home the Pips

Posted: Wed May 14, 2008 7:01 pm
by pitboss
TRO this one is for you :)

A bias exists between a period's opening and the previous period's middle significant enough to make the forex beatable. As an inllustration:

A $10,000 starting bankroll becomes $1,500,000 in 8 years!

My analysis program accessible via the internet at:
http://forex.azrewards.com

You will find that each of the major 4 pairs show this bias in the time periods of 15m, 30m, 1hr and daily. Additionally, their is a analysis of the 4 pairs combined together for daily.

Nope, I'm not selling anything - just my little contribution. Take a look and check out what i'm saying isn't so.

For the folks that like the nitty details:

Data was obtained from Metatrader history files,
The 'spread' is a percentage from the previous middle,
'Commission' is either ur broker's pip spread or mbt's commission,
A running kelly criterian is used for determining bankroll risk and sizing, &
The minimum BR risk is 1% and max at your pain threshold.

Illustrations

Posted: Thu May 15, 2008 3:32 am
by pitboss
Here is the daily EUR from 14 Jun 00 thru 12 May 08.

Observe how a change in the users input effects the various summary components (i.e. win%, spread, advantage and etc).

No Spread - No Commission - 3% Advantage


.0015 Spread - No Commission - 6.9% Advantage


.0015 Spread - 1/2 pip Commission (both ways) - 5.3% Advantage

Here is the 1 1/2 million bucks!

Posted: Thu May 15, 2008 3:48 am
by pitboss
Duh!! Guess I should have posted this with the first posting.


Posted: Thu May 15, 2008 7:08 pm
by TheRumpledOne
Could you please explain what this means?

What is the entry/exit criteria?

Should we FADE THE MIDDLE or trade the other way?

More complete explaination:

Posted: Fri May 16, 2008 2:40 am
by pitboss
The analysis is calculated using absolutes, there are no suppositions. The previous period's high/low and the current period's opening/closing prices are a matter of fact. You will find that the summary information for each of the individual pairs are very close and simular.

This is what the following chart tells you:

Entry:
Open above previous period middle - go short
Open below previous period middle - go long
No other entry criteria

Exit:
Close of the period being used (in this case, daily)
No other exit criteria

Summary:
2,068 possible trade days
Approx 1/2 qualify as long the other 1/2 short
Whether long or short, over a 50% win
There is a positive expectation return of 3% on every dollar



This is what the following chart tells you:

Entry:
Open above previous period middle * 1.0015 - go short
Open below previous period middle * .9985 - go long
No other entry criteria

Exit:
Close of the period being used (in this case, daily)
No other exit criteria

Summary:
2,068 possible trade days with 1344 qualifying (65%)
Approx 1/2 qualify as long the other 1/2 short
Whether long or short, well over a 52% win
There is a positive expectation return of 6.9% on every dollar



If you run a 'combined' analysis report, you will receive a summation of each of the pairs having being acted upon during the same exact time period (i.e. day 1 looks at each individual pair's results for that day to make a summary for that day, day 2 does the same, as does day 3, etc).

The meat of this is in the combined analysis, rather than looking at one single pair and comparing that pair's result to another pair's result. It would be similar to playing 4 hands of blackjack at a time using the exact same play strategy per hand. 1 of the hands may do extraordinarliy good, 1 extremely pourly with the other 2 being middle of the road but the mathematical probability for each hand is actually identical. It is only the combination of the individual results that provide the true results whereas selecting one of the individual hand's results would be misleading.

This is what the following chart tells you:

Entry:
Open above previous period middle * 1.0015 - go short
Open below previous period middle * .9985 - go long
No other entry criteria

Exit:
Close of the period being used (in this case, daily)
No other exit criteria

Costs:
Pay MBT commission of .00005 for each side of trade

Summary:
2,064 possible trade days with 2 1/2 pairs qualifying each day
Approx 1/2 qualify as long the other 1/2 short
Whether long or short, well over a 53% win
There is a positive expectation return of 8.7% on every dollar


Posted: Fri May 16, 2008 5:55 pm
by TheRumpledOne
Thank you for sharing.

Testing Results in Demo A/C

Posted: Sun May 18, 2008 3:04 am
by pitboss
Have traded this in a demo account for 15 days.

I make entries around 19:30 EST (after margin rollover costs) and exit the next day around 15:30 EST (just prior to NY closing).

Made 44 trades, averaging 2.93 daily

Results include commission both ways

22 long - 12 win, 10 lose = 54.5% win
22 short - 11 win, 11 lose = 50.0% win

Total win $3,716.07, average $168.91
Total lose $2,608.28, average $118.33

Net advantage 21.4%

Clearly I am doing extraordinary and cannot expect this margin of advantage to be maintained. Of course, if it did I'd be packing for the a south sea island and a life of leisure, lol.

Posted: Mon May 19, 2008 5:31 pm
by Skyline
Hi Pitboss,
I like your system and since I'm a metatrader coder it could be a good idea to translate your method into an EA to backtest/forward test.
Could you summarize all details of your method pls ?

Thank You :)

Skyline

Posted: Mon May 19, 2008 11:24 pm
by pitboss
Skyline:

You are welcome to write an EA but duplicating work using exact data source info. The elements of my method are in an earlier post. Ony thing I overlooked posting was that I set a catastrophic stop loss of 3 stdev for individual pair's daily average.

I just chose to import Metatrader historical data, transfer into a database and just wrote a stand alone program for the analysis. Found it was easier to modify and play around with various senerios. I even made my work publicly available via the internet:

http://forex.azrewards.com

I used TRO's mid and modified it to indicate the entry points for short and long to make it eaiser to visually see qualifying trades on Metatrader charts. Make actual trades on MBTrading.

My philosophy is KISS - "Keep it simple stupid", lol

As of today:

Total 48 trades - 16 days
24 long - 13 win 54.2% - 11 lose
24 short - 13 win 54.2% - 11 lose

Total 26 win 54.2% - 22 lose
Total win $4,195 - lose $2,603

Average win $161.37 - lose $118.03
Average daily net $99.22

Net expectation 27.9%

Posted: Tue May 20, 2008 8:27 am
by Skyline
Yes I see PitBoss , btw just for the fun to do it I already coded the EA :)
Be also aware that history downloaded metatrader data are far to be accurate (even if you are using daily data) infact I did my test using Alpari data with an accuracy of 90% in the modelling quality backtest.
Btw I tested only EURUSD from 1 Jan 2008 up to 1 May 2008 and in attachment you can find my result that seems worst than yours.

cheers,
Skyline