GBP Patterns, Facts, Trivia | Page 4 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

GBP/AUD, GBP/CAD, GBP/CHF and GBP/NZD thread 126 replies

The facts 4 replies

Market & Trading Facts 60 replies

CHF, EUR and GBP Calendar Trading Patterns 0 replies

Attachments: GBP Patterns, Facts, Trivia
Exit Attachments

GBP Patterns, Facts, Trivia

  • #61
  • Jan 11, 2007 8:13pm Jan 11, 2007 8:13pm
I'm on a demo trade right now- trying what I learned out of curiosity just yesterday.

Get a daily chart and find every "THURSDAY" candle going back to June of 2006.

Look at the "FRIDAY" candle next to it.

How many times do you see green next to green and red next to red?

My calculations are that fully 70% of Thursday's trends continue through the trading day Friday.

In fact, if you had simply placed a trade at the close of the daily Thursday candle in confirmation of its trend (going long on a bull candle, short on a bear) and then held your position until the end of the Friday candle and then closing it out you would have averaged 50 pips a week on a simple, mechanical entry and exit- five minutes work- tops.

Yep, tis so.

But this strategy would have been a train wreck in the trading weeks prior to June of 2006. The market changed and now it seems that Thursdays are good confirmation for the direction Fridays will go- at least on Cable.

Is your SL my TP?
  • #62
  • Jan 11, 2007 8:27pm Jan 11, 2007 8:27pm
Great !!! Another little trick to add to the tool box !!
  • #63
  • Jan 11, 2007 8:32pm Jan 11, 2007 8:32pm
I likem' simple, don't you?


Quoting Tiki Trader
Great !!! Another little trick to add to the tool box !!
Is your SL my TP?
  • #64
  • Jan 11, 2007 8:32pm Jan 11, 2007 8:32pm
This means there is a good chance I should be going LONG tonight... which is fine with me since other than that tonight... I have no clue !!!
  • #65
  • Jan 11, 2007 8:33pm Jan 11, 2007 8:33pm
Hey Nice
Quoting kenmania
I'm on a demo trade right now- trying what I learned out of curiosity just yesterday.

Get a daily chart and find every "THURSDAY" candle going back to June of 2006.

Look at the "FRIDAY" candle next to it.

How many times do you see green next to green and red next to red?

My calculations are that fully 70% of Thursday's trends continue through the trading day Friday.

In fact, if you had simply placed a trade at the close of the daily Thursday candle in confirmation of its trend (going long on a bull candle, short on a bear) and then held your position until the end of the Friday candle and then closing it out you would have averaged 50 pips a week on a simple, mechanical entry and exit- five minutes work- tops.

Yep, tis so.

But this strategy would have been a train wreck in the trading weeks prior to June of 2006. The market changed and now it seems that Thursdays are good confirmation for the direction Fridays will go- at least on Cable.

  • #66
  • Jan 11, 2007 8:55pm Jan 11, 2007 8:55pm
Here's a worksheet I threw together to come up with the pip spreads for the Thursday trades. It's not formatted well, but you can get the drift.

"Close" is the Thursday close. "Weekly Cl." is the Friday close. I subtracted Thursday's number from Friday's to come up with the pip totals.

Don't panic! It doesn't get "good" until page two (like I said, until June it was a mess!)
Attached File(s)
File Type: doc Thursday Trading.doc   35 KB | 565 downloads
Is your SL my TP?
  • #67
  • Jan 11, 2007 9:46pm Jan 11, 2007 9:46pm
Quoting goodthings
What does the excell spread sheet mean?
<table x:str="" style="border-collapse: collapse; width: 693pt;" border="0" cellpadding="0" cellspacing="0" width="923"><col style="width: 56pt;" width="75"><col> <tbody></tbody><col style="width: 42pt;" width="56"><col style="width: 47pt;" width="62"><col style="width: 37pt;" width="49"><col style="width: 51pt;" width="68"><col style="width: 48pt;" span="2" width="64"><col style="width: 44pt;" width="59"><col style="width: 48pt;" span="5" width="64"><col style="width: 37pt;" width="49"><col style="width: 43pt;" width="57"><tbody><tr style="height: 12.75pt;" height="17"></tr></tbody></table>
OK, well, it starts at the beginning of the European session, 8AM GMT, and it then waits the amount of time given by "Wait Mins". If the change in price is more than "Pips", it starts recording. The change in price from that point in time until Midnight GMT is then recorded.
This is done for every day for which the condition holds in each year, and then the numbers are combined to give a mean and standard deviation.

Let me convert the row 145 of the sheet into words: "In 2005, for instrument GBPUSD, whenever the price moved at least 20 pips within the first 180 minutes, the average continuation in pips for the rest of the day was 12.7 pips, with a standard deviation of 72.8 pips".
I will also explain the "significance" column and the orange cells if you like.

This analysis proved to me that objectively cable has the greatest tendency to follow the intraday trend.
  • #68
  • Jan 11, 2007 10:07pm Jan 11, 2007 10:07pm
Waxwing, I am very interested. When you get the time please explain it all.
This thread is very interesting, keep up the good work people.
Pip By Pip
  • #69
  • Jan 11, 2007 11:42pm Jan 11, 2007 11:42pm
Sure, happy to oblige.
The spreadsheet is a condensed result of a lot of work, so unfortunately it needs a lot of explaining.

First, how I produced it: Used Forexite 1 min bar OHLC data on USD major pairs, for 2005 and 2006 up to July (sorry but not more recent).
This data is good enough quality for this analysis (actually it's great, imo, although I'm using OANDA tick data nowadays).

The analysis coding was done in C++ and output to text files, then amalgamated into the spreadsheet. I will try to explain what the code is doing below.

I am trying to test the hypothesis that there is a statistically significant tendency for intraday moves to continue rather than reverse.

Still with me? OK, each column one at a time.
Instrument - self explanatory, Year - likewise.

"Wait mins", "Pips": So, the program will flag "start" at 8AM GMT, and it will add the "wait mins" figure in minutes. When we reach that point, it will find the size of the move in pips. Example: wait minutes is 80, then it will record the price at 8AM GMT, record the price at 9:10GMT (80 minutes later) and subtract the difference.
Suppose that gives us +28 pips.
It will then check the "Pips" value, 4th column. If the Pips value is 30, then this point will be discarded since we didn't get 30 pips difference. But, if the "pips" value is 20, then this data point will be included.

Next, the program will step forward until midnight GMT, the end of that trading day. It will calculate the total return from 9:10GMT (if wait mins is 80, as above) until midnight.
Note that if the original move was +28, as above, the program is going to try going LONG; so it will record the positive return from 9:10 to midnight. That is what's recorded in "Mean profit", the 5th column. If on the other hand the return from 8AMGMT to 910GMT was negative, the program will go SHORT and we will record the negative return. I'm sure you get the idea


Before I describe the rest of the columns, (which are just derived statistics), I suggest you turn on Autofilter in Excel and scan through the results for different currency pairs, and see if you can spot any pattern
  • #70
  • Jan 12, 2007 2:57pm Jan 12, 2007 2:57pm
Waxwing, this is very impressive! Thanks a lot. I'd love to see more of such serious work here.
Although GBPUSD, indeed, seems to be the "trendiest" pair, the results look quite weak to me. We would expect that 2% of the results to be "significant" by chance (judging by the orange marks). In reality, about 6% are "significant". Since the data points are not independent and thus the type II error is probably inflated, I'd say that these results are not very far from random.
In addition, these data per se cannot serve as a basis per trading. For example, the most convincing case is GBPUSD 70 pips after 210 minutes occurs only about 30 times a year. I tried to calculate, what would be the optimal strategy based on this - and could find nothing better than taking profit of 29 p. with a stop loss at -27 p. The success rate (2005-2006) is 58%, so after a 4 p. spread this would make 2 p. per trade...
sans peur et sans reproche
  • #71
  • Jan 12, 2007 3:04pm Jan 12, 2007 3:04pm
waxwing,

thanks for the explanation. continue when ever you are ready.
Pip By Pip
  • #72
  • Jan 12, 2007 10:33pm Jan 12, 2007 10:33pm
Quoting ilanr
Waxwing, this is very impressive! Thanks a lot. I'd love to see more of such serious work here.
Although GBPUSD, indeed, seems to be the "trendiest" pair, the results look quite weak to me. We would expect that 2% of the results to be "significant" by chance (judging by the orange marks). In reality, about 6% are "significant". Since the data points are not independent and thus the type II error is probably inflated, I'd say that these results are not very far from random.
In addition, these data per se cannot serve as a basis per trading. For example, the most convincing case is GBPUSD 70 pips after 210 minutes occurs only about 30 times a year. I tried to calculate, what would be the optimal strategy based on this - and could find nothing better than taking profit of 29 p. with a stop loss at -27 p. The success rate (2005-2006) is 58%, so after a 4 p. spread this would make 2 p. per trade...
ilanr,
Thanks for the critique, it is appreciated
My understanding of statistics is not as deep or broad as it might be. But I would suggest the following:
If you take the mean profit for all the GBPUSD results, you will notice that:

The average value for all results is 16.3 pips
There are only 4 negative results out of 271.
The lowest is -5 pips.
The highest is 45 pips.

For my case, it is impossible to believe that this is random.

If on the other hand you look at USDJPY you see a very different picture! This is more normal, it looks like there might be some very small dependence, for large moves, but nothing that could be relied upon.

I just thought of another way to look at the data. Take the product of wait mins * pips, for all the GBPUSD. Then calculate the correlation with mean pips. It is 39%. I believe this result is not affected by the non-independence of the data, but that's only my intuition. Anyway, it was just an idea.

If the overall number of orange cells (as you can see, I have just looked for cases where the profit is greater than 2*standard deviation of the mean, a pretty arbitrary measure) is not too big compared with random for the whole spreadsheet, I don't find that surprising because when you look at the cross pairs EURJPY and GBPCHF there is basically no evidence of dependency.

Now some details about the construction I previously missed.
Each row is, of course, an average of several days' results. I only included the data if there were more than 20 days data to average, as otherwise it didn't seem worth it.
That's why you don't see a row for all possible combinations of data.
<sorry have="" go="" to="" the="" beach="" will="" continue="" later="">.. Sorry, have to go to the beach, will continue later
</sorry>
  • #73
  • Jan 13, 2007 4:43am Jan 13, 2007 4:43am
OK, so a little more. Sorry if this is annoying, don't want to swamp the thread here with stuff.

So as I was saying I only included rows which were averaging a minimum of 20 data points, as statistics are too dodgy below that. I'm not at home right now and don't have data to hand, but basically "Sdev" records the standard deviation of the profit, as you can see there is plenty of variability compared with the mean value always, as you would expect for such a simple rule. "Sdev mean" records the theoretical standard deviation of the mean value, which is not the same as the standard deviation of the sample. It is recorded as (Sdev/sqrt(n)) where n is the number of data points found for each row. As I said it should be greater than 20 to be recorded. That's as I recall it, it was about 3 months ago I did this.

The columns "MAE" and "MFE" refer to maximum adverse excursion and maximum favourable excursion during the theoretical trade. I think these figures are pretty interesting too. The MAEs aren't bad at all.

Now as to ilanr's suggested usage, I am puzzled about the suggestion of such small targets as 20-30 pips. I can't imagine many intraday strats that are going to get anywhere with orders so close to entry, especially given retail spreads from 2.8 up to 4 and beyond. I don't see why you were thinking to look at such numbers? Also there are other more "smeared" ways to play the effect that would yield much higher trading frequencies.

I have built a strategy based *very loosely" around the effect shown in this data and been trading it for 1.5 months so far.
Results have been somewhat poorer than expected, I'm up only about 3% of account on that strat, I would have expected somewhat more but I also have some evidence of these effects getting weaker, generally, so I'm not hugely surprised.

You may note that the "HANS" strategy on strategybuilder forum, is based on a similar effect.

It is a very uncertain business
The real eureka would be to understand why cable is different from the other pairs! (although eurusd and usdchf also could be said to show the effect, but weaker).

I've probably said all I can for now, let me know if you have other questions about it.
  • #74
  • Jan 14, 2007 3:34am Jan 14, 2007 3:34am
waxwing:
" Now as to ilanr's suggested usage, I am puzzled about the suggestion of such small targets as 20-30 pips. I can't imagine many intraday strats that are going to get anywhere with orders so close to entry, especially given retail spreads from 2.8 up to 4 and beyond. I don't see why you were thinking to look at such numbers? Also there are other more "smeared" ways to play the effect that would yield much higher trading frequencies."

I believe that the important thing is not the target size or pips per day, but rather the winning percentage, the expected win and its proportion to the expected loss size (the risk size). These determine the percent of your capital that you can afford to risk (Kelly's formula). With a substantial winning ratio and a small risk, you can have large profits on small investment - and that is what counts, doesn't it?

Just a small example. I'm trying now a strategy that, in the last two years, ended in winning 96% of cases with TP target of 21 and SL of 46 pips. Apply Kelly's formula and see that you can risk 87% of your capital, that is, theoretically, with a $1000 balance, you can open a position of 18 minilots and to expect to win $324. Of course, there are margin requirements, and natural caution, etc., so winning a quarter of that is more realsitic, which will still make an 8% profit per trade. Well, the right set up for this strategy occurs only once in two weeks, so I still don't have enough experience to say if it really works, but you get the idea: I'll grab any strategy that provides tiny gains with large certainty...
sans peur et sans reproche
  • #75
  • Jan 14, 2007 6:21am Jan 14, 2007 6:21am
Monday the first Day of the week the GbpUsd has:
<?xml:namespace prefix = o ns = "urn:schemas-microsoft-com:office:office" /><o></o>
Moved 100 pips are more 30 times in the last 53 Mondays for a 56.7% probability of 100 pip movement
The average daily range on Monday is 124 pips since 2006
Moved 100 pips are more 12 times in the last 20 Mondays for a 60% probability of a 100 pip movement
The average daily range for the last 20 Mondays is 112 pips.
On a bull run the average movement from previous days high to the days high is 69 pips.
It has moved 70 pips more from days high to next days high 54 times for a probability of 41%
  • #76
  • Jan 14, 2007 7:04am Jan 14, 2007 7:04am
Man you're up early with stats today !!!
  • #77
  • Jan 14, 2007 9:36am Jan 14, 2007 9:36am
Quoting ilanr
waxwing:
" Now as to ilanr's suggested usage, I am puzzled about the suggestion of such small targets as 20-30 pips. I can't imagine many intraday strats that are going to get anywhere with orders so close to entry, especially given retail spreads from 2.8 up to 4 and beyond. I don't see why you were thinking to look at such numbers? Also there are other more "smeared" ways to play the effect that would yield much higher trading frequencies."

I believe that the important thing is not the target size or pips per day, but rather the winning percentage, the expected win and its proportion to the expected loss size (the risk size). These determine the percent of your capital that you can afford to risk (Kelly's formula). With a substantial winning ratio and a small risk, you can have large profits on small investment - and that is what counts, doesn't it?

Just a small example. I'm trying now a strategy that, in the last two years, ended in winning 96% of cases with TP target of 21 and SL of 46 pips. Apply Kelly's formula and see that you can risk 87% of your capital, that is, theoretically, with a $1000 balance, you can open a position of 18 minilots and to expect to win $324. Of course, there are margin requirements, and natural caution, etc., so winning a quarter of that is more realsitic, which will still make an 8% profit per trade. Well, the right set up for this strategy occurs only once in two weeks, so I still don't have enough experience to say if it really works, but you get the idea: I'll grab any strategy that provides tiny gains with large certainty...
You are absolutely right. I was generalizing too much about small targets. The strategy you describe is pretty amazing, even with such low frequency/opportunity
It would have been more careful of me to say that trying to play simple tendency to continue will not allow you to use such tight stops. The effect I observed is hidden behind a *lot* of noise.
  • #78
  • Jan 14, 2007 1:43pm Jan 14, 2007 1:43pm
Quoting goodthings
Monday the first Day of the week the GbpUsd has:
Moved 100 pips are more 30 times in the last 53 Mondays for a 56.7% probability of 100 pip movement
The average daily range on Monday is 124 pips since 2006
Moved 100 pips are more 12 times in the last 20 Mondays for a 60% probability of a 100 pip movement
The average daily range for the last 20 Mondays is 112 pips.
On a bull run the average movement from previous days high to the days high is 69 pips.
It has moved 70 pips more from days high to next days high 54 times for a probability of 41%
Check post number thirty-five on this thread from OrangeRoshan for more info on Sunday/Monday trends. Just follow the thread for a few posts to the following page to get the "Jedi Answer".
Is your SL my TP?
  • #79
  • Jan 14, 2007 1:51pm Jan 14, 2007 1:51pm
nang...!!!! are u there?
  • #80
  • Jan 14, 2007 5:32pm Jan 14, 2007 5:32pm
See Chart
Attached Image
0 traders viewing now