Scoreboard
Intersting.....
Good Luck! 
Daniel:
Can you post your data as plain text files or csv (comma seperated values) files? I perfer the csv file. They are the common version of data structure to apply current statistical softwares. If you have csv file of the data, I can use my tools to have a look of it.
niva
Daniel:
Can you post your data as plain text files or csv (comma seperated values) files? I perfer the csv file. They are the common version of data structure to apply current statistical softwares. If you have csv file of the data, I can use my tools to have a look of it.
nivamt4 saves the candle data in csv as a default, but then you need to make sure that you get the same candles for each pair. Here's the main sheet on the document in csv (hopefully without errors)
me too mate, me too.
Time for an update. Updated test file is in the first post
I've been working on interpreting the results of the test.
I had two ideas: first one was to check whether one currency could be traded against the others the following day it scored a "4". The results are on sheets marked with single currencies.
The test checks how many fours were scored, and what came after: how many threes, twos, ones and zeros. The score table is followed with simple stats showing numbers and %s of events.
Each sheet has also tests of what happened after scoring 3s, 2s, 1s and zeros.
These stats aren't conclusive in my opinion, although I would appreciate a discussion because I thought the results should be similar for all five currencies but they aren't. The GBP seems to like to follow the previous day's direction, and USD has as many continuations as reversals and therefore one day's win or loss is of no matter the next one.
One solution to this would be using more data, because this test was done on a period of about three recent years only. I am in the process of preparing twelve years of data (so that daily OHLC values match for all pairs).
The more interesting part of the test is checking the instances when one currency scores a 4 another a 3, and what comes next. This test checks how many times a pair of 4 and 3 occured and how many times it was followed with 4 or more wins out of 7 trades taken the next day.
For instance EUR scores a 4 and GBP a 3. The next day we take 4 trades on eur pairs (long eur) and three on gbp pairs (all excluding eurgbp, because we would go long on it). The test counts all the times when our 7 trades bring 4, 5 6 or 7 wins.
There are ten tests as there are ten pairs. IMO these tests are a little more conclusive as I would trust a bit of a diversification to trading one currency only.
Please analyse the results yourselves if interested and I'll try to answer any questions as you post them. I'm absolutely open to suggestions as how to use them to form a trading system.
I will extend the test with more data in the next few days and update the first post with the new file.
I'm also thinking of opening positions in the direction of winners and leaving them to run for longer than one day.
Any way, good luck and happy pip hunting all 
Here's just my 2 cents
Daniel asked me to post my 25% correction rate failure model. For instance, you can see in this decision tree, when CHF=4, JPY=1, USD=0, the predicted value is EUR=3; with 45 wrong predictions, 63 total predictions under this category . I just forgot this predicted EUR is day1, day2 or day3 prediction.
CHF = 0: 2 (180.0/120.0)
CHF = 1
| JPY = 0: 2 (70.0/43.0)
| JPY = 1: 0 (4.0/3.0)
| JPY = 2: 1 (39.0/26.0)
| JPY = 3
| | USD = 0: 2 (13.0/8.0)
| | USD = 1: 2 (0.0)
| | USD = 2: 2 (11.0/4.0)
| | USD = 3: 2 (0.0)
| | USD = 4: 1 (37.0/26.0)
| JPY = 4
| | USD = 0: 1 (14.0/5.0)
| | USD = 1: 3 (0.0)
| | USD = 2: 3 (17.0/9.0)
| | USD = 3: 3 (27.0/15.0)
| | USD = 4: 3 (0.0)
CHF = 2
| JPY = 0: 2 (82.0/55.0)
| JPY = 1
| | GBP = 0
| | | EUR = 0: 2 (0.0)
| | | EUR = 1: 2 (0.0)
| | | EUR = 2: 2 (0.0)
| | | EUR = 3: 2 (8.0/2.0)
| | | EUR = 4: 1 (6.0/3.0)
| | GBP = 1: 2 (0.0)
| | GBP = 2: 2 (0.0)
| | GBP = 3
| | | EUR = 0: 3 (6.0/2.0)
| | | EUR = 1: 2 (0.0)
| | | EUR = 2: 2 (0.0)
| | | EUR = 3: 2 (0.0)
| | | EUR = 4: 2 (6.0/1.0)
| | GBP = 4
| | | EUR = 0: 2 (3.0/1.0)
| | | EUR = 1: 1 (0.0)
| | | EUR = 2: 1 (0.0)
| | | EUR = 3: 1 (19.0/12.0)
| | | EUR = 4: 1 (0.0)
| JPY = 2
| | EUR = 0: 1 (0.0)
| | EUR = 1: 1 (0.0)
| | EUR = 2: 1 (2.0)
| | EUR = 3: 1 (0.0)
| | EUR = 4: 2 (2.0/1.0)
| JPY = 3: 3 (31.0/21.0)
| JPY = 4
| | USD = 0: 3 (26.0/16.0)
| | USD = 1
| | | EUR = 0: 2 (6.0/3.0)
| | | EUR = 1: 0 (0.0)
| | | EUR = 2: 0 (0.0)
| | | EUR = 3: 0 (6.0/4.0)
| | | EUR = 4: 0 (0.0)
| | USD = 2: 3 (0.0)
| | USD = 3: 2 (11.0/5.0)
| | USD = 4: 3 (0.0)
CHF = 3
| EUR = 0
| | USD = 0: 2 (0.0)
| | USD = 1: 2 (5.0/2.0)
| | USD = 2: 2 (0.0)
| | USD = 3: 1 (1.0)
| | USD = 4: 4 (2.0)
| EUR = 1: 3 (28.0/18.0)
| EUR = 2: 2 (89.0/62.0)
| EUR = 3: 0 (5.0/3.0)
| EUR = 4
| | GBP = 0
| | | USD = 0: 2 (0.0)
| | | USD = 1: 1 (3.0/1.0)
| | | USD = 2: 2 (5.0/1.0)
| | | USD = 3: 2 (0.0)
| | | USD = 4: 2 (0.0)
| | GBP = 1: 1 (23.0/16.0)
| | GBP = 2
| | | USD = 0: 1 (15.0/9.0)
| | | USD = 1: 2 (11.0/7.0)
| | | USD = 2: 1 (0.0)
| | | USD = 3: 1 (0.0)
| | | USD = 4: 1 (0.0)
| | GBP = 3: 1 (0.0)
| | GBP = 4: 1 (0.0)
CHF = 4
| JPY = 0: 3 (39.0/23.0)
| JPY = 1
| | USD = 0: 3 (63.0/45.0)
| | USD = 1: 3 (0.0)
| | USD = 2: 0 (5.0/3.0)
| | USD = 3: 3 (1.0)
| | USD = 4: 3 (0.0)
| JPY = 2
| | USD = 0: 2 (26.0/12.0)
| | USD = 1: 0 (6.0/3.0)
| | USD = 2: 2 (0.0)
| | USD = 3: 2 (1.0)
| | USD = 4: 2 (0.0)
| JPY = 3
| | GBP = 0: 4 (7.0/3.0)
| | GBP = 1: 1 (4.0/1.0)
| | GBP = 2: 2 (6.0/3.0)
| | GBP = 3: 1 (0.0)
| | GBP = 4: 1 (0.0)
| JPY = 4: 3 (0.0)
Correctly Classified Instances 270 27.8064 %
Incorrectly Classified Instances 701 72.1936 %
Niva! Thank you for the input, I'm only having a bit of trouble understanding what are you trying to model here. Are you trying to predict which currency will be strong based on the order of strenght of other currencies? It's a much different approach to what I did. If you have the time, could you explain what your test does? I'm totally unfamiliar with it. 
edit: I've been sitting all weekend over it, but I can't seem t find what I was hoping for. If there are no suggestions, this idea will die.
Have a look here: https://www.mql5.com/en/forum/176969/page7
hmmmm 
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
You agree to website policy and terms of use
I've been trying to produce statistics of single currency strenght in the form of a scoreboard, where there are five currencies against each other. It bases on daily open/close values.
Here's the idea: Five currencies (gbp, eur, chf, jpy and usd) form ten pairs:
EURUSD
EURGBP
GBPUSD
EURCHF
EURJPY
GBPCHF
GBPJPY
USDCHF
USDJPY
CHFJPY
If the eurusd pair closes higher than the open, we allocate 1 point to eur and zero to usd. We collect all the points from all pairs to form the scoreboard:
For instance, all pairs containing eur closed in the direction of eur:
EURUSD 1 0
EURGBP 1 0
EURCHF 1 0
EURJPY 1 0
Similarly for other for four other currencies we check the scores that could look like this:
EUR GBP USD CHF JPY
4 0 3 2 1
which means that eur had 4 wins, gbp lost across the board, and other three were in the middle.
I've been building those stats in an open office document attached below. The scores are in the sheet called "test" at the very right. The question is how to use this data to form a trading edge. My original idea was to check how many points did they score the next day and therefore trade once a day at the EOD in the most likely direction and most likely pairs. My concern is that I might not have enough data, so I will be updating the calc sheet with 12 years of data within the next few days as time permits.
I have seen most other threads with currency strenght indicators, but they do not suit this idea. I would like to open a discussion on how these stats could be used whether on an end of day basis or finding a prevailing trend.
If you can't open this file (you would probably need Open Office (free) for it), give me a shout and I will email you the excel file which is over 15 MB.
It would be interesting to see these scores on a graph, maybe even as an indicator in mt4.