Market Statistics & Probabilities (requests welcome) | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Odds and Edge - Probabilities in Day Trading 1 reply

Best book on randomness in markets and market probabilities? 4 replies

Trading and Probabilities 3 replies

enoc's psychology&probabili ties thread 23 replies

Trader Statistics and Broker Statistics 0 replies

Attachments: Market Statistics & Probabilities (requests welcome)
Exit Attachments

Market Statistics & Probabilities (requests welcome)

  • #1
  • First Post: Edited Aug 27, 2011 5:47pm Jan 3, 2011 12:01pm | Edited Aug 27, 2011 5:47pm
I am going to start considering requests for statistics and probabilities

I will not take all requests but leave a comment and I will consider what I want to test..

The testing is based on 10 years of data. (EUR/USD.. see post #7 more details)

I will mostly consider statistics of daily TF and some higher TF.. (your welcome to request anything tho.. convince me)

----------------------------------------
HOW TO SUBMIT A REQUEST:

Start your post with a title "REQUEST" and kindly leave the exact criteria.. something like "statistics that 2 days with higher highs will be followed by 3rd day of higher high" (like I said I might not test it but I'll consider if it looks like promising data)

(ALSO.state the TF and YEARS somewhere.. otherwise I will test on Daily when you might be thinking 1M or something.. and I will test 10 years of data when you may only want the most recent 2 years..)
----------------------------------------


PS: Others are welcome to share there data if any.. and one last thing... if any of these probabilities help you create a good system the least you could do is share your system with the thread or kindly PM me

AGAIN: Post #7 Raises some important considerations



---------STATISTICS AND PROBABILITIES---------
Here will go the collection of data we come up with... I'll start with something basic...

 

  1. Condition (current day makes Higher High than previous day & price > 21 SMA > 50 SMA)(10 years of data)
    Current Day will Close Above Previous High = 55.21%
    Current Day will Close Below Previous High = 44.79%
    Next Day Will Make A New High = 60.26%
    Next Day Will Fail to Make New High = 39.74%

    Condition (current day makes Lower Low than previous day & price < 21 SMA < 50 SMA)(10 years of data)

    Current Day will Close Below Previous Low = 52.96%
    Current Day will Close Above Previous Low = 47.04%
    Next Day Will Make A New Low = 60.81%
    Next Day Will Fail to Make New Low = 39.19%

  2. Price Returns to Open after Making a New High or Low(10 years of data)

    Price Returns to Open after making New High = 61.46%

    Price Returns to Open after making New Low = 59.74%

    (NOTE: for example, this means after making a new high, price will return to the opening price at least 1 time 60% of the time)

  3. what is the probability of having at least 50pips of breakout distance from either high or low of 2nd bar, provided that the retracement from breakout point is less than 50pips?(10 years of data daily charts)

    result = 47.62%

    Condition: bar3 (anything) , bar2 ( makes higher high & higher low compared to bar3) , bar1 (insidebar), bar0 (today breaks bar1 high)..

    (testing data on 1 hour chart for 10 years)

    probability it will go up 20 pips = 45.94%

    probability it will go down 20 pips = 54.06%

  4. Regarding #3.. of the 60% of times..(10 years of data)

    price makes new high returns to open, chance it.. Returns to last days high again = 68.26%

    price makes new low returns to open, chance it.. Returns to last days low again = 66.58%

  5. Daily candle being the same direction as candle from the previous day.

    45.36%

    Weekly candle being the same direction as candle from the previous week.

    50%

  6. LONDON and NEWYORK stats

    Average pips per breakout:
    (I used net pips to determine trend/breakout.. but results are in actual pips)
    (London breakout define as 3-6am(EST).. Newyork breakout define as 8am-12pm(EST)

    (10 years data)

    London up = 40 pips
    (this means if London breaksout upward.. it averages 40 pip upmove)

    London down = 39 pips
    (this means if London breaksout Downward.. it averages 40 pip down move)etc. for rest of results

    NewYork up = 57 pips
    NewYork down = 56 pips


    (Last 2 years data)

    London up = 54 pips
    London down = 57 pips

    NewYork up = 76 pips
    NewYork down = 73 pips


    NEWYORK & LONDON Copy eachother? :

    (10 Years)

    NEWYORK copied LONDON = 49.19%

    LONDON copied NEWYORK = 48.23%

    (Last 2 Years)

    NEWYORK copied LONDON = 49.92%

    LONDON copied NEWYORK = 49.14%


    (10 years)
    London's direction today will be London's direction tomorrow = 47.73%

    NewYork's direction today will be NewYork's direction tomorrow = 47.82%

  7. Regarding #3 and #5

    I ran the test again.. this is the result when you use the previous bars middle point rather than the newbars open price (which is same as last bars close price)

    so the middle point of previous bar..

    it makes new high.. chance it returns to previous bars mean = 38.92%

    so now its made a new high and return to the mean point.. chance it makes it back to the high = 30.29%

    and if you want to know the percent of times over a 10 year period that price made a new high.. return to mean of previous bar.. and then made it back to the high.. 10.22%

    Just wanted to show the relationship to risk to reward.. or the fact that in previous test 3 and 5 the open price was often near the high resulting in high winning chances.. but considering the risk to reward.. you draw your own conclusions

  8. Will the weekend gap fade?

    Yes it will fade = 53.97% of the time

    NOTE: This is done with 1:1 risk to reward.. so if the weekend gap was 50 pips.. will it fade the 50 pips.. or go the opposite direction 50 pips..

  9. Standard Deviation
    if price goes to +1 SD or -1 SD.. you immediately place a trade in direction of the average, with TP at the average, and SL at +2 or -2 SD (to get 1:1 risk to reward)

    chance you will get your TP = 54.23%

  10. Its Friday and just after London close

    the week was up.. does the EUR/USD retrace in direction before forex closes for the weekend? Will everyone take their profits before the weekend?

    NO! more often than not the last few hours of Friday will continue in the direction of the weeks trend..

    Friday after London closes: (10 years data)

    Continues in weeks direction = 56.10%

    Retraces direction = 43.90%


  • #2
  • Edited 4:05pm Jan 3, 2011 3:51pm | Edited 4:05pm
Added another probability

so far the numbers would suggest.. if you look at Bulletin 2

consider day0 is current day.. day1 is last day

on day0 you would have a 60% chance of hitting a TP 1 pip above day1 high.. (when day1 high > day2 high & day1 price > 21sma > 50sma)

now am I saying to do this.. not really.. its just statistics and just barely better than a coin flip.. plus where do you set your SL (risk to reward)

but don't we all want probabilities in our favor?
  • #3
  • Jan 4, 2011 1:50pm Jan 4, 2011 1:50pm
new data.. about 60% chance price will return to the open at least one time after it makes a new high or low
  • #4
  • Jan 4, 2011 3:38pm Jan 4, 2011 3:38pm
No one else care about statistical / probability advantages in there trading but me?

lol
  • #5
  • Jan 4, 2011 4:20pm Jan 4, 2011 4:20pm
I'm interested.

Which pairs are you studying ?
  • #6
  • Jan 4, 2011 4:23pm Jan 4, 2011 4:23pm
Quoting newyear498
No one else care about statistical / probability advantages in there trading but me?

lol
Well since you are offering such a kind service, I would like to know that given an inside bar, what is the probability of having at least 50pips of breakout distance from either high or low of 2nd bar, provided that the retracement from breakout point is less than 50pips?

Thank you.
  • #7
  • Jan 4, 2011 4:59pm Jan 4, 2011 4:59pm
Quoting Paddy9
I'm interested.

Which pairs are you studying ?
So far I have run the studies on EUR/USD only


but for GPD/USD here are the stats for bullet 3 with same test


Price Returns to Open after making new High = 56.14%

Price Returns to Open after making new Low = 57.06%



As you can see the stats are roughly the same.. I mean its not that dramatic of a shift.. 100% to 20% or something crazy..

so while my public results will be EUR/USD results.. they should be very close statistics to the 4 majors at the very least.. and probably close to any currency stats

I say take things with a grain of salt and always add and subtract 5% to the results

also all reports I make will be 80-95% accurate.. (depending on the type of test and so forth..) so the stats are not 100% themselves.. so again back to adding and subtracting 5% for worst and best case scenarios

EXPECT THE WORST!
  • #8
  • Jan 4, 2011 5:11pm Jan 4, 2011 5:11pm
Quoting ForexQuant
Well since you are offering such a kind service, I would like to know that given an inside bar, what is the probability of having at least 50pips of breakout distance from either high or low of 2nd bar, provided that the retracement from breakout point is less than 50pips?

Thank you.
Hey,
I need to confirm what your saying before I test it..

I'm not sure if you mean 2nd bar, as the bar before or after the inside bar..

I believe your saying that.. we get an inside bar and you want to know if the next bar after the inside bar will breakout atleast 50 pips

by "retracement from breakout point is less than 50pips?" what did you mean?


Id be happy to test this but like I said.. don't want to give you the wrong stats so please lay it so were both on same page

thanks

PS: your referencing daily data as well correct?
  • #9
  • Jan 4, 2011 5:17pm Jan 4, 2011 5:17pm
Nice thread, I will throw some things your way in the near future. Thank you for taking time out of your day to do this.
  • #10
  • Jan 4, 2011 5:21pm Jan 4, 2011 5:21pm
Quoting Sonofsin5
Nice thread, I will throw some things your way in the near future. Thank you for taking time out of your day to do this.
Thank you for your kind remarks..

yes I figured doing statistics would be a good way to actually give something more then a 75% chance of being bullshit back to FF (like some crappy system I make or something) lol

Plus.. with everyones suggestions I get some good things to test that I might not normally think of.. so thats why I started!

I think this thread may one day become a gold mine of information.. because I have yet to find anything like this on the web.. other than some candlestick pattern probabilities..
  • #11
  • Jan 4, 2011 5:32pm Jan 4, 2011 5:32pm
Quoting newyear498
I believe your saying that.. we get an inside bar and you want to know if the next bar after the inside bar will breakout atleast 50 pips

PS: your referencing daily data as well correct?
This is correct. My 2nd bar actually refer to the bar that 'inside'.

Quote
by "retracement from breakout point is less than 50pips?" what did you mean?

It means if the price already down 50pips from the High or up 50pips from the Low but it yet to reach minimum 50 pips of breakout distance then it fails.

Quote
PS: your referencing daily data as well correct?

Up to you.
  • #12
  • Jan 4, 2011 9:20pm Jan 4, 2011 9:20pm
Thanks a lot for this valuable information.
I have a couple of requests:

1) The average number of pips of the london breakout, and the new york breakout.

2) How much of the time both london and ny move in the same direction.

I am looking for data on the 1H TF.
Thanks
  • #13
  • Jan 4, 2011 9:35pm Jan 4, 2011 9:35pm
Quoting stockey
Thanks a lot for this valuable information.
I have a couple of requests:

1) The average number of pips of the london breakout, and the new york breakout.

2) How much of the time both london and ny move in the same direction.

Thanks
Hi Stockey.. these are good questions

but I'll need more specific requirements.. some questions I would have to ask you is..

what is a london breakout.. I mean..

how long do I track the breakout after london opens?

what pip level above and below the open of london is our breakout start?


as for newyork playing micky mouse with london...

do I need to ignore the crossover of london and newyork to track this data?

should I track the entire london and entire newyork... or only the opening hours (since I suspect near end of hours things could wind down)

I have some trouble tracking newyork and london do same thing since if either one ranges it could skew results.. like london open to close is clearly up.. but newyork opens goes up high and near end of day comes flying back down below its open.. which way did it really move? and was the london overlap affecting it..

you.. or anyone else interested for that matter.. comes up with the specific details into this let me know and I will test it.. but I see some problems with number 2.. number 1 you mention should be easier for you to define I imagine

but either way let me know!
  • #14
  • Jan 4, 2011 9:49pm Jan 4, 2011 9:49pm
Quoting ForexQuant
Up to you.
Well I think the TF matters since a 1M can't really produce 50 pips like a daily could

I am testing it on daily.. I am curious to know tho how you figured 50 pips.. why not 25 or 100 and on the 1 hour what would be your x amount of pips

I suppose 50 pips on daily makes since tho since its about 25% of the daily range generally on EUR/USD.. but in your words.. why 50

Oh.. and I'll have the results shortly!
  • #15
  • Jan 5, 2011 12:12am Jan 5, 2011 12:12am
Quoting ForexQuant
Well since you are offering such a kind service, I would like to know that given an inside bar, what is the probability of having at least 50pips of breakout distance from either high or low of 2nd bar, provided that the retracement from breakout point is less than 50pips?

Thank you.
This is results from last 10 years.. you may be disappointed

result = 47.62%


the problem with this test as well is that 10 years of data is only about 120 cases of this scenario since inside bars are rare and then even rarer setups for your condition

hope this helps!
  • #16
  • Jan 5, 2011 2:41am Jan 5, 2011 2:41am
Quoting newyear498
This is results from last 10 years.. you may be disappointed

result = 47.62%


the problem with this test as well is that 10 years of data is only about 120 cases of this scenario since inside bars are rare and then even rarer setups for your condition

hope this helps!
Did you mean that the total number of insider bars is more than 120 cases? Well they should be same because the initial condition is same since I was asking "given an insider bar...".
  • #17
  • Jan 5, 2011 3:01am Jan 5, 2011 3:01am
All I can say is thanks for this thread. This is very helpful and I say more greese to your elbow.

Am sure more request will come around.

Cheers.

Ibuks
  • #18
  • Jan 5, 2011 4:08am Jan 5, 2011 4:08am
One must also take into account the fact that probabilities change as time changes. If you are using 10 years of data for statistics one should have a histogram showing also the time value. E.g. even though there were a 60% for new high when price is above sma50 it might have been that this method would have produced 70% probability 5 years ago and now only 30%. So my point is that these kind of things are dangerous.
  • #19
  • Edited 10:24am Jan 5, 2011 10:06am | Edited 10:24am
Quoting ForexQuant
Did you mean that the total number of insider bars is more than 120 cases? Well they should be same because the initial condition is same since I was asking "given an insider bar...".
(example as high reverse for short)
the way I tested was.. if there was an inside bar.. and the next day broke say the high of the inside bar.. and the retrace before that was less than 50 pips.. the probability that it would break the high + 50 pips..

so there was only about 120 cases of the next day breaking the high of the insidebar while retracement was less than 50 pips.. I thought thats what you ment by "given the retrace is less than 50 pips".. so if it was breaking the high of the inside bar but it was down 51 pips previously from the inside bars high.. no test and excluded from results

so I did filter out some inside bars.. I thought I clarified with you on the proper test.. I should be able to run test again if you correct me

So...

do you want me to just test if given any inside bar it makes 50 pips breakout without factoring in any retrace.. then we would have all inside bars

I could also give you the stats for average breakout..
  • #20
  • Edited 10:29am Jan 5, 2011 10:13am | Edited 10:29am
Quoting serpo
One must also take into account the fact that probabilities change as time changes. If you are using 10 years of data for statistics one should have a histogram showing also the time value. E.g. even though there were a 60% for new high when price is above sma50 it might have been that this method would have produced 70% probability 5 years ago and now only 30%. So my point is that these kind of things are dangerous.
its a good point and one I thought about as well

you may specify what years you want me to test.. I have no problem testing less data.. or maybe I will begin showing results for last 10 years and also last 2 years?

what would you like to see?
0 traders viewing now