Statistical Analysis of TZs | Page 4 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Voodoo Science-Statistical Analysis 45 replies

Statistical analysis of trading (PA, MM and more...) 3 replies

Best tools for statistical analysis of tick data and news events? 8 replies

Statistical Analysis of Trading History 9 replies

Statistical analysis fallacy 33 replies

Attachments: Statistical Analysis of TZs
Exit Attachments

Statistical Analysis of TZs

  • #61
  • Apr 28, 2015 6:48pm Apr 28, 2015 6:48pm
Quoting Saturday
{quote} Thanks a lot! I wonder if the probabilities are somewhat similar on lower timeframes too. Installing ninjatrader aws, thank you for the work you've put in.
You're welcome. And good luck with NinjaTrader, there's a bit of a learning curve at first. The main reason I use it is because my broker doesn't support MT. It's also more powerful than MT from what I've heard. But I only use it for analysis - all my live trading is done on my broker's platform.

By the way, the very-long-awaited NT8 (major release!) is supposed to be available next week. I don't know how different it will be to NT7, but you might want to wait for it and skip NT7 entirely

Good luck with it all.
You reap what you sow.
  • #62
  • Apr 28, 2015 8:23pm Apr 28, 2015 8:23pm
Sorry, my information about NinjaTrader 8 was wrong. It looks like the May 4th release will only be to an initial group of beta testers. No release date yet for the general public
You reap what you sow.
  • #63
  • May 1, 2015 1:36am May 1, 2015 1:36am
Do you want to know an interesting statistic about TZs? Well, a couple of weeks ago Jay mentioned a high probability setup for EUR.AUD on his thread. He said that out of 59 similar setups since Jan 2008 on that pair, ALL of them reached target. He also kindly confirmed for me onlist that this was for H4 (M240), and all 60 setups (including the latest one) were based on TZs.

I for one was blown away by that figure. 59 out of 59, in fact 60 out of 60 after his target TZ was swiftly resolved. It's not a huge number, but a 100% hit rate over 7 years is hugely impressive, is it not?

Anyway, if you count the bars om the chart where his target TZ was, you can see that its h value is about 115 (give or take 1 depending on your platform and provider). I didn't think the hit rate was based on First Opposite TZs, due to the less than perfect results posted so far. But I ran the stats for h=115 anyway, just in case.

Bombshell #1
Since Jan 2008 (the same period), there were 60 h=115 first opposite TZs, and 59 of them hit TP! There were zero SLs, and only 1 EX (on 27-Nov-2009) which eventually became an SL. So the hit rate over more than 7 years was 98%. Truly amazing! And this is a fact! You can verify it by scanning the chart for first opposite h=115 TZs if you don't believe me (though it might take a while ).

Bombshell #2
I asked Jay offlist if his stats were based on First Opposite TZs, and he replied that they weren't. They are based on a more complicated system that I won't reveal here, because it's not my system to reveal. And yet, we were able to produce very similar stats using two different TZ systems. Unbelievable!
Some of our target TZs were undoubtedly the same ones, but some weren't. After all, there are several hundred h=115 PTZs on the chart since Jan 2008.

So this is why I'm still interested in TZ analysis. They do seem to hold a few gold nuggets after all, though it takes a lot of digging to find them.
My quest continues
You reap what you sow.
  • #64
  • May 2, 2015 11:36am May 2, 2015 11:36am
Quoting NorthTrader
Do you want to know an interesting statistic about TZs? Well, a couple of weeks ago Jay mentioned a high probability setup for EUR.AUD on his thread. He said that out of 59 similar setups since Jan 2008 on that pair, ALL of them reached target. He also kindly confirmed for me onlist that this was for H4 (M240), and all 60 setups (including the latest one) were based on TZs. I for one was blown away by that figure. 59 out of 59, in fact 60 out of 60 after his target TZ was swiftly resolved. It's not a huge number, but a 100% hit rate over 7 years...

Amazing results NT! Im glad to see you are making a lot of progress in your work. It all has deeply intrigued me. Based on the stats you posted I figured it may be a good idea to my own analysis on the stats. Before I share what I have found I wanted to make sure I had the basic idea of the setup correct. From what I have read is that once price touches the top TZ for the first time, while a bottom TZ is forming, the goal is to trade at the high for the top TZ, and we are only trading the first opposite TZ's that form not any of the ones that form after correct? We are then using the Base TZ as your SL and the New TZ as our TP. Is this correct? Below is a picture of what I mean.

Attached Image (click to enlarge)
Click to Enlarge Name: Screen Shot 2015-05-02 at 10.24.13 AM.png Size: 22 KB


is this the correct setup you guys are talking about?

Also since you posted 98% hit rating on EURAUD through 2009 i thought I would look into this as these are some amazing if not "holy grail" stats.

I have modified an indicator posted by Kprsa that shows the stats of the opposite TZ resolving regardless of its the first one or second one or 3rd. As long as it formed while the base TZ was incomplete, it measures the hit rate of the opposite TZ's. My data sample is relatively small as I am not proficient enough in coding to modify it to look back further but the stats are still pretty interesting.

Here is the hit rate of the opposite TZ's regardless if they are first or second etc. with an h value of 115 like you mentioned.
Attached Image


Here are the stats for the h value = 48 as you and jay previously talked about.
Attached Image


my first impression is i am not surprised to see that taking longer term trades or trades based on a large picture of the market produces more accuracy. Its been proven many times, even in studies on FF that trading any system on a day chart compared to an hour(Larger pictures of the market) will be more accurate.

I am interested to hear your opinion/input also if my first question in regards to how the setup works is correct.

sis.

EDIT: forgot to mention all of those numbers are based on EURAUD 4hr chart.
  • #65
  • Edited 1:16pm May 2, 2015 12:13pm | Edited 1:16pm
Here is something interesting. My calculations may be flawed in the indicator but here are the numbers for base TZ being H 115 Left & Right with the Opposite or target TZ being h left 115 and h right 48.

EURAUD
Attached Image


EURUSD
Attached Image


EURJPY
Attached Image


One flaw in the logic of these numbers is that some of the trades "expire" with out the SL being hit, counting it as a win technically via the program. so the 100% rate on EURUSD means that the stop loss TZ was never hit before the set up expired This does't mean every setup would have been a winner. I need to figure out how to count the expired trades as losers to get a better picture of what the true values are, but it is pretty dang interesting as is.

Again these are based on all opposite TZs that form while the base is still forming, not just the first one.
  • #66
  • May 2, 2015 8:19pm May 2, 2015 8:19pm
Quoting Sis.yphus
Amazing results NT! Im glad to see you are making a lot of progress in your work. It all has deeply intrigued me. Based on the stats you posted I figured it may be a good idea to my own analysis on the stats.
Thank you for picking up the baton and producing some stats of your own, sis. Great stuff! As well as generating new ideas, generating stats on different platforms - in order to confirm or deny mine - was one of the goals of starting this thread. And thank you also for subscribing to me, I really appreciate it. And that goes for everyone who subscribes to me!

Before answering your questions, I have some good news of my own, especially if you can use NinjaTrader. I've decided to post my main indicator here, with source code. I spent some time yesterday taking out the stuff I'm still working on, and tidying it up a bit. It just needs a bit more documentation, then it's ready. If anyone wants to modify it, or convert it to MetaTrader, then that's fine. I'll post it within the next few days hopefully.

And so, on to your questions and observations. Yes, the way you described with your chart example is the way its supposed to work. The Base TZ doesn't have to become a CTZ (confirmed TZ) before the Target TZ can be traded towards. And the first Target TZ after the Base should in theory produce the highest chance of success.

But having said that, there are many ways of doing this and the way mentioned above may not be the best. My indi also measures the other Target TZs (after the first one), which I call "Subsequents". The success rate for these is surprisingly high too, even though eventually the last one will become a CTZ and fail. There are also various parameters to consider, such as how many bars after the Base the first Target appears, and how far price has retraced back to the Base following the Target TZ.

My EUR.AUD 98% figure (over 7 years) is based on First Opposite TZs only, not subsequents, and little else. Retrace can be anything, and bars after the Base can be anything. This means that some of the Targets didn't retrace much, and may not have been tradeable (without huge R:R). But with a success rate of 98% that becomes less important, because it's almost certain that the first target will be resolved whatever the retrace.

And a word of warning: that 98% figure is only for EUR.AUD. I tried h=115 on a few more pairs and the probability was more like 90%. Still pretty good, but allowing for min retrace and R:R you have to be careful.

Interesting stats you posted with Target h-right=48. It's true that most of them resolve within 50% of h-left, but as you pointed out without taking into account all the EXs (expirys) those stats aren't accurate. Personally I've found that allowing every Target TZ 3 weeks to work (fixed h-right = 90) produces best results.

Thanks,
NT
You reap what you sow.
  • #67
  • May 2, 2015 8:27pm May 2, 2015 8:27pm
Quoting NorthTrader
{quote} Ha ha, nice one! If we'd done only that for the last 6 years on EU we'd have been in green Seriously though, if that holds true over many pairs then you might be on to something. Something like "if price moves away from the TZ tip by more than 10%, then there's a good chance it will move even further away (e.g. 30%) before resolving". Like you said earlier, TZs are big supply and demand areas. If one wins over the other, then momentum should carry it further, at least for a while.
Intuitively, if trend following strategies work, then this must be true to some extent. After all the whole trend following idea rests on the premise that as price moves further away from a given point, it has tendency to move even more.
Build good relationships with others.
  • #68
  • May 3, 2015 12:16am May 3, 2015 12:16am
Quoting shellsnail
{quote} Intuitively, if trend following strategies work, then this must be true to some extent. After all the whole trend following idea rests on the premise that as price moves further away from a given point, it has tendency to move even more.
Yep, it's all about momentum. In trading, like in real life, there's no getting away from it. But there are scales of momentum. A retrace after a strong move typically has weak momentum. You can scalp it if that's your style, but such counter-trend trades aren't for everyone. The main move still has plenty of momentum even during the retrace, and if you're patient it can be better to wait for the retrace to finish before entering in the main (stronger) direction. I think a big part of successful trading is knowing the difference between weak momentum and strong momentum.
You reap what you sow.
  • #69
  • May 4, 2015 7:11am May 4, 2015 7:11am
Here is the the latest stable version of my TZJ3 indicator. This one produces the reports posted earlier (but not the double sequence stats near the start of the thread), and the data for the parameter optimisation spreadsheets. It also marks the TZs and various other details on the chart.

You can import it into NinjaTrader using the "File > Utilities > Import NinjaScript..." menu item. Or you can just look at the source code by unzipping the file and opening the TZJ3.cs file in a text editor.

The parameter settings are described in the description panel of the Indicators dialog box (the box that opens when you right-click on the chart and select "Indicators..."). Some descriptions are too big to fit in the standard-size panel, so increase the size of the panel by dragging its top border up to see the full description. Most parameters were also described earlier in this thread.

I've changed the code slightly in an attempt to make it more fair and accurate, such that if a bar hits TP and either Min Retrace or Max Retrace on the same bar, and it didn't close beyond the TZ tip, then it becomes invalid (reason = ambiguous). So some min and max retrace combinations won't exactly match those posted earlier, due to the new ambiguous ones, but they're pretty close.

I've also added a warning message box when running in Optimise Mode, if the number of iterations will be greater than 100. It gives you the chance to abort the optimisation run in this situation, because it might freeze your program (and PC?) for too long.

Note that the optimisation output is a bit crude and only outputs the results to the Output window in CSV format. It's up to you to copy the output to a CSV file, import it into Excel, and smarten it up.

One more thing, the indi is designed to analyse historical data and not to run in live trading (realtime) mode, so it won't refresh the chart and stats as new bars are painted. If you want to refresh the chart and stats just press F5.

Here it is:
Attached File(s)
File Type: zip TZJ3.zip   15 KB | 704 downloads


Please feel free to modify or convert it any way you like, for personal use. If you want to redistribute it please let me know. And please keep my name at the top of the code, because a lot of hard work has gone into it

I'm already working on the next version, which may or may not be released publicly, and I don't want to maintain both versions simultaneously. So this version is released as is. Minor bug fixes and enhancement requests will probably not be done, sorry.

If you have any questions (after you've played with the indi and read this thread), please post them here. Good luck!
You reap what you sow.
  • #70
  • May 4, 2015 11:45am May 4, 2015 11:45am
Quoting NorthTrader
Here is the the latest stable version of my TZJ3 indicator. This one produces the reports posted earlier (but not the double sequence stats near the start of the thread), and the data for the parameter optimisation spreadsheets. It also marks the TZs and various other details on the chart. You can import it into NinjaTrader using the "File > Utilities > Import NinjaScript..." menu item. Or you can just look at the source code by unzipping the file and opening the TZJ3.cs file in a text editor. The parameter settings are described in the description...
Great thread. Thank you very much for your hard work, NT ...
Good friends are more valuable than gold ...
  • #71
  • May 5, 2015 2:17am May 5, 2015 2:17am
Quoting pedma
{quote} Great thread. Thank you very much for your hard work, NT ...
You're welcome pedma

Forgot to say, to get that great EUR.AUD probability I mentioned earlier, try h Min = 114 and h Max = 115, over a range of Jan 2008 to now on M240. All other parameters set to default, though to get 60 valid first opposite TZs you'll need to set min retrace to 0%. A higher min retrace will result in less TPs but still 0 SLs and 1 EX.
You reap what you sow.
  • #72
  • May 13, 2015 12:58am May 13, 2015 12:58am
Hey everyone, how's it going? Have any of you found anything interesting using the indicator I uploaded? Here's a quick update from me.

I've been pretty busy the last couple of weeks enhancing the indicator to test new ideas, then optimising the parameters and analysing the results. I've also been trading more, and have had a good run so far

Regarding the First Opposite TZ system, my aim is to produce positive results (positive expectancy) from the indicator, over many pairs and many years. I also want a good number of setups to trade, not just 1 or 2 per month. I'm not looking for spectacular results, because I have doubts that an automated system can produce those consistently over time. But even mediocre results would be encouraging, because an experienced trader should be able to turn mediocre automated results into decent live trading results by managing the trades (or avoiding them) based on current market conditions.

To that end, this is what I've done:

1) Added more pairs and more years to the test data. There are now 9 pairs (7 majors + 2 crosses) and data since Jan 2008 (almost 7.5 years) for each.

2) Built a simple P/L (profit and loss) metric into the indicator. This is based on the assumption that you risk 1% of your initial account per trade. It's a bit crude and doesn't include compounding, trading costs or margin requirements, but it's only to provide a rough gauge of positive expectancy.

3) Added a Base TZ h-right parameter, to test whether Target PTZs that appear shortly after the Base PTZ are more likely or less likely to be resolved.

4) Added several Dynamic Retrace parameters, to avoid relying on fixed % min and max retrace levels for each trade.

By the way, the indicator only reports 1 trade per setup, whereas in reality you might trade the same setup several times. Again, this doesn't bother me. I'm not trying to build a fully automatic trading system - just something which gives me the confidence to trade the system live.

Have to work now so I'll post the results later.
You reap what you sow.
  • #73
  • Edited 6:56am May 13, 2015 5:14am | Edited 6:56am
So, before I started testing dynamic retrace levels, I spent a bit more time with fixed retrace levels to optimise the existing parameters. A good combination for EUR.USD was 25% min / 90% max. In other words, enter at 25% and SL at 90%. The following P/Ls are for 25-90% across all 9 pairs and 7.35 years, M240 bars:

-14% for First TZs (738 trades = approx 11 trades per year per pair)
-51% for All TZs (3010 trades = approx 45 trades per year per pair)

(note that both of the above figures are negative)

And just to remind you what these results are for:

1) h = 30 to 120+
2) First TZ = first Target TZ within h range following a Base TZ
3) All TZs = first and subsequent Target TZs within h range
4) All other parameters set to default, except Target h-right = 90 (max 3 weeks per trade)

As you can see, the results aren't good. There might be better fixed retrace combinations than this, but I didn't want to spend much time on it because I knew from experience that fixed retrace levels aren't a good way to trade (sorry Fib lovers!). The market is dynamic so the best place to enter and place a stop loss must vary by trade.

I'll post the dynamic retrace results soon.

Edit: corrected min retrace from 20% to 25%
You reap what you sow.
  • #74
  • May 13, 2015 5:37am May 13, 2015 5:37am
Quoting NorthTrader
So, before I started testing dynamic retrace levels, I spent a bit more time with fixed retrace levels to optimise the existing parameters. A good combination for EUR.USD was 20% min / 90% max. In other words, enter at 20% and SL at 90%. The following P/Ls are for 20-90% across all 9 pairs and 7.35 years, M240 bars: -14% for First TZs (738 trades = approx 11 trades per year per pair) -51% for All TZs (3010 trades = approx 45 trades per year per pair) (note that both of the above figures are negative) And just to remind you what these results are...
Hi NorthTrader

How to find the best retrace is too difficult for me,Puzzled me for a long time

Do use what method or indy find a retrace?

thx a lot.
  • #75
  • May 13, 2015 5:40am May 13, 2015 5:40am
Quoting NorthTrader
So, before I started testing dynamic retrace levels, I spent a bit more time with fixed retrace levels to optimise the existing parameters. A good combination for EUR.USD was 20% min / 90% max. In other words, enter at 20% and SL at 90%. The following P/Ls are for 20-90% across all 9 pairs and 7.35 years, M240 bars: -14% for First TZs (738 trades = approx 11 trades per year per pair) -51% for All TZs (3010 trades = approx 45 trades per year per pair) (note that both of the above figures are negative) And just to remind you what these results are...
Looking forward to those results...


Thanks
Solutions ONLY!!!
  • #76
  • May 13, 2015 7:08am May 13, 2015 7:08am
Thanks Bill and blueface. The method of dynamic retraces I'm currently testing is very simple and uses only one indicator. I'm pretty sure that this indi is available on any charting platform. But I don't want to say any more about it than that, sorry... at least not right now.

It probably isn't the best method anyway, and I encourage you to play around with other methods (such as session S/R levels and pivots) in the hope of finding a good one that suits your own style.

Results coming soon!
You reap what you sow.
  • #77
  • May 13, 2015 7:29am May 13, 2015 7:29am
So, here are the dynamic retrace results so far. All other parameters and details are the same as before:

+46% for First TZs (269 trades = approx 4 trades per year per pair)
+86% for All TZs (1429 trades = approx 22 trades per year per pair)

This is much better than fixed retrace! But not spectacular - 86% over 7.35 years isn't exactly stellar. And you have to take all 1429 trades to get that. But for an automated system result, it's pretty encouraging I think.

Still to do:

- finish optimising the retrace parameters
- consider completely different entry and SL techniques
- add basic SL management, such as trailing the stop when price nears TP
- add variable TPs (at the moment they're always fixed at the TZ tip)

I'll be back when I have more results
You reap what you sow.
  • #78
  • May 13, 2015 9:05am May 13, 2015 9:05am
Nice results NorthTrader!

I am still here playing with the idea of TZ and so far have had great results. I am not trading them the same exact way as you are as my indicator measures a different pattern than the first opposite TZ Trade. Based on back test results, if you manually manage the trades based on current market conditions I found it can be very profitable. Also I recommend people do lot limit themselves to just 1 time frame. TZ's are nothing more than price patterns, and patterns are fractal. If you are measuring the right pattern you cans till have 90%+ hit rate between 15m, 30, 60m 240m and 1440m charts. I have taken it to forward testing for May and so far the results have been great.

Testing Began May 1st 2015.

Attached Image (click to enlarge)
Click to Enlarge Name: Screen Shot 2015-05-13 at 8.02.16 AM.png Size: 36 KB


Going to keep froward testing through may and will post the results in here.
  • #79
  • May 13, 2015 7:42pm May 13, 2015 7:42pm
Quoting Sis.yphus
Nice results NorthTrader!
Nice results too, Sis! And your different pattern method is intriguing . But yes, there are many ways to trade these things. Looking forwad to the end of May results!

Your mention of different timeframes reminded me of something I missed off my list of things still to do. That is: investigate statistically (if possible) the effect of trading smaller TFs once a First Opposite TZ appears on a larger one. I'm thinking along the lines of:

1) a valid h30 Tip TZ appears on M240 chart
2) drop down to M1 chart and attempt first trade in direction of TZ tip as conditions allow (decent retrace but not too strong); tight-ish stop (maybe 50% of reward)
3) if M1 works (which it should more times than not), then that's the TZ resolved so wait until the next valid one on M240; if it doesn't work, try again on M5
4) repeat on M15 to M240 until one of them works

Not sure how this will play out in backtesting or live trading, but I intend to find out. It seems rare for price to fly away from a TZ and never retrace back to it at some point, so at least one of the lower TF trades should yield a few pips if handled correctly. On the other hand, this is a kind of "add to the losers" system so it may be better to trade the lower TFs away from the TZ (with, not against, the lower TF trend) until conditions improve. This will make it difficult to analyse programatically because a lot of discretion is required.

Also, I intend to run the current indi on lower TFs with the same h values as above (30 to 120+), which is what I think you were alluding to. I only chose M240 because that TF proved best in initial studies. Also, I'm not a full-time trader so lower TFs are harder for me to focus on. But I'll do it when I have the time, and let you know the results.
You reap what you sow.
  • #80
  • May 13, 2015 10:08pm May 13, 2015 10:08pm
Quoting NorthTrader
So, here are the dynamic retrace results so far. All other parameters and details are the same as before: +46% for First TZs (269 trades = approx 4 trades per year per pair) +86% for All TZs (1429 trades = approx 22 trades per year per pair)
Something else I wanted to mention was that the win rates (% of trades that hit TP) for these positive results varied between 43% and 57% depending on the pair (for the "All TZs" result - 1429 trades). In other words, you had to endure a lot of losers in order to win overall.

The maximum number of consecutive losers per pair was:

AUD.JPY: 5
AUD.USD: 5
EUR.AUD: 6
EUR.USD: 6
GBP.JPY: 4
GBP.USD: 7
USD.CAD: 7
USD.CHF: 5
USD.JPY: 5

And since they are correlated to some extent, you might get a string of losers (or a string of winners) on several pairs at the same time. This could be hard to take both emotionally and financially.

Anyway, just pointing this out. Now back to the analysis... :nerd:
You reap what you sow.
2 traders viewing now