Home › Forums › Trading Systems Discussion › Transient & Recurrent Zones
This topic contains 293 replies, has 47 voices, and was last updated by despacito 2 years, 1 month ago.
-
AuthorPosts
-
right :) To be honest I don’t know how some get 90%+ win rate by trading TZ concepts. I’m trading TZ setups with a hit rate around 65-85%, maybe because of the SL? But any profit is good profit
I too am unsure how we can really reach 90% and have found myself in the 60-70% range of TP being hit instead of SL level. but again these are more than enough odds to reap pips all day with proper R:R as you were saying.
right :) To be honest I don’t know how some get 90%+ win rate by trading TZ concepts. I’m trading TZ setups with a hit rate around 65-85%, maybe because of the SL? But any profit is good profit
Honestly: it’s fairly easy to achieve a 90%+ success rate by trading TZ setups. Just trade H1 or H4 timeframe, reduce your profit expectation per trade to only a handful of pips (2 … 5 + spread) and you will have 93 / 95 / 97 %. But imagine the poor R:R ratio!
I also agree that for proper trading with a reasonable R:R the expectations for success rates should be reduced to something between 65 and 80 %. On the long run this is enough to succeed.
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top- and bottom-fishing to people on an ego trip. (Dr. Alexander Elder)
Following statistics are about the PTZ expect setup.
Very good analysis!
One concern that I have is, when we actually take a trade after a TZ is confirmed towards another opposite pTZ, the R:R could be really off erasing any edge we had. That’s the problem that I’m running into. With SL, the edge is almost nonexistent because those few losses can easily erase the wins. The market is highly volatile going towards a level. We need statistics on the actual trades that we would take, the ones with proper R:R. What is the success vs failure of those trades. Then is there an edge? If not, what can we do to have an edge? What I’m thinking is that when the R:R is good, then the chance of it becoming TZ might be around 50% compared to RZ. Know what I mean? I really don’t know the answer to this yet, I need to find out.
To be honest I don’t know how some get 90%+ win rate by trading TZ concepts.
The way to have 90%+ is by increasing the H_Right size. The longer the H_Right is, higher the probability would be.. but more the DD (drawdown) could be
Focus, Patience, Determination & Order in chaos
Hi simplex,
Conclusion 1: The probability of fully resolving is basically independent of the h_past interval chosen.
Conclusion 2: The wider the h_future interval, the higher the probability of fully resolving.
I agree with those conclusions from my analysis. Because of #1 I stopped focusing on h left: see my post on ff in reply to NorthTrader’s post that started me on this way of thinking:
http://www.ff.com/showthread.php?p=7930467#post7930467Basically, the odds of any bar becoming recurrent is approximately the same as any PTZ becoming recurrent. So h right seems to be all that matters.
Conclusion #2 raises a (unsolved) question for me and it sounds like for many others. The goal is to favorably resolve the risk/return part of the equation, rather than just the win% part. In reality the summation of the positive / negative excursions is more important than the actual win% because it’s the summation of these moves that adds up to net profit or loss.
Hi Saver0,
We need statistics on the actual trades that we would take, the ones with proper R:R. What is the success vs failure of those trades. Then is there an edge? If not, what can we do to have an edge? What I’m thinking is that when the R:R is good, then the chance of it becoming TZ might be around 50% compared to RZ. Know what I mean? I really don’t know the answer to this yet, I need to find out.
These are some really good questions, and I think they focus very well on the profit making aspect of identifying these patterns. The real trick is to identify those patterns that have a positive edge over time, rather than just the ones that have a high win % but a poor R:R ratio. And you are right, to take that into account you basically have to test it in terms of maximum positive and negative excursion based on a specific strategy rather than just general statistics. So I think the basic statistics only get us so far.
Maybe we can implement a SL with some statistics added to it. Hope i can get my thoughts into words that will be understandable…otherwise i have to use my “incredible” paint skills :)
Considering a confirmed BTZ with h=30… we got a potential Top TZ as our target. Couldn´t we now expand the h-right to fit the End of our Potential Top tz..
Then we could calculate the probabilty of price hitting the now free space on the right of our confirmed tz..
I think Kprsa talked about that in the similarity thread. … I think i just can´t understand by myself what i ve written…Time for some Paintskillz :)
This wouldn´t help with a bad R:R ratio per se, but that would mostly depend on where price is at the time. I don´t even know if this is mathematical sound..
-
This reply was modified 4 years, 11 months ago by
Bartleby.
"A dream you dream alone is only a dream. A dream you dream together is a reality." (John Lennon)
Anybody who can see with what happens after the red vertical lines?
-
This reply was modified 4 years, 11 months ago by
GU_Night.
Attachments:
You must be logged in to view attached files.There are children playing in the streets who could solve some of my top problems in physics, because they have modes of sensory perception that I lost long ago. ~Oppenheimer
Another great example, right now NZDCAD. Already happend, but it was another great oppotunity with minimal risk.
There are children playing in the streets who could solve some of my top problems in physics, because they have modes of sensory perception that I lost long ago. ~Oppenheimer
right :) To be honest I don’t know how some get 90%+ win rate by trading TZ concepts. I’m trading TZ setups with a hit rate around 65-85%, maybe because of the SL? But any profit is good profit
Honestly: it’s fairly easy to achieve a 90%+ success rate by trading TZ setups. Just trade H1 or H4 timeframe, reduce your profit expectation per trade to only a handful of pips (2 … 5 + spread) and you will have 93 / 95 / 97 %. But imagine the poor R:R ratio! I also agree that for proper trading with a reasonable R:R the expectations for success rates should be reduced to something between 65 and 80 %. On the long run this is enough to succeed.
What we are talking about is including TP and SL. The Kprsa recurrent indi does not take any of these factors into account. To have a real trade setup where you TP has a 90% chance of being hit instead of your SL(so only 10%) is very hard or at least for me it is. so if you trade based on that indi alone there are no probabilities associated with your SL so the chance of PTZ resolve given by the indi is not the true probability of a trade using TP & SL
Thanks myfriend Juri, good work
-
This reply was modified 4 years, 11 months ago by
jacky.
Maybe we can implement a SL with some statistics added to it. Hope i can get my thoughts into words that will be understandable…otherwise i have to use my “incredible” paint skills :) Considering a confirmed BTZ with h=30… we got a potential Top TZ as our target. Couldn´t we now expand the h-right to fit the End of our Potential Top tz.. Then we could calculate the probabilty of price hitting the now free space on the right of our confirmed tz.. I think Kprsa talked about that in the similarity thread. … I think i just can´t understand by myself what i ve written…Time for some Paintskillz :)
You named it – in two different ways!
- Yes: we have to implement both TP and SL and their respective hit probabilities in our stats. Easy to state, sort of challenging to code, IMO.
- In your drawing you mention the need to look at hit probabilities as a function of time. You’re right: this is essential, IMO. At the moment, mostly based on kprsa’s stats indi, we’re looking at hit probabilities in a stationary way. This is a good point to start, but we should try to move on.
I have no stats about this yet, but my feeling from visual analysis says that the probability of a potential fractal zone being revisited will generally degrade over time. I believe that there are two possible driving forces behind this degradation: time itself and / or difference between current price and the outer border (high or low) of the potential zone. Again my feeling says – without having hard stats – that price difference plays the major role.
If this can be proven statistically, and a function describing the degradation process over time can be found, this knowledge may serve as a refinement of a basic formula. So from current point of view, this is second priority, IMO.
First priority should be to define a basic stationary formula that incorporates basic TP and SL.
I started to work on stats about sequences of confirmed top and bottom zones. Intermediate result:
- For most pairs, timeframes, and reasonable ranges of h the probability for a confirmed top being followed by a second confirmed top is somewhere between 20 and 25 %. Same for bottoms.
- For most pairs, timeframes, and reasonable ranges of h the probability for two confirmed tops in a row being followed by a third confirmed top is somewhere between 12 and 18 %. Same for bottoms.
- Those results are valid even for the shortest possible range of h = 1.
Now I have several questions in mind to decide what to code and analyze next. I just throw them in to start a discussion:
- Is it interesting to have some stats about the mean / min / max distance between a pair of subsequent tops / bottoms?
- Is it interesting to have some stats about the mean / min / max number of bars between a pair of subsequent tops / bottoms?
- Which possible TP / SL values are reasonable and worth the effort of coding detailed hit statistics?
-
This reply was modified 4 years, 11 months ago by
simplex. Reason: formatting - this forum software still puzzles me a lot!
-
This reply was modified 4 years, 11 months ago by
simplex. Reason: More precise: this forum software drives me nuts! Doesn't even display my line feeds correctly!!!
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top- and bottom-fishing to people on an ego trip. (Dr. Alexander Elder)
Anybody who can see with what happens after the red vertical lines?
Nice pic, but what’s your point?
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top- and bottom-fishing to people on an ego trip. (Dr. Alexander Elder)
What we are talking about is including TP and SL. The Kprsa recurrent indi does not take any of these factors into account. To have a real trade setup where you TP has a 90% chance of being hit instead of your SL(so only 10%) is very hard or at least for me it is. so if you trade based on that indi alone there are no probabilities associated with your SL so the chance of PTZ resolve given by the indi is not the true probability of a trade using TP & SL
Absolutely correct!
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top- and bottom-fishing to people on an ego trip. (Dr. Alexander Elder)
Hi simplex,
Conclusion 1: The probability of fully resolving is basically independent of the h_past interval chosen. Conclusion 2: The wider the h_future interval, the higher the probability of fully resolving.
I agree with those conclusions from my analysis. Because of #1 I stopped focusing on h left: see my post on ff in reply to NorthTrader’s post that started me on this way of thinking: http://www.ff.com/showthread.php?p=7930467#post7930467 Basically, the odds of any bar becoming recurrent is approximately the same as any PTZ becoming recurrent. So h right seems to be all that matters. Conclusion #2 raises a (unsolved) question for me and it sounds like for many others. The goal is to favorably resolve the risk/return part of the equation, rather than just the win% part. In reality the summation of the positive / negative excursions is more important than the actual win% because it’s the summation of these moves that adds up to net profit or loss.
FXEZ,
Thanks a lot for pointing me to NorthTrader’s post at FF! This is really important, and I’d like to recycle his post here so it doesn’t get lost:
Following on from this post (www.forexfactory.com/showthread.php?p=7830420#post7830420), it seems safe to assume that the probability of returning to (aka “resolving”) a potential TZ within h bars to the right of the middle bar is almost 80%, if h is set such that 3% of all the bars contain full TZs.
But what about the probability of returning to a bar that isn’t a PTZ (potential TZ bar)? In other words, a current bar that is h-left-recurrent, and that might stay that way (only h-left-recurrent), or might become h-fully-recurrent? Well, using the same example bar counts from the last post, we can calculate the empirical probability as 680 / (680 + 180) = 79%.
This figure is almost the same as the other one (78.5%) – is this just coincidence or is there a reason for that!
Anyway, since the current bar can only be one of two types – PTZ and non-PTZ – we can say that the probability of returning to any bar on the chart is always 79%. But this is just an example with 1000 bars. What about a much larger sample with real data?
Well, according to my calculations of 20,000 bars per timeframe, over 6 timeframes (M1, M5, M15, M30, H1 and H4), the average figures for EURUSD are as follows:
1) Probability of revisiting a PTZ is 77.8%
2) Probability of revisiting a non-PTZ is 79.7%<b>Therefore, the probability of fully revisiting any bar on the chart is about 79%.</b>
What an amazing proposition! Why wait for a fairly rare PTZ to form when any bar will give you same the edge?
In my previous posts a stated that I even checked my stats down to h-values as low as 1.
What NorthTrader states here is nothing else than providing stats for h = 0. I already had the idea of doing so before, but I didn’t dare to try!
His results are very interesting, but are they really relevant when it comes to proctical trading and introducing TP and SL hit probabilities in TZ stats?
Just starting to think about this now … hope more of you will join in …
EDIT: no – the above said is not true (partially, at least)! I’d better start thinking before posting, right? There is no such thing as h = 0. So what does ‘revisiting non-PTZ’ mean practically?
-
This reply was modified 4 years, 11 months ago by
simplex. Reason: Addendum. Formatting still drives me nuts here
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top- and bottom-fishing to people on an ego trip. (Dr. Alexander Elder)
- Yes: we have to implement both TP and SL and their respective hit probabilities in our stats. Easy to state, sort of challenging to code, IMO.
- In your drawing you mention the need to look at hit probabilities as a function of time. You’re right: this is essential, IMO. At the moment, mostly based on kprsa’s stats indi, we’re looking at hit probabilities in a stationary way. This is a good point to start, but we should try to move on.
I have no stats about this yet, but my feeling from visual analysis says that the probability of a potential fractal zone being revisited will generally degrade over time. I believe that there are two possible driving forces behind this degradation: time itself and / or difference between current price and the outer border (high or low) of the potential zone. Again my feeling says – without having hard stats – that price difference plays the major role. If this can be proven statistically, and a function describing the degradation process over time can be found, this knowledge may serve as a refinement of a basic formula. So from current point of view, this is second priority, IMO. First priority should be to define a basic stationary formula that incorporates basic TP and SL.
Simplex, regarding your “feelings” I know at least Saver0 and I have done some stats that seem to back up your intuition. Saver’s post is here (see the two picture attachments):
http://www.ff.com/showthread.php?p=7884880#post7884880
This basically shows that as you increase h, the hit probability (probability that price comes back and fills in to the transient zone) increases basically linearly (pic 1), however the # of opportunities decreases roughly exponentially (pic 2).
I ran some slightly different statistics here:
http://www.ff.com/showthread.php?p=7918351#post7918351
Attachment / picture #1 basically shows that the cumulative probability of filling in a potential transient zone increases basically logarithmically in the first half of the curve, but in the 2nd half of the curve, the cumulative probability flattens out much more than the log curve. My interpretation is that in the early part of the curve (lower hright values) it is more favoralbe (from hit rate persepctive, not a R:R perspective) to bet on recurrence, but in the later part of the curve (higher hright values), it is more favorable to bet on transience.
Attachment #2 basically says the same thing by showing non-cumulative probabilities of a PTZ filling in.
In my next post in that thread, I show maximum excursion. Max excursion appears to be fairly linear increasing function over time. Note how the general shape it is fairly similar to Saver0’s 1st picture showing a roughly linear increase in hit rate over time.
And in my next post in the thread, I attempted to put the “fill in” probability together with excursion, so it is more like a system. However I think to get meaningful stats you really have to test a specific strategy in a more systematic manner (keeping track of which comes first the TP hit or the SL hit and then summing up the P&L along the way).
http://www.ff.com/showthread.php?p=7918754#post7918754
The bottom half of the picture / attachment shows:
The bottom chart shows average theoretical profit based on the implied strategy of playing from the average maximum deviation for a given hright, back to the partial transient bar (low). The upper/lower red lines are +/- 4 pips respectively and the yellow line is zero. I note how unstable this seems, underscored by making the avg profit chart a line chart. Not very many consecutive points are outside the +/- 4 pip zone. It appears that a strategy based on the simple concept described by Eurusdd isn’t what those few who have figured it out, are basing their strategy upon. Unless my analysis is incorrect or I made a mistake, either of which is a definite possibility.
I had a disappointing result from this particular experiment but would love for someone to take the R code, tweak it and report back their findings. Or if someone has a better suggestion that builds on this idea I would love to hear it.
-
This reply was modified 4 years, 11 months ago by
FXEZ. Reason: Yes line spacing is messed up
[ Rahat Lukum ]December 19, 2014 at 3:05 pm #1886Now I have several questions in mind to decide what to code and analyze next. I just throw them in to start a discussion:
- Is it interesting to have some stats about the mean / min / max distance between a pair of subsequent tops / bottoms?
- Is it interesting to have some stats about the mean / min / max number of bars between a pair of subsequent tops / bottoms?
- Which possible TP / SL values are reasonable and worth the effort of coding detailed hit statistics?
Nr 2, I did something along those lines, my modest example here… Hmm, how to attach images? Ok, I’ll link it somehow.
Here it is : http://postimg.org/image/opvses9m5/
At one point I looked into nr 3 as well, but proved to be a challenge which I solved in a primitive way, therefore results only scratched the surface. So much I can say that directional probability I got was around 50% for my predetermined SL/TP sets.
-
This reply was modified 4 years, 11 months ago by
Rahat Lukum.
-
This reply was modified 4 years, 11 months ago by
Rahat Lukum.
-
This reply was modified 4 years, 11 months ago by
Rahat Lukum.
-
This reply was modified 4 years, 11 months ago by
Rahat Lukum. Reason: I give up
Anybody who can see with what happens after the red vertical lines?
Nice pic, but what’s your point?
I have no point other than showing good risk reward opportunities with a good probability, in my opinion then.
Another example from today, EURCAD against the short term trend. Stop is a candle close against the entry. Take profit a couple of pips away from totally clearing.
There are children playing in the streets who could solve some of my top problems in physics, because they have modes of sensory perception that I lost long ago. ~Oppenheimer
However I think to get meaningful stats you really have to test a specific strategy in a more systematic manner (keeping track of which comes first the TP hit or the SL hit and then summing up the P&L along the way).
Absolutely right! This leads us directly to our best-loved MetaTrader limitation: tick-based backtests, right?
But for TZ analysis, it should be sufficient to start at H4 timeframe to define the zones and then analyze ‘price action’ on M5 or so. Shouldn’t happen too often that TP and SL are being hit by the very same candle. This is why I’m currently working on a TZ indicator that will visualize zones from a higher timeframe. Didn’t succeed yet. It’s more challenging (for me, at least) than I expected.
This kind of stats wouldn’t be tick based, but it could provide meaningful stats without coding an EA and running backtests.
I had a disappointing result from this particular experiment but would love for someone to take the R code, tweak it and report back their findings. Or if someone has a better suggestion that builds on this idea I would love to hear it.
Hmmm – honestly: R coding is definitely NOT one of my core skills.
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top- and bottom-fishing to people on an ego trip. (Dr. Alexander Elder)
Anybody who can see with what happens after the red vertical lines?
Nice pic, but what’s your point?
I have no point other than showing good risk reward opportunities with a good probability, in my opinion then.
Ahhh – got it, thanks! Could you provide a bit more info about your trading ideas behind those giant pics? That might help to develop this thread towards a sound TZ strategy that finally could be coded.
At the moment you’re posting ‘great opportunities’ in hindsight. Nice to look at, really! But what’s the strategic idea behind? This information would really drive this thread in an interesting direction and I would love to hear more about that!
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top- and bottom-fishing to people on an ego trip. (Dr. Alexander Elder)
I have no point other than showing good risk reward opportunities with a good probability
Is your subsequence indicator something available on the internet somewhere or did you make it? If you made it, would you be interested in sharing it
Looks pretty neat
That was a very nice trade!
-
This reply was modified 4 years, 11 months ago by
Saver0.
Focus, Patience, Determination & Order in chaos
Forgot to quote simplex, sorry. simplex, this one is for you!
Okay simplex, I give it a try with my English. The Darkslategray rectangles are H=20, the arrows from the recurrence indicator are H=36, but the recurrence arrows are not that important to me. So I will not talk about H=36.
I wait for two opposite PTZ’s, once one is confirmed I bet that the other one is cleared within H-bars (or a couple of pips away from totally clearing to raise the changes for a succesful trade), but only if I can find a good Risk Reward opportunity. By the way, I don’t take trades where I think that the PTZ has not a good change to be taken out within H-bars. A PTZ that is totally left behind for example. I only enter if I think that the trade has a good chance to be successful. This has nothing to do with calculations.
Ideal for me is the entry at the confirmed TZ or a couple of pips above the confirmed TZ, the EURCAD entry in the last picture is a good example of the most ideal entry for me.
The vertical lines in the pictures show the confirmation from the H=20 (rectangles) TZ’s. After the vertical lines -and this is how I see it- (another person can see it totally different) does price after a couple of bars ‘realize’ that it have to go the other way to clear the non-confirmed PTZ on the other side. This is in my opinion a very good place and time to enter the trade.
In short; I want to see a (P)TZ on one side that is not taken out within 20-30 bars and a “fresh” PTZ on the other side that is not taken out within only a couple of bars as shown in the EURCAD picture.
At the time from my entry the “old” PTZ is not taken out within 23 bars, and the “fresh” PTZ on the upper side is at the time from entry not taken out within 4 bars. Ideal situation for a good probability(estimating, discretionary) and good risk reward setup, imo.
So far my two cents for now, without the very huge pictures.
-
This reply was modified 4 years, 11 months ago by
GU_Night.
There are children playing in the streets who could solve some of my top problems in physics, because they have modes of sensory perception that I lost long ago. ~Oppenheimer
I have no point other than showing good risk reward opportunities with a good probability
Is your subsequence indicator something available on the internet somewhere or did you make it? If you made it, would you be interested in sharing it
Looks pretty neat
That was a very nice trade!
Hey Saver0, it is MTH (Kiads) indicator. Don’t you have it yet? I don’t know if he shares all his indicators on this forum? Otherwise there must be a way to get it from me or Kiads in another way!
There are children playing in the streets who could solve some of my top problems in physics, because they have modes of sensory perception that I lost long ago. ~Oppenheimer
it is MTH (Kiads) indicator. Don’t you have it yet?
Oh ok, thank you! I will reach out to him
Focus, Patience, Determination & Order in chaos
Forgot to quote simplex, sorry. simplex, this one is for you! …
Thanks a lot, GU_Night! So basically you’re waiting for two opposite PTZs and then set up your trade heading for the older one? Did I get that right?
If yes, does it work and do you have any success stats?
At first you state ‘Darkslategray rectangles are H=20‘, and later ‘I want to see a (P)TZ on one side that is not taken out within 20-30 bars‘. So there seems to be some kind of fine-tuning involved. How do you find your particular value of h? Just visually? Maybe statistically?
And most important: may I interpret your post so that there’s more intuition involved in your entries than hard rules or math? I’m asking this because intuition is a fine tool but I don’t dare to code this in an EA or indicator! When math is involved, coding becomes much simpler!
Cheers, simplex
A good trader is a realist who wants to grab a chunk from the body of a trend, leaving top- and bottom-fishing to people on an ego trip. (Dr. Alexander Elder)
hanks a lot, GU_Night! So basically you’re waiting for two opposite PTZs and then set up your trade heading for the older one? Did I get that right? If yes, does it work and do you have any success stats?
I think it´s waiting for the confirmed older tz..and then targeting the opposite potential tz.
Zelo posted some statistics about that (Zelo please correct me , if i´m wrong) on ff.com
i link this post here.
http://www.ff.com/showthread.php?p=7904030#post7904030
make sure to correct the ff into .. you know what ;)
-
This reply was modified 4 years, 11 months ago by
Bartleby.
"A dream you dream alone is only a dream. A dream you dream together is a reality." (John Lennon)
-
This reply was modified 4 years, 11 months ago by
-
AuthorPosts
You must be logged in to reply to this topic.