This topic has been reported for inappropriate content
- This topic has 246 replies, 23 voices, and was last updated 7 years, 6 months ago by
.
- Posts
This reply has been reported for inappropriate content.
that sums it up nicely but nothing is safe as is. strictly speaking about inefficiencies the probability of them correcting can be different for each one. the area before them can give a clue on the quality of the move and likelihood of resolution. you can get your entry and exit from previous points of liquidity usually preceding an inefficiency. imho once you fully understand it it is not any more or less safe than any other Singular method.
“I believe the very best money is made at the market turns. Everyone says you get killed trying to pick tops and bottoms and you make all
your money by playing the trend in the middle. Well for twelve years I have been missing the meat in the middle but I have made a lot of
money at tops and bottoms.”
– Paul Tudor JonesThis reply has been reported for inappropriate content.
Hi members, I NEED YOUR HELP! I’ve just realized that there’s a big gap of the posts (missing) in the “The similarity system” thread. It seems that posts between about the 1st of August 2013 (end of this page) and 16th August 2013 (beginning of that page) have been deleted nearly completely. However, it seems that during that time Eurusdd tried to explain how his revolution principle works (cf. this post).
Has someone of the senior members saved those post and would you please share it?!
Many thanks in advance!!!
This reply has been reported for inappropriate content.
Anti: Try http://penguintraders.com/forums/topic/archive-treasure-chest/, December 09, 2014, link from Jurij, there would be the whole thread
But i am little bit sceptic about eurusdd “secret”. I think he had very interesting and revolutionary view on the market, but some pieces of puzzle missed. Similarity and transient zones systems had big drawdowns.
Maybe, you will find something “extra”. Good lack and happy trading
This reply has been reported for inappropriate content.
@geminicz: Nice try. However, as the document hast been downloaded at the end of 2014 and those posts have been deleted during August 2013, they are not available there. Nevertheless, thanks for your effort!
Yes, you are right. Some parts of the puzzle are missing. But look over the whole thread. There were many people who seem to profited from single strategies (e.g. similarity or TZ theory). Eurusdd said may times that all the concepts he introduced should be connected and linked. For instance, one could use (dis)similarity to identify when a new trend has started and just trade back previously transient prize levels. It’s just one idea how to apply it.
This reply has been reported for inappropriate content.
Hi members, I come back to share some findings regarding the 3-linear subsequence theorem of Eurusdd from the Similarity System thread with you. Here’s what Eurusdd posted:
First of all I’ve tried to reproduce the relative frequency of subsequences that fulfill at least one of those conditions. I did it by writing down all possible permutations of ups (1)/downs (0) of length 5 (using the R command
expand.grid(0:1,0:1,0:1,0:1,0:1)) and counting. There are 32 unique combinations and just 2 of them didn’t fulfill at least one of the above four conditions. Thus, the relative frequency (or probability) equals (32 – 2) / (32) = 0.9375. Thus, this statement seems to be valid.The next step was to survey real forex data in order to study how the sample probability fits this theoretical number (93.75 %). As data source I’ve used open, high, low, and close prizes of the EURUSD M1-, H1-, and D1-chart. Here are the results:
source: Open ————— M1: 0.899, H1: 0.934, D1: 0.920
source: High —————- M1: 0,908, H1: 0.948, D1: 0.940
source: Low —————– M1: 0.906, H1: 0.953, D1: 0.938
source: Close —————- M1: 0.894, H1: 0.928, D1: 0.906
These findings imho suggest to apply the idea of the analysis of 3-linear subsequences especially on higher TFs as the estimated probability for not fulfilling at least one of the four conditions is worse for the M1 data. Additionally, the high and low prizes delivered highest probability estimates.
From that point we could ask how to apply those 3-linear subsequences to our trading. It seems to be quite obvious how Eurusdd would have used it. He wrote in a post which was deleted (fortunately a quotation has been delivered that part):
“For example, suppose you chose your reals [comment: randomly chosen real numbers] to be high of bars. Then on the daily chart of you set n=2 then you need to look in the DNA for 5 bars. Let us suppose the highs of the last 4 bars are 1.2344, 1.2566, 1.2211, 1.2222, ????? What will be ????? Since there must be a decreasing or increasing 3-subsequence, the high of the next bar must be above 1.2222 or below 1.2211. Just two options.”
What does it mean? If we find at any time a cluster of 4 consecutive (open, high, low, or close) prizes and the conditions i) and iii) are not yet met, then we should expect that the next bar (which has not been matured) will fulfill either condition ii) or iv).
Thus, my last question was how big the probability for obtaining condition ii) or iv) was when i) and iii) haven’t been fulfilled. The following numbers give the results for the same data sets used before:
source: Open ————— M1: 0.699, H1: 0.750, D1: 0.718
source: High —————- M1: 0.715, H1: 0.741, D1: 0.763
source: Low —————– M1: 0.708, H1: 0.778, D1: 0.749
source: Close —————- M1: 0.688, H1: 0.739, D1: 0.709
Here you have an edge. Regarding the TF and the data source you chose, you’ll have a success probability of 68 % to 77 %. That is not bad. However it is far away from the suggested 93.75 %.
This reply has been reported for inappropriate content.
It took a long time to understand some of the theories behind Eurusdds similarity methods. Although I’ve just understood a fraction of the whole thing, I was able to create some high probability setups. Two of them are shown below. The first image just depicts ZZ tops and bottoms that are suspect real turning points as they don’t correlate with high/low stochastic readings.
There are two ways to trade those suspect turning points: 1. You could entry whenever prize brakes the high/low indicated by green/orange arrow. 2. You can trade towards those prize levels as prize is expected to take them out by at least 1 Tick with a probability of 0.94.
The second application of the similarity idea is based on a modification of the concepts introduced during the development of the similarity system thread. The arrows in the image below mark some local maxima and minima in real time. Whenever an arrow appears, there’s a 88 % chance to make a profit as least as big as the spread.
Those methods are based on the analyses of my broker’s data (prize, spread).
This reply has been reported for inappropriate content.
Hi Anti
Did you manage to get the missing posts, I was following this thread almost daily when it was happening and have just checked my records which I thought would be complete.
Between post numbers 4785 and 4831 are missing, is this what you mean?
Post 4785 I downloaded on 27/07/2013, we then went on holiday and I picked up downloading again on 15/08/2013 at post 4831. Got to be honest I missed this gap, maybe thats why I “never got the Eureka moment”.
Sorry my forensics do not offer the missing pages, I will keep looking as i did do some reading when on holiday and I wonder if an old phone has by chance some history on it, i will do some digging. If you have located these missing pages then I would appreciate a copy for my records, i do keep going back to this research as I’m certain some nuggets are in there, I just haven’t dug them up yet
Kind regards
This reply has been reported for inappropriate content.
Hi @TradeFinder, yes – that’s exactly the gap I meant. However, I was not able to dig the missing posts. Thus, whenever you find some of those posts, I’d be happy if you’d share it. Many thanks in advance!
This reply has been reported for inappropriate content.
Hi Anti
Well thats a result, I remembered I took a laptop on holiday and got in trouble from the wife about working on holiday!!
Anyway I updated them from XP to windows 7 and dutifully did a backup in windows old, as you do!!
And here are the missing posts. enjoy and feed back if this provides any help. As I mentioned I am looking at this again but from a very simplistic visual approach, early days but welcome any help along this path
This reply has been reported for inappropriate content.
Hi Anti
I am having a problem uploading the 3 pdfs I have of these posts, any idea’s?
This reply has been reported for inappropriate content.
I am having a problem uploading the 3 pdfs I have of these posts, any idea’s?
Thanks for your efforts! Try to send those pdfs by PM to Anti or to me. We’ll post them then. At the moment I have no idea why it does not work.
s.
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)
This reply has been reported for inappropriate content.
I am having a problem uploading the 3 pdfs I have of these posts, any idea’s?
Thanks for your efforts! Try to send those pdfs by PM to Anti or to me. We’ll post them then. At the moment I have no idea why it does not work. s.
Hi
I think I have sent them, I had to send separate as it wouldn’t let me pick more than one, I got two fail to sends so just resent them. Hopefully you have them, please let me know
In the posts there are some attachments, are these needed?, no guarantees but I can have a look if needed
Kind regards
This reply has been reported for inappropriate content.
Thanks @tradefinder! Got 3 pdfs now and try to upload them myself.
Ok, I’m afraid there’s something wrong with the forum software. I can only upload one file at a time.
s.
EDIT: No, the reply dialog accepts exactly one file, but does not show it in the forum. I’ll drop a line @Saver0. Hope he’ll find the time to fix it.
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)
@TradeFinder! That’s really great! Thanks a lot. Could you or @simplex send them to me?
This reply has been reported for inappropriate content.
Hi @Billyon, unfortunately it turned out that these were not the lost posts. Sorry.
This reply has been reported for inappropriate content.
Ok thanks.
I’ve been searching for them as well. I’ll post if I find something.
This reply has been reported for inappropriate content.
@Anti Could you please point out why you consider 2^5 = 32 binary combinations for proving the theorem? What exactly do they encode?
This reply has been reported for inappropriate content.
Hi @Felix,
it’s a logical consequence. The idea is that we are looking at a sequence of 5 candles. We are only interested in the relative rise/fall of each bar, not it’s exact value (e.g. close prize). Thus, we could encode the candles with 0 (candle closes below previous candle’s close) and 1 (candle closes above previous candle’s close). As you’ve realized, for a sequence of size 5 there are 2^5 = 32 unique combinations of those 0’s and 1’s. Now my question was how many of those 32 sequences don’t contain a increasing or decreasing 3-linear subsequence. And that can be only the case in 2 of those 32 sequences – in the sequences 01010 and 10101. But yes, that’s just an approximation as in the sequence 10101 the 2nd 1 could be higher than the 1st 1 and the 5th 1 could be higher than the 3rd 1, too.
Is your question answered?
This reply has been reported for inappropriate content.
Well, here are some other results. Eurusdd said that it is very important to know how prize reached its recent state (prize level) (cf. here). That means one thing: Prize history matters. Eurusdd gave the reason why:
Because top and bottoms like B occur with probability less than 1%, if you define a period which is greater than three bars. Check it yourself. Therefore, since the probability that B is a turning point is less that 1%, price is expected to continue to do what it was doing before it got to B.
Imho that is a very powerful statement as it means that we can estimate the reversal probability if we know how long the current trend existed. Thus, I wanted to test this.
Surely, it is not easy to define trends. But I used ZigZags (arbitrary settings 30-1-1) to determine the begin and end of a trend for used historical data (EURUSD, M1, 238,859 bars). As mentioned above, the number of bars a trend remains intact are of most importance to me as this may allow to calculate reversal probabilities. Thus, I just counted the length of each trend. The results as well as the estimated cumulative distribution function are shown below.
As you can see, the distribution of length of trends is very right skewed. That means that most trends are not that long (median = 34 bars, mean = 42.77). But some of the trends turn out to be very long (maximum trend length in sample = 235 bars, minimum trend length = 1 bar). For instance, from that distribution now it is easy to calculate the probability for the occurence of a particular trend, e.g. a trend of 130 bars just has a probability of p = 0.023. Therefor, whenever a trend with more than 130 bars occur, the probability that a reversal is near becomes quite high.
However, note that the average occurence of real turning points (for the used ZZ settings) is about 2.553 % and thus bigger than those given by Eurusdd.
This reply has been reported for inappropriate content.
Thanks for your extensive reply. I think I understand your encoding but I disagree with the implicit assumpion of a uniform distribution of binary combinations. Computing the relative frequency with (32 – 2) / (32) requires that each binary pattern occurs with the same probability of 1/32. As you pointed out, at least one inequation (i) – (iv) must hold for all binary patterns except for 01010 and 10101 (resp. x1010 and x0101 as the pattern encodes 4 <-relations between 5 elements) which may or may not fulfill at least one condition depending on the relative ordering of non-adjacent elements. But these patterns occure more frequently than 00000 and 11111, representing the strictly monotonic sequences, for example. The {0,1}-encoding omits the relations between those non-adjacent elements.
You can also look at it from a perspective of sorting. There are 5! = 120 permutations for 5 randomly chosen real numbers and we are matching the sorting sequence with 4 (8) partial sorting sequences arising from inequations (i) – (iv): ..1..2..3.., ..1..3..5.., ..2..3..4.., ..3..4..5… (plus all reversed sequences for the >-relations) where (..) is a wildcard for 0 or more indices. The {0,1}-encoding maps 120 permutations non-uniformly to 16 (32) binary patterns. But there are 20 out of 120 permutations where none of the above inequations hold if you choose x1 .. x5 randomly as stated in the theorem. I checked that with some pseudo random numbers and EURJPY data and for the former as well as for data on really small time scales (tick data) the relative frequency converges to 83.33%. On higher time frames, trends seem to be responsible for a stronger ordering which may push the frequency to 93% and above.
This reply has been reported for inappropriate content.
@Felix: Regarding higher probabilities of particular patterns: Yes, you are right. However, I used Eurusdds assumption that you even have an edge if market turns out to be totally random.
You can also look at it from a perspective of sorting. There are 5! = 120 permutations for 5 randomly chosen real numbers and we are matching the sorting sequence with 4 (8) partial sorting sequences arising from inequations (i) – (iv): ..1..2..3.., ..1..3..5.., ..2..3..4.., ..3..4..5… (plus all reversed sequences for the >-relations) where (..) is a wildcard for 0 or more indices. The {0,1}-encoding maps 120 permutations non-uniformly to 16 (32) binary patterns. But there are 20 out of 120 permutations where none of the above inequations hold if you choose x1 .. x5 randomly as stated in the theorem. I checked that with some pseudo random numbers and EURJPY data and for the former as well as for data on really small time scales (tick data) the relative frequency converges to 83.33%. On higher time frames, trends seem to be responsible for a stronger ordering which may push the frequency to 93% and above.[\quote]
A very interesting perspective. Thank you for it! I’m not sure if I should agree with your conclusion. Are higher TFs more sorted? Or are tick charts less random?! You could argue with both cases …
Have you ever used that subsequence thing for trading?
This reply has been reported for inappropriate content.
Are higher TFs more sorted? Or are tick charts less random?! You could argue with both cases …
Maybe this blog post will add some insight to the latest topic of discussion.
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)
This reply has been reported for inappropriate content.
Thanks, my friend! Those findings exactly reflect my thoughts. Micro timeframes are less random than some higher TFs as there are micro trends nearly everywhere (also if the higher TF is ranging).
- You must be logged in to reply to this topic.