Home › Forums › General Discussions › The similarity system – discussion
This topic has been reported for inappropriate content
Tagged: calculating probabilites, market DNA, similarity, transient
- This topic has 246 replies, 23 voices, and was last updated 3 years, 7 months ago by
Anti.
-
AuthorPosts
-
Hello everybody,
I start this thread as I read over Eurusdd’s posts at Forexfactory (especially the similarity system) and try to understand what I can. Thus, I’d be grateful if other members of this forum which once studied Eurusdd’s work or those that are currently reading some of his posts would contribute and discuss. Let me start that discussion with one of my unclear points:
Eurusdd mentioned many times that prize will revisit prize areas where dissimilarity started. Those dissimilarities are market inefficiencies. But what are those inefficiencies? What do they mean? And why do market makers have to correct it (although they sometimes seem to forget about it …)?
-
This topic was modified 5 years, 11 months ago by
Anti.
This reply has been reported for inappropriate content.
i m also intrested as it really have postential me not good in english but will try to contrubet
This reply has been reported for inappropriate content.
Let me give you some examples of what Eurusdd said:
- “In theory, efficient markets should close these kinds of windows.” [view post]
- “There will always be loop-holes.” [view post]
- “Market players have many things to correct but sometimes they can’t correct it all and so, after a period they will forget about it.” [view post]
Note, that those inefficiencies are called “windows”, “loopholes”, “sinkholes”, and “wormholes” by Eurusdd.
Does that mean that the market is efficient by means of efficient-market hypothesis?
@fasttrade: What are your questions? I invite you to ask here …This reply has been reported for inappropriate content.
Maybe that forum is really dead. Nevertheless, another attempt to start some discussion …
My next question is regarding the theorem of subsequences, especially on the application of that theorem on 5 randomly chosen numbers.
From this point of view I’d check the last 5 bars (including the most recent one) and if conditions of theorem 1 are not satisfied, I’d check if x4 and x5 or x2 and x4 may give a hint if next candle closes higher or lower. Sometimes x4 and x5 can suggest another direction for next candle than x2 and x4. How would you act in that case? I assume that those situations cause a skulls/star/smiley (can’t remember) during the lifetime of next candle in SSS indicator of @MTH2014 if the current value is above/below the OHLC/4?!
However, Eurusdd wrote further: “So for 10 consecutive highs if the theorem is not true for the first five, then the probability it is false for the second five also is 0.3%” (to be fair, I calculated 0.39%). How does it fit? 10 consecutive highs seem to mean 10 consecutive and thus distinct candles. How would you deal with 10 candles? Would you check if the first five failed and then if the next four don’t satisfy one of the four conditions, you suggest the direction of 10th candle given the values of x7/x9 resp. x8/x9?
-
This reply was modified 5 years, 11 months ago by
Anti.
[ pipatronic ]June 23, 2016 at 8:57 pm #12702This reply has been reported for inappropriate content.
Hi Anti, yes quiet here but I do keep popping in, I am just so busy at the moment, I am finishing off my trading system and have a super programmer doing the clever bit with the code, hope to be trading full time at the end of the year once I have some commitmants out of the way.
All the best
Pip
skype : pipatronic
This reply has been reported for inappropriate content.
@pipatronic: Good luck!
This reply has been reported for inappropriate content.
I think all Theorem 1 says is that out of 5 randomly chosen candles there is a probability of roughly 94 % that anywhere within this bar sequence there is a moment in time which will provide the opportunity to take profit, move your SL, or whatever you think is suitable in that moment.
It only says that you will be able to take an at least tiny bit of profit if you’re waiting for the right moment within 5 bars to do so. Trading cost have to be taken into account after applying this Theorem.
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.
@simplex: An interesting way I’ve never thought about.
This reply has been reported for inappropriate content.
@pipatronic goodluck indeed! Would you be willing to talk about it at all (possibly in private if you’re more comfortable?) or is this all solo secret stuff?
This reply has been reported for inappropriate content.
@Anti: maybe this way of looking at it is not as obvious as I thought it is.
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.
Hmm, just discovered that post and subsequent posts. Seems that SSS is more complex.
This reply has been reported for inappropriate content.
Thank you @fasttrade. The definition of Blackhole in that paper reminds me on absorbing states (see transient/recurrent states introduction).
[ pipatronic ]June 26, 2016 at 7:19 pm #12717This reply has been reported for inappropriate content.
Thanks Anti
CSendo, will put something together when I get chance, will be away fro a bit, unfortunately I will not be able to be specific
Good to see some action here
Regards
pip
skype : pipatronic
This reply has been reported for inappropriate content.
@pipatronic, hope you’re fine …
This reply has been reported for inappropriate content.
I won’t give up. Thus, here’s a more practical question (on the use of the cycle identifier, CI). I refer to that post of Eurusdd, where he wrote about the use of the Length and PrizeActionFilter in combination of EMAs. However, I’m not sure how his description should be understood. As far as I understand it, we have to use the CI with X-1-4 and 1-Y-4, where X and Y refer to different EMA periods. And than?! If we are able to choose the right values of X and Y then most of the spikes of both CIs will occur at the same time or subsequently? Should only those spikes assumed to be valid?!
Would be great if someone could give a hint how to understand the use of Length in conjunction with PrizeActionFilter … and how to find correct values for lower TFs (i.e. M1, M5, M15).
This reply has been reported for inappropriate content.
@Anti: An interesting way I’ve never thought about.
If your interpretation of the theorem regarding trading is different, would you post it?
By the way: I’m well aware of the fact that my interpretation does not take into account intrabar price action, so it’s true only if the range of each candle is zero. For ranges greater than zero, the probability will increase.
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.
My interpretation is that given in post #12685.
Your statement is that one could make a small profit with probability of 0.94 if you’d know the right entry point within 5 candles.
My idea is the prediction of that entry point whenever the conditions are not met for 4 out of 5 bars. Assume that we have 4 candles 1, 2, 3 and 4 (referring to time) and neither condition (i) or (iii) are fulfilled yet. As in 94 % of every 5-candle-chunk at least one of the four conditions (i), (ii), (iii), or (iv) should be met, we can expect that maybe candle 5’s value will give a
- lower value x5 if x1 > x3 or x3 > x4 or a
- higher value x5 if x1 < x3 or x3 < x4
and thus fulfill at least condition (ii) or (iv). However, before we trade we have to check the relation of x1 to x3 (i.e. x1 > x3 or x1 < x3; x5 should be higher than x3 or lower) and compare it to conclustion from x3 and x4 (i.e. x3 < x4 or x3 > x4; x5 should be lower than x4 or higher). If both both cases indicate the same direction for candle 5 we can act accordingly.
Hope, it’s understandable. If not, please feel free and send me a PM.
This reply has been reported for inappropriate content.
Thanks, got it! You wrote:
we can expect that maybe candle 5’s value will give a … higher … lower … value and thus fulfill at least condition (ii) or (iv)
So what would be the probability behind that ‘maybe’? If candles 1 to 4 did not fulfill that 94 % criterion, is the probability for candle 5 to fulfill it higher, or lower, or equal?
IMO, it’s much lower. If those 94 % are correct, then this probability is correct only for the complete set of 5 bars, not for a subset.
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.
No, I don’t think that the probability will be affected. If 1 to 4 don’t fulfill the condition, then we should have a quite high probability that 5 will do it. The probability should be still 94 %. The only thing in which I see a problem is that we are not looking on independent candles if we apply it the way described above. Thus, if we first look on candles 1 to 5 and after 5 closed to 2 to 6, then we double-check candles 2 to 5. Thus the probability of 2 to 6 is not the same as for 1 to 5.
This reply has been reported for inappropriate content.
Ok, so we have different perceptions of that Theorem as related to practical trading.
I see it as a theoretical construct, ignoring intrabar price action for the sake of simplification, used as an example to state that even some randomly chosen real numbers will show certain ‘patterns’ with high probabilities. Please note the term ‘randomly chosen real numbers’: it means that those numbers are not tied to each other, hence the ‘greater than’ logic will show exactly the same probability as the ‘smaller than’ logic at an identical set of numbers.
So, looking at random sets of numbers, there’s nothing wrong with looking at a certain subset twice, e.g. x1 … x5 and x2 … x6, obviously subset x2 … x5 being a member of both complete sets.
All it says is: pure noise (random numbers!), when exploited fast, can lead to a high percentage of winning trades.
Up to this point, it’s purely theoretical, ignoring trading cost, practical terminal limitations, possible margin calls due to temporary drawdown, etc. As a model to think about possible strategies, it’s ok, though.
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.
Yes, that’s true. It is quite theoretical atm. I believe that your randomness assumption if fulfilled if we assume that forex prizes are generated by a Markovian process. However, even Eurusdd said that the assumption of random markets is not true.
This reply has been reported for inappropriate content.
Ahhh yes, and one more thing has to be considered: quotes don’t come in randomly, that’s why the direction of a trade does matter.
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.
‘Randomness’ of those 5 real numbers is not ‘my’ assumption. This is part of the theorem. If those numbers are not random, the probability will be completely different, depending on the function that connects those numbers.
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 topic was modified 5 years, 11 months ago by
-
AuthorPosts
- You must be logged in to reply to this topic.