Piligrimus is a neural network indicator. - page 9
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
int red1 = iCustom(NULL,0, "Kristograf",Threshold1,Threshold2,Threshold3,Kn,1000,3,1);
int red2 = iCustom(NULL,0, "Kristograf",Threshold1,Threshold2,Threshold3,Kn,1000,3,2);
int aqu1 = iCustom(NULL,0, "Kristograf",Threshold1,Threshold2,Threshold3,Kn,1000,0,1);
int aqu2 = iCustom(NULL,0, "Kristograf",Threshold1,Threshold2,Threshold3,Kn,1000,0,2);
int blu1 = iCustom(NULL,0, "Kristograf",Threshold1,Threshold2,Threshold3,Kn,1000,2,1);
int blu2 = iCustom(NULL,0, "Kristograf",Threshold1,Threshold2,Threshold3,Kn,1000,2,2);
if(red2<blu2 && red1>=blu1){ condition on the short position opening and long position closing
if(red2>aqu2 && red1<=aqu1){ condition for long position opening and short position closing
Since this indicator is a variation of moving averages, it probably can't do alone.
Thanks for the tip, we will try it, but since this indicator is one of the variations of moving averages, it might not be enough.
Try it in combination with "Piligrimus".
So far the results are not particularly impressive, even in combination with Piligrimus. But, I can say that the signals are much better than with ordinary (and even smoothed) mash-ups.
What do you want? Do you want a primitive indicator to give you a complete picture of the market? To get decent results you need to create a decent expert system, the market is a very complex phenomenon, where economic, political and psychological factors and laws are mixed up, but nevertheless the processes in it are not random, and we perceive them as chaotic and random only because of our limited ability to analyze and trace the deep links that shape them.
int red1 = iCustom(NULL,0, "Kristograf",Threshold1,Threshold2,Threshold3,Kn,1000,3,1);
int red2 = iCustom(NULL,0, "Kristograf",Threshold1,Threshold2,Threshold3,Kn,1000,3,2);
int aqu1 = iCustom(NULL,0, "Kristograf",Threshold1,Threshold2,Threshold3,Kn,1000,0,1);
int aqu2 = iCustom(NULL,0, "Kristograf",Threshold1,Threshold2,Threshold3,Kn,1000,0,2);
int blu1 = iCustom(NULL,0, "Kristograf",Threshold1,Threshold2,Threshold3,Kn,1000,2,1);
int blu2 = iCustom(NULL,0, "Kristograf",Threshold1,Threshold2,Threshold3,Kn,1000,2,2);
if(red2<blu2 && red1>=blu1){ condition on the short position opening and long position closing
if(red2>aqu2 && red1<=aqu1){ condition for long position opening and short position closing
Do not use 2 and 3 bars for analysis, but 2 and 4, because in this case there will be less false triggers when touching, rather than breaking signals.
int red1 = iCustom(NULL,0, "Kristograf",Threshold1,Threshold2,Threshold3,Kn,1000,3,1);
int red2 = iCustom(NULL,0, "Kristograf",Threshold1,Threshold2,Threshold3,Kn,1000,3,3);
int aqu1 = iCustom(NULL,0, "Kristograf",Threshold1,Threshold2,Threshold3,Kn,1000,0,1);
int aqu2 = iCustom(NULL,0, "Kristograf",Threshold1,Threshold2,Threshold3,Kn,1000,0,3);
int blu1 = iCustom(NULL,0, "Kristograf",Threshold1,Threshold2,Threshold3,Kn,1000,2,1);
int blu2 = iCustom(NULL,0, "Kristograf",Threshold1,Threshold2,Threshold3,Kn,1000,2,3);
if(red2<blu1 && red1>blu1){ condition for opening a short position and closing a long one
if(red2>aqu1 && red1<aqu1){ condition for opening a long position and closing a short one
I didn't expect anything supernatural from your indicator, just answered the question
Shniperson 29.05.2009 12:49
And what are the results of EA on this indicator ?
About the 4 bar - tried it, the results are indeed better than the 3, but sometimes there is a lag.
I didn't expect anything supernatural from your indicator, just answered the question
Shniperson 29.05.2009 12:49
What is the result of the Expert Advisor on this indicator?
About 4 bars - tried it, the results are indeed better than 3, but sometimes there is a lag.
I have not tried to use it for TS, so I can't say anything, just experiment.
Try to shift all the values by 1 bar, i.e. instead of 1 -3, 0-2, if there is strong jitter on small timeframes, enter the condition to trade only on the formed bars, but not in the batter, but in the indicator code itself.
Dear ladies and gentlemen, traders and traders, I have to tell you the most unpleasant news: the publication of an article based on the Piligrimus indicator, as the second part and continuation of the article “Is it possible to predict the Forex market, and how to create your own trading strategy” - is cancelled. At the request of working traders, I have already started writing it, all the necessary material for the article was already ready, I had practically completed work on the Piligrimus indicator, it only remained to conduct a few more experiments with the dynamic part built on retraining neural networks, the static part on polynomials, and some modules of the dynamic part are completely completed, but it's not destiny .... My hard drive flew and all my developments have sunk into oblivion. I say this not as an excuse, I really sincerely wanted to summarize my work before the end of my activity in this field, and pass on some of my experience to others. In private correspondence, I have long informed some of the participants in this forum that I have decided to end my activity in the Forex market. The only thing I wanted to do was complete the work on the “ Piligrimus ” indicator, and use a specific example to show that the market is predictable, and the use of neural networks and regression polynomials allows you to get a significant advantage in your work.
Apparently, it was not by chance that I entered the Forex market. Back in the late seventies, I, overwhelmed by a thirst for profit, decided to make a sports lotto prediction system, and become such a simple Soviet millionaire. I managed to develop a system, it reliably gave a forecast for the next draw by 3-4 numbers out of 5, this was enough to earn a stable and decent income when buying a large number of tickets. But the problem turned out to be a large amount of calculations and low performance of computers of that generation. For a full calculation cycle, it was necessary to count more than a day with 100% processor load, I could do this only on weekends, and the completed tickets had to be sent before Friday. So I had to wait with millions, but this work taught me a lot.
Although everyone I spoke to about my work claimed that it was impossible to predict lotto sports, but from the very beginning I intuitively felt that there was a solution, and it was not just my desire, namely, as an inner knowing that I would do it. Human psychology is such that most decisions are at the subconscious level, and if we believe that something is not possible, it will not be possible for us, our subconscious will not try to look for a solution in a direction that is a priori impossible, so it is arranged. Take, for example, a person under hypnosis, he does such things and finds such solutions that he would never do in a normal state, because hypnosis removes barriers, that it is “not possible” to do this. I say this to the fact that what you allow for yourself in the Forex market as achievable, it potentially becomes achievable, if not, then it’s a dead end, no matter how hard you put in, doubts as a result will not allow you to find a solution.
In the late eighties, I decided to go into business and earn millions again, and this time I succeeded, although the motivation was already different, I was no longer interested in personal enrichment, I needed funds to implement my ideas. I assume that my following explanations on this issue will cause a furious wave of indignation on the part of adherents of fundamental science and their energetic finger movements near the temple, but I will venture to say the following ...
Since childhood, fragments of knowledge from past lives have been surfacing in my memory, for me reincarnations are not some exotic theory, this is my own life experience, and although these memories are fragmentary, this does not give reason to doubt their truth. You may vaguely remember some events from your childhood, but you have no doubt that this happened, and you can draw a fairly clear line between what was real and what you invented. One of the most significant memories for me is connected with Atlantis. With all my esoteric worldview, I have a technocratic mindset, and memories of some technical solutions that existed on Atlantis always caused a great desire to implement this in our lives, although everything is certainly not so simple, in order to translate vague and fragmentary knowledge into reality, you need to spend a lot experiments require large funds. In general, the civilization of Atlantis was an order of magnitude superior to our current level of achievements, although it was not technocratic, like us, but a different path of development. Such disciplines in the modern sense as alchemy, magic, science - there were a single whole. Power plants made it possible to convert cosmic energy into controlled gravitational fields and electrical energy. Aircraft used the gravitational field for their flights, nuclear energy and lasers were used, the ability to transmute some materials into others reached a level that modern science does not even dream of, and this extended not only to non-living matter, but also to biological structures. For example, the legends about mermaids, centaurs, satyrs that have come down to us are nothing more than the results of genetic experiments on Atlantis, where people were crossed with animals to get an inferior race of slaves, missing that such mutations affect not only bodies, but also souls, and this already contradicts the Divine plan of evolution, and they paid for it in the end, being destroyed.
I created a whole holding, which included about a dozen enterprises, commercial, manufacturing, research, there were joint ventures, offshore, and the purpose of which was to create an infrastructure for the development and implementation of technologies based on my memories. By today's standards, an annual turnover of $ 3,000,000 is of course pennies, but in 91-92. when today's oligarchs sewed jeans in cooperatives, the factories did not receive salaries for a year, and I privatized a whole mechanical plant for $ 7,500, it was substantial money, and I expected that I would be able to realize my plan. But the time was vague and unpredictable, showdowns with the lads and with the fiscal authorities - were commonplace, and our rulers did not let us relax, arranging one crisis after another. In addition, not all of my team shared my interests and goals, many had more short-term and mercantile goals. In short, problems arose at every turn. Only a number of years later, after analyzing all the past events, I realized that it was not the time to implement these ideas, it was necessary to wait 50-100 years, perhaps then the ground for their implementation would be ready.
I first got acquainted with stock trading in 92, when I graduated from a business school in the USA. But at that time there were no stock exchanges or the Internet in Russia, so I remembered this when I left the business in 2000. When I started working in this field, money ceased to be an end in itself for me, not because I had a lot of them, I had already managed to lose everything that I had by that time, it’s just that their amount could no longer affect my lifestyle, nor on my interests. But, faced with Forex, I felt a challenge, it was said everywhere that it was impossible to predict it, and my intuition, like with sports lotto, clearly told me that I could do it. To be clear, I trust my intuition, many times it literally saved my life. And I undertook to solve this problem, although of course I did not expect that it would drag on for such a long time. I was never interested in trading itself, I wanted to prove the theorem: Forex Market - Predicting! And now, when I was about to publish the results and, having finished this business, start a new life - the hard drive fails and all the data on it disappears, the viruses did what I planned to do a little later. I am not a superstitious person, but in my life I got used to analyze what is happening with all the causal relationships, symbolically perceive certain events, this saved me from having to break through a locked door. And now a situation has arisen similar to the one I described earlier, about attempts to develop new technologies, it seems that the time has not yet come for Forex forecasting. After completing this work and writing an article, I planned to clean the computer from all sorts of Matlabs, Polyanalysts, Terminals, and hedgehogs with them all my developments, so that their spirit would not be left, and so that after a while, when new ideas appeared, there would be no temptation start all over again, otherwise I already had a similar situation. This happened when the metaquotes abandoned the . Before that, I did all the development in Matlab. In it, I made a terminal simulator that could connect to the DC server in real time, receive quotes, make calculations and trade, bypassing the terminal. The rejection of the API was a blow for me, all the developments went to the trash, for almost a year I did not approach the computer at all and wanted to quit everything, but the work had not yet been completed, the theorem had not been proven, I decided to finish the job. But now the situation is somewhat different, I found the answer to all my questions, I lost all interest in doing this a long time ago, so now for sure - that's it. Yes, it looks like the time has come. For the last forty years, it has turned out for me that every ten years I drastically change the scope of my activity. It has long been thought to write an esoteric, science fiction novel about Atlantis, the technologies that were once used on it and that I wanted to implement, I hope that many years later this will inspire someone, as the novels of Jules Verne inspired in their time to create new devices. So I already have a job for the next 10 years.
As a small compensation for not writing an article, I can cite some materials that were prepared for it and were on my old laptop, on which I performed calculations using PolyAnalyst , the license for it was tightly tied to the hardware of this laptop, and besides I haven't used it for anything.
In general, I recommend everyone, especially those who work with neural networks, to use PolyAnalyst . In addition to the fact that it can be used to obtain formalized polynomials of neural networks and linear regression, although the capabilities of the package are much wider and, if desired, it can be used to solve many problems, it can also be used to prepare an optimal set of input features for ordinary retraining neural networks. For example, when training the NN, I used 600 input variables, such a number of input signals for a regular NN is too much, after training I got the polynomial NS corresponding to NS with 4 layers and 28 nodes, i.e. out of 600, the most effective 28 features for solving this problem were selected, which can later be used for retraining neural networks (for this purpose, you can also use the features selected in the LR polynomials). In this example, I present the results of NN training (screenshots of reports at different training intervals), and a LR training report.
According to my observations, the polynomials obtained using PolyAnalyst , although they are not retrained, are not inferior to the retrained NNs obtained on other packages in terms of forecast accuracy and stability. I trained almost all polynomials in the “ Piligrimus ” indicator on EURUSD M1 data from April 28-29, the sample length is 1500 bars, a little more than a day and only a few recent models on 2740 bars data. After training, all models showed stable results on all instruments. and timeframes from M1 to H4, while on H4 the history was for 3 years, and on all timeframes until May 19, i.e. in fact, 1 day of training and 3 years of a stable forward test, although over the past 3 years the market has changed very much. And I am absolutely sure that in the next 10 years the models would show the same stable results. Of course, it is very important for stable operation and effective forecasting to train models correctly, and for this, first of all, you need to correctly form the input data. To this end, I first scaled the input data by analogy with what I did in the article “Principle of superposition and interference of financial instruments”, and then set the scale grid offset in such a way that no matter how the market changes, the data is constantly in the same dynamic range, I refused traditional normalization methods, they distort the data too much. At the next stage, I tried to ensure that the vector with respect to which the training was carried out was completely covered by the input variables, in Fig. 1. - poor overlap, in Fig. 2. - much better, and, accordingly, the training accuracy will be significantly higher (the black line is the vector with respect to which training is carried out, the remaining lines are input signals).
Fig 1.
Fig 2.
" PolyNet Predictor " Report
Text
The last result was obtained on 10.06.2009, 2:44 for variable: PR1_601. 3 hours and 50 minutes elapsed since the start. The process is running on the'World' table with the attributes enabled: PR1_1, PR1_2, PR1_3, PR1_4, PR1_5, PR1_6, PR1_7, PR1_8, PR1_9, PR1_10, PR1_11, PR1_12, PR1_13, PR1_14, PR1_15, PR1_16, PR1_17, PR1_18, PR1_19, PR1_20, PR1_21, PR1_22, PR1_23, PR1_24, PR1_25, PR1_26, PR1_27, PR1_28, PR1_29, PR1_30, PR1_31, PR1_32, PR1_33, PR1_34, PR1_35, PR1_36, PR1_37, PR1_38, PR1_39, PR1_40, PR1_41, PR1_42, PR1_43 PR1_44, PR1_45, PR1_46, PR1_47, PR1_48, PR1_49, PR1_50, PR1_51, PR1_52, PR1_53, PR1_54, PR1_55, PR1_56, PR1_57, PR1_58, PR1_59, PR1_60, PR1_61, PR1_62, PR1_63, PR1_64 PR1_65, PR1_66, PR1_67, PR1_68, PR1_69, PR1_70, PR1_71, PR1_72, PR1_73, PR1_74, PR1_75, PR1_76, PR1_77, PR1_78, PR1_79, PR1_80, PR1_81, PR1_82, PR1_83, PR1_84, PR1_85 PR1_86, PR1_87, PR1_88, PR1_89, PR1_90, PR1_91, PR1_92, PR1_93, PR1_94, PR1_95, PR1_96, PR1_97, PR1_98, PR1_99, PR1_100, PR1_101, PR1_102, PR1_103, PR1_104, PR1_105, PR1_106, PR1_107, PR1_108, PR1_109, PR1_110, PR1_111, PR1_112, PR1_113, PR1_114, PR1_115, PR1_116, PR1_117, PR1_118, PR1_119, PR1_120, PR1_121, PR1_122, PR1_123, PR1_124, PR1_125, PR1_126, PR1_127, PR1_128, PR1_129, PR1_130, PR1_131, PR1_132, PR1_133, PR1_134, PR1_135, PR1_136, PR1_137, PR1_138, PR1_139, PR1_140, PR1_141, PR1_142, PR1_143 PR1_144, PR1_45, PR1_146, PR1_147, PR1_148, PR1_49, PR1_150, PR1_151, PR1_152, PR1_153, PR1_154, PR1_155, PR1_156, PR1_157, PR1_158, PR1_159, PR1_60, PR1_161, PR1_162 PR1_163, PR1_164, PR1_165, PR1_166, PR1_167, PR1_168, PR1_169, PR1_170, PR1_171, PR1_172, PR1_173, PR1_174, PR1_175, PR1_176, PR1_177, PR1_178, PR1_179, PR1_180, PR1_181, PR1_182, PR1_183, PR1_184, PR1_185, PR1_186, PR1_187, PR1_188, PR1_189, PR1_190, PR1_191, PR1_192, PR1_193, PR1_194, PR1_195, PR1_196, PR1_197, PR1_198, PR1_199, PR1_200, PR1_201, PR1_202, PR1_203, PR1_204, PR1_205, PR1_206, PR1_207, PR1_208, PR1_209, PR1_210, PR1_211, PR1_212, PR1_213, PR1_214, PR1_215, PR1_216, PR1_217, PR1_218, PR1_219 PR1_220, PR1_221, PR1_222, PR1_223, PR1_224, PR1_225, PR1_226, PR1_227, PR1_228, PR1_229, PR1_230, PR1_231, PR1_232, PR1_233, PR1_234, PR1_235, PR1_236, PR1_237, PR1_238, PR1_239, PR1_240, PR1_241, PR1_242, PR1_243, PR1_244, PR1_245, PR1_246, PR1_247, PR1_248, PR1_249, PR1_250, PR1_251, PR1_252, PR1_253, PR1_254, PR1_255, PR1_256 PR1_257, PR1_258, PR1_259, PR1_260, PR1_261, PR1_262, PR1_263, PR1_264, PR1_265, PR1_266, PR1_267, PR1_268, PR1_269, PR1_270, PR1_271, PR1_272, PR1_273, PR1_274, PR1_275 PR1_276, PR1_277, PR1_278, PR1_279, PR1_280, PR1_281, PR1_282, PR1_283, PR1_284, PR1_285, PR1_286, PR1_287, PR1_288, PR1_289, PR1_290, PR1_291, PR1_292, PR1_293, PR1_294 PR1_295, PR1_296, PR1_297, PR1_298, PR1_299, PR1_300, PR1_301, PR1_302, PR1_303, PR1_304, PR1_305, PR1_306, PR1_307, PR1_308, PR1_309, PR1_310, PR1_311, PR1_312, PR1_313, PR1_314, PR1_315, PR1_316, PR1_317, PR1_318, PR1_319, PR1_320, PR1_321, PR1_322, PR1_323, PR1_324, PR1_325, PR1_326, PR1_327, PR1_328, PR1_329, PR1_330, PR1_331 PR1_332, PR1_333, PR1_334, PR1_335, PR1_336, PR1_337, PR1_338, PR1_339, PR1_340, PR1_341, PR1_342, PR1_343, PR1_344, PR1_345, PR1_346, PR1_347, PR1_348, PR1_349, PR1_350 PR1_351, PR1_352, PR1_353, PR1_354, PR1_355, PR1_356, PR1_357, PR1_358, PR1_359, PR1_360, PR1_361, PR1_362, PR1_363, PR1_364, PR1_365, PR1_366, PR1_367, PR1_368, PR1_369, PR1_370, PR1_371, PR1_372, PR1_373, PR1_374, PR1_375, PR1_376, PR1_377, PR1_378, PR1_379, PR1_380, PR1_381, PR1_382, PR1_383, PR1_384, PR1_385, PR1_386, PR1_387, PR1_388, PR1_389, PR1_390, PR1_391, PR1_392, PR1_393, PR1_394, PR1_395, PR1_396, PR1_397, PR1_398, PR1_399, PR1_400, PR1_401, PR1_402, PR1_403, PR1_404, PR1_405, PR1_406, PR1_407, PR1_408, PR1_409, PR1_410, PR1_411, PR1_412, PR1_413, PR1_414, PR1_415, PR1_416, PR1_417, PR1_418, PR1_419, PR1_420, PR1_421, PR1_422, PR1_423, PR1_424, PR1_425, PR1_426, PR1_427, PR1_428, PR1_429, PR1_430, PR1_431, PR1_432, PR1_433, PR1_434, PR1_435, PR1_436, PR1_437, PR1_438, PR1_439, PR1_440, PR1_441, PR1_442, PR1_443, PR1_444 PR1_445, PR1_446, PR1_447, PR1_448, PR1_449, PR1_450, PR1_451, PR1_452, PR1_453, PR1_454, PR1_455, PR1_456, PR1_457, PR1_458, PR1_459, PR1_460, PR1_461, PR1_462 PR1_463, PR1_464, PR1_465, PR1_466, PR1_467, PR1_468, PR1_469, PR1_470, PR1_471, PR1_472, PR1_473, PR1_474, PR1_475, PR1_476, PR1_477, PR1_478, PR1_479, PR1_480, PR1_481 PR1_482, PR1_483, PR1_484, PR1_485, PR1_486, PR1_487, PR1_488, PR1_489, PR1_490, PR1_491, PR1_492, PR1_493, PR1_494, PR1_495, PR1_496, PR1_497, PR1_498, PR1_499, PR1_500 PR1_501, PR1_502, PR1_503, PR1_504, PR1_505, PR1_506, PR1_507, PR1_508, PR1_509, PR1_510, PR1_511, PR1_512, PR1_513, PR1_514, PR1_515, PR1_516, PR1_517, PR1_518, PR1_519 PR1_520, PR1_521, PR1_522, PR1_523, PR1_524, PR1_525, PR1_526, PR1_527, PR1_528, PR1_529, PR1_530, PR1_531, PR1_532, PR1_533, PR1_534, PR1_535, PR1_536, PR1_537, PR1_538, PR1_539, PR1_540, PR1_541, PR1_542, PR1_543, PR1_544, PR1_545, PR1_546, PR1_547, PR1_548, PR1_549, PR1_550, PR1_551, PR1_552, PR1_553, PR1_554, PR1_555, PR1_556 PR1_557, PR1_558, PR1_559, PR1_560, PR1_561, PR1_562, PR1_563, PR1_564, PR1_565, PR1_566, PR1_567, PR1_568, PR1_569, PR1_570, PR1_571, PR1_572, PR1_573, PR1_574, PR1_575 PR1_576, PR1_577, PR1_578, PR1_579, PR1_580, PR1_581, PR1_582, PR1_583, PR1_584, PR1_585, PR1_586, PR1_587, PR1_588, PR1_589, PR1_590, PR1_591, PR1_592, PR1_593, PR1_594 PR1_595, PR1_596, PR1_597, PR1_598, PR1_599, PR1_600
Computing process has been started on the local computer.
Parameter
Value
Restriction on F-Ratio
2
Minimum percentage of missing values, %
0
Degree
3
Significance index:
1807
Standard error:
0.01101
R-squared:
0.9999
Standard deviation:
4.287e-005
Number of processed points:
2740
Number of network layers:
1
Number of network nodes:
3
Rule text:
(4.70234e-006 +PR1_471*(1.20598+PR1_471*(8.60911)) +PR1_501*(-0.119136+PR1_501*(3.53506) +PR1_471*(-11.5888)))
Predicted from actual
Predicted and actual from number
Residuals