I don't see anyone following up on maestro's concept of the arcsin or median deviation ratio concepts. Play with them and ask how they might translate to market strategies(or not). Also, why are the results not intuitive (from a coin toss perspective?). I don't think he'll spoon feed answers, but if you show him work, I think he'll move you closer to the simple answer you want (and everyone benefits). There are times I've disagreed with him, even citing academic papers -- but then, after playing over and over with his ideas I've stumbled upon things I didn't formerly believe. Until I brought up splines, the mean reversion thread would have gone nowhere (he teased for a long time). Now that he acknowledged someone has stumbled upon his thinking, he's not only continued to put the theme in play, but elevated common TA vernacular by about 10X in more ways than you might think. I don't know why you are feeling generous, Maestro, but if you want to divulge more of your wisdom for any reason, I'm very eager to hear it. ---------------------------- If posters are bored, here's something to play with. Since many posters are looking at this from a mean reversion perspective, see if you can understand the reversion directional probability rule. Take a ts generated from a gaussian process (a large one like 1000 steps). Then count how many times the following occur a) p(t-1) < median AND sign(p(t)-p(t-1))= POS b) p(t-1) > median AND sign(p(t)-p(t-1))= NEG Sum both results and divide by the total number of trials. Run it over and over. Does the directional indicator only give a 50% result? Remember, you are generating a probability rule based on random walk distribution.
I guess it is time for me to explain why ⦠As my dear friend Ex University of Toronto professor Burke Brown (he is 82 now and still going strong) I have always believed that opening eyes to the wonderful world of Randomness is my duty and a âhigher purposeâ if you will. And what is a better way to encourage learning and to get people to appreciate this almost completely ignored science other than through the Physical, tangible and so desirable world of trading. I have said it many times: my purpose is EDUCATION. I simply want to make a difference in the minds of a few people as naïve as it sounds. So, for those few who are willing to learn I will gladly share my findings that, believe it or not, are as important intellectually as they are practical. I have chosen this venue simply because there is no better application field than trading to demonstrate the power of true science of Randomness. I am a head of research of a good size hedge fund so I cannot literally disclose the trading algorithms, however, I believe I can seed the hints that could be picked up by a few gifted people (such as yourself) and serve my main purpose without harming my business. I think this is straight forward enough. Here are a few links to my favorite prof home-based sites. http://www.sciencecanada.ca/ http://www.aakkozzll.com/ http://www.eipiphiny.org/ Please pay him a visit and support the wise one! P.S. I am a proud member of eipiphiny Society and we do have a few secrets!
This book http://www.amazon.com/Handbook-Anal...=sr_1_3?ie=UTF8&s=books&qid=1258141096&sr=1-3 has pascal code for natural cubic splines & smoothing splines, if you know a little matrix algebra you should be able to put something together with any general purpose matrix library.
Modern day Pythagorean society. And you sure seem to be fond of the golden mean. hmmmm. As is above, so is below? I also like the little animated quincunx machines used as wallpaper on the aakkozzll site. Thank you.
Naa, it hasn't even started yet.. Dtrader, I've wanted to bang heads with you for awhile, touche... Before this turns into a circle jerk with all of us singing "Kum ba yah" with Maestro as the conductor.. "Take a ts generated from a gaussian process (a large one like 1000 steps)." While I do agree with the idea that counting/medians are better with market data than means...What is the point of an experiment with a prior of assuming gaussian anything? If we know anything we know from t1 to t2 will not be gaussian...the other thing we know is from t2 to t3 will not follow the same distribution as from t1 to t2... If you connect the means of these two distributions with a straight line or a NURB..who cares...a NURB will look cooler, but thats about it as far as information goes.
I think the idea is to get people to focus more on the hypothesis that randomness and edge are not mutually exclusive. And to posit an objective exercise that unlike common TA *ahem* cherry picking, demonstrates something that everyone can reproduce and there is no discrepancy in the conclusions. Doing so flies in the face of common TA, similar to the way that cfd's visual exercise might get some to perceive things differently. It may seem like a simple exercise, but the results are pretty profound, IMO. And how often does someone share something like that? I'm sticking my neck out here subject to the review of everyone viewing. And I know there are sticklers for detail, along with a few very sharp mathematicians, engineers, quants...etc. Is it directly translatable? Maybe, Maybe not, but for starters, it can hopefully give ideas and undo the misconception so many have shown on the board that, bah... are you saying a time series can be random and i can still generate a positive expectation rule set. NO WAY. That is a very bold assertion to make. And is at the heart of one's personal philosophy on markets and gambling. Here is an opportunity to actually prove it to yourselves and the board. An exercise that all should arrive at the same conclusions, without subjectivity. How many TA posts do you see that are not subject to such scrutiny... Well, I made a lot of money, or you just can't see it... etc .. etc.. I don't know about you, but I often begin with simple models and see how markets deviate from the conclusions I find in those models. Then I ask, whether or not there might be any underlying cause or reason to expect that behavior to continue forward, and whether or not I can profit from it. Other people might approach things differently, but that is one approach I have found not only useful, but not subjective. I've been through the TA hoops, and have spent a lot of time deconstructing what I personally believe is a lot of baloney. Regardless, whether I'm wrong or not in that perspective, I think it's only fair to show something subject to scrutiny without subjectivity. I know people get annoyed when others don't give direct systems (fish), but the way I see it, it doesn't do anyone any good to give away the exact recipe, as it kills it for themselves and everyone who struggled to find it. Kind of ironic is it not? What if everyone had 'the' holy grail, what do you think would happen to said recipe. Otherwise, there are ways to engage in beneficial discourse/exchange that do not necessarily rely on giving exact, precise instructions. There is something I kind of feel was given away, but at the same time as Mandlebrot commented; everyone sees their own view depending on their experience, etc.. Regardless, engaging in such topics is useful, IMO. Cheers, dt P.S. Markets are very, very close to gaussian processes, making it a good underlying model to engage. Noble prizes have been awarded based on such simplified modeling. ----------------------------------------------- In so far as a scientific statement speaks about reality, it must be falsifiable; and in so far as it is not falsifiable, it does not speak about reality. Karl Popper
And if one doesn't have access to MatLab, they can trial & error in R at no cost... http://stat.ethz.ch/R-manual/R-patched/library/splines/html/00Index.html http://www.r-project.org/ Slight caution, documentation not nearly as good as ML.
HUH ! And I though my math/stat knowledge is superior -51 -149 -3 -167 -143 out of 20K X 5 random steps. Must be excel bug and does not really generate random stuff.
Well first I will give in and admit it..im a 2 bit pirate out 100% for myself. I care not at all about what anyone else thinks about TA or improvements on it, or experiments that show its whatever. I would love for you to prove markets have anything to do with gaussian process, based off a series that I randomly pick the start and end points to lets say 40 time series on ES data over the next 90 days..I'll even pick them once they already happened(ie historic data)...we can extend this to whatever sample size you want..you know this will never happen though because the entire idea is absurd.. "I don't know about you, but I often begin with simple models and see how markets deviate from the conclusions I find in those models." This line of thought doesn't even compute to me. Strikes me as the optimal way to come up with nonsense. I prefer to first start with market data that is naked as possible, find a pattern, then try to find as many tools as possible that proves I'm full of shit. When I can't find that, I bet.