Bravo, Jack! Well done! A tour de force! It grieves me that I can teach you nothing, you who misled me to teach myself so much. Beware ossification. An even worse fate than obfuscation. Exeunt all.
You sure hit the nail on the head with that snack/turd analogy!!! Here's an example. THE SNACK: Use the P,V Boolean relation to catch up with tomorrow's paper today!!! See attached document from http://sputnick5.www8.50megs.com/ Wow, I'm impressed! Base 2 (binary) scoring, Jokari windows, adobe llamingos publishing... it must be good, right? So I analyzed buying the 0 to 7 turn (see diagram on last page) on 1000 stocks from 2000 to 2005 -- a total of 5000 stock-years. How did it do? See next post...
THE TURD: Buying the 0 to 7 turn of the "P,V Boolean relation" and exiting 5 days later* produced the attached equity curve when tested on 1000 stocks over five years. So much for having tomorrow's paper today! * 5 days worked better than 1,2,3, or 4 days.
Does grob speak sideways inclination to twist meaning? Understanding reciprical logic throws most venom.
Thank you for testing that, always was suspicious of this fellow. Reading this post: http://www.elitetrader.com/vb/showthread.php?s=&postid=1168103&highlight=parasitic#post1168103 gives me the impression that he is front running his followers...... Your test results are the confirmation of something that I have noticed in my charts where I consider his entry suggestion to be of unfavourable risk: reward. Looks like another pipedream exposed. wD
So what are the numbers for this back test. 5 years Thats 240 times 5 which is 1200 days. 1000 stocks. that 1000 time 1200 days for 1, 200,000 stock- days. We see that a 5 day hold was "best". For real backtesting the "best" is about 8 days. 23,000 holds occurred of five days each. everyone notices that the hold is a constant number of days for all stocks and 23, 000 events occurred. 23,300 times 5 days is about 116.500 days of being in the market and some days out of the market. The % of time that people are in the market for the back test is: 116,500 divided by 1,200,000 which is under 10% of the time available. Is anyone trading doing position trading where they are in the market 10% of the time according to their back tests? And losing all of the time (net)? 23,300 trades net losses are 400,000 dollars. This is a loss per trade of 17 dollars and 18 cents. How does this compare with any position trader's trading losses or profits? take a look at how you are stopped out on your protection on the last 23,300 trades you did. Something may be missing on the chart. Was each trade for one share? Well 17.18 is a big screw up. Was it a round lot of stocks? A round lot is 100 shares. Well 17.18 is a loss to be sure and it is 17 cents a share on a losing trade. None of can speculate on this backtest; we do not know. People who are successful backtesters do use coding, etc to come as close to the truth as possible. It is not possible to review the coding on this backtest, however. From a practical viewpoint we do have enough information to conclude a few things. As has been pointed out here in ET to the backtester, he has used a data set that is not applicable. Over a five year period it would be possible to have about 100 stocks available at any time and none of them would be workable for the whole period of time. So the basic five year sample of stock days available is 100 times 5 times 240 which is 120,000 stock-days. You cannot use 1000 stocks when their are only about 100 at any given time. The back tester chose about 900 extra unqualified stocks for any given time and , for some reason he kept the selection going for 5 years which is out of the question as well. During that time you would use for trading, in parallel, streams of stocks that were held for a period of time that varied for each stock during its cycle of use. There are always, on any day, several of the 100 that are going to start their respective cycle. So this means that the active money makers at any time may be 5 or 6 or so and that they would be held for a period to complete their half cycle performance. The smaller value would be 5 times 240 times 5 which is 6,000 stock-days of activity. the coorresponding number of cycles 5 divided into 240 and that result times 5 years. We see, obviously that this is 240 and not 23,300. The graph of a backtest would have a horizontal axis of 240 trades over 5 years. The average target for making money is a portion of the cycle that is a minimum of 10%. Two things happen when trading is done. Some trades are exited early and some are held longer. Exited trades have their stock replaced with another available stock. Trades that go longer create fewr trades over time for that stream but not less income or profits. So what is the backtester doing and why is what he does so different than trading the approach recommended? Probably the reason is this. The scoring is a tool and it is not a trading method. Almost any trading method can be used in conjuction with scoring. A way of looking at scoring would be to look at the annotated chart of the trading cycle that includes scoring. This is available on many many websites including ET. The chart included the four variables that are used to get the score. These are what the backtester uses on the 100 or so stocks that are always available to be tested. The Universe of stocks to use for the backtesting is found by ranking 15,000 or so stocks and cremeing off relatively few that are distributed through the 15,000. This wasn't done by the back tester; so far we do not know how he got the list of 1000 instead of a correct list for each day of the five year period. What if a person wanted to test the concept and practice of scoring? How would he go about it? Historically, the work of creating scoring was based upon using three variables at first and then adding the fourth variable as a finness control. Refer to the scoring write up. P, V and Time did not provide a fineness and if fineness were added it could minimize risk. So A/D was added and how it was represented mathematically went through several iterations. This stuff does not appeal very much to some kinds of people. That is clear and will not change. The group of people who make money with high money velocities is very selective by the process of self selection. It does not depend on much more than not being blocked from proceeding further in gaining an understanding of markets and traders. being blocked from going to higher levels of performance comes as a simple consequence of not being able to test personal learning by measuring how you are learning. Some people are learning; others are blocked from learning. In trading, a person can get better and better over time in a variety of ways right up until the person is blocked. Looking at an equity curve, it is easy to see that their are two components. One is the amount of capital that is applied the other is the learning curve of the trader. I do not regard any equity curves based upon edges as being valid. Some people say edges come and go and equity curves reflect that. This type of curve is a curve in the absence of learning. All equity curves continually increase their slope if the person has learned to compound and they further steepen as the person learns (a synthesis of learned elements). Backtesting charts are not equity curves. This one is especially strange since it would ordinarily go horizontal as all failing trading methods do. It is hard to say what kept it going the way it did. It was disconnected from any capital consideration; I can see that. The current DAS (Daily Analysis Sheet) for trading the PVT method allot one column to scoring and it is done automatically from the data feed. The rest of the sheet contains 23 columns. By adding by hand the symbol, you automatically get the score which originally was four columns done by hand. You also get 17 columns of the 22 filled in as a starter. 7 of those columns are from coded routines that are much more complicated that getting the score column. 5 different people from different locations contributed the coding for these 7 columns. the remaining columns(formations buy peak watch for and action are typed in by hand (or hand written. The formations, watch for and action columns follow sequences of word trains. Some assortments (sequences) of possible answers are as long as 30 or more. This means that all of these column fillin can be derived from look up tables and the selection can be made by coding. That work is happening now. Their are three mechanical fully coded alternative software programs available for this now. What it comes down to here is levels of expertise. This repost of an old backtest that was handled elsewhere by successful backtesters and coders is one that doesn't have any expertise related to it. mostly because it is simplistic and arbitrary. The current DAS is a nice and automated sheet that affords anyone using it the opportunity to have a high velocity equity curve that is going to appear to be "unbelievable" to a large segment of the public that is not able to or chooses to not learn enough to understand what the market offers to traders and how expert traders do trading. How things geet to be unbelievable is largely through the comparison process that people do with incoming information. They simply drag out their knwoledge, skills and experience and use it as their standard of measure. Since they also operate in the "rather be right than rich modus", they simply chuck anything that they have not made part of their standard as "unbelievable. Here we see in this thread two persons who are doing this as they agree with the backtester who meets their evaluation standard and they summarily dump on the persons who are doing the trading at the "unbelievable" levels and are posting their work every day in ET. All but two people who are coding the DAS are posters in ET and they are trading at "unbelievable" levels as compared to this backtester, etc.. our IBD Meetup tonight is to get ready for the second half of the year (profitwise). The local office is closed as of tomorrow until after Labor Day. All accounts will get to cash by tomorrow.
Your "tomorrow's paper today" paradigm is still a turd with an 8 day hold. What I tested was YOUR methodology (attached). And it contains NO reference to 23 columns or ANY screening criteria that are to be applied before scoring based on P,V, and A/D. One would think that something touted as "tomorrow's paper today" would at least provide practitioners with a tiny edge but it does not. No amount of obfuscation can hide that simple fact.