Do you think a paper like that would be accepted for publication anywhere? The reason I ask is because of the pattern in figure 6 obtained from the results in figure 5 based on the thinking in figure 3. Further, I am assuming some additional work was done to conclude the range of alpha values suggested. Do you know of any data that is available that qualifies the pattern search software and/or how much capital was put into developing it? TIA
Great POINT!!! I do NOT recommend relying on ANY back-testing if you are going to MANUALLY TRADE. Back Testing ALWAYS assumes automation. If you can't automate the trade - then the Back Test is pointless in creating any expectation. Simply - you can never account for discipline, human error, etc.
Here is a snagit for the pattern's first occurance today on ES. (bars17, 18 and 19) I know those watching got a kick out of the preceding "gyrations. How michael expects anyone to make use of the paper is what I am most puzzled about.
Back Testing is like expecting tyson to win the new fight based on his last bout. Or betting that The Dolphins will win the Superbowl IX because they had a victory win in the VII and VIII super bowls. Past performance does not predict future performance.
Here is one which will not pan out. The reference is bars 21 and 22. For those watching the same order of events occurs on this one with respect to the lateral container that appears before the failure.
for those who wish to avail themselves of powerful aids in trading, back testing is very fruitful. Applying back testing can be done in many ways and it can be used to go from model to development. The person I am responding to is correct that back testing does make it easy to automate. the proposition is that back testing code can be used to set up some automated trading. The paper of Michael Harris is an exception as shown by figures 4, 5 and 6. I made the point by showing a working example and a failure. I know this is all (both winning and losing) within the data of the pattern to net a positive result. The point made was that in the pattern of the example, the pattern has a context and the context determines the pattern's success more than the pattern itself. I am interested in whether or not there are more serious efforts being made that render MH's type work moot.
The last part of the day had three more of these patterns. The first one was the pm BO after settlement (it is attached) The last one is going on as I type. Taking a back test discovery which is at this level probably takes some doing. I posted this as a caution for anyone who might think about trying to find patterns using an analyzer of some sort. An alternative mode of discovery could be critical thinking. The practical matter of trading through a day doesn't bring up very many possibilites for consideration. adults are apable of thinking things through as well If you use this pattern and it fits in to your trading, the first thing to edo is figure out how to "trade it". Obviously, it is not a stand alone tradable from the open of the first bar. Does it ever gain in value for a trader's use?
What you are mistaking is the purpose of statistics. The purpose of statistics is to allow you to interpret data in an intelligent manner. Statistics does not foretell the future. For example. Let's say tomorrow a 1 in a 100 year event happens. When can the next 1 in a 100 year event happen? The day after. And if then it happens again, when can the next 1 in a 100 year event happen? The day after that. In other words the 1 in a 100 year event can happen each and every day for a 100 years. BUT what would happen is that as time passes that event is declassified from 1 in a 100 to 1 in 1 days. Once you can make that classification you would have gone broke and kicked yourself for not understanding the data better. Thus backtesting can only tell you that in the past you would have made X. If the back test is statistically significant, then it says in the past your assumption is pretty much on the money. But it will not tell you what the future holds. So why backtest? So you don't make idiot moves...
Jack, Just forget about the patterns and read the first 7 pages. I think the approach is quite good, rigorous and the author explains well his points. The pattern search software mentioned in that paper is an older version of APS Automatic Pattern Search. It is one of the many tools I use and maybe one of the best I have ever come across in its domain of application. I have no idea how much capital was invested to develop it. I guess you could contact the authors to find out.