Traders' Roundtable

I have added a tutorial to my site on trading equities( stocks) . The tutorial is a case study of a Nasdaq 100 stock-trading system I developed based on the work of Toby Crabel as well as Intermarket analysis. This system also shows the power of being able to optimize across a basket of market.

In this tutorial we discuss many issues including the problem of using split adjusted stock data can lead to errors in back tested results of 300% to 500%. This tutorial shows how we solved these issues. I think it would be fun to discuss some of the issues involved with getting a realistic backtest of a equity trading system in this topic. I am willing to share some of my thoughts with others in this forum.

This system has done well for the past 20 years averaging a compund rate of return of about 22% annually during this period, without dividends.so returns would have been slightly higher with them included. You can see this tutorial at:
http://tradersstudio.com/Default.aspx?tabid=126

Everyone can read the first few pages. If you have registered on TradersStudio.com, which is free you can download the whole tutorial in pdf, which fully discloses the rules to this system.

Murray Ruggiero wrote: Everyone can read the first few pages. If you have registered on TradersStudio.com, which is free you can download the whole tutorial in pdf, which fully discloses the rules to this system.

The problem with stock splits is well-known to anyone who has desired to properly test systems on stock data. One problem I have noticed is that even when backing out splits, unless the underlying data was carefully controlled prior to be being multiplied by the split ratio, there is a loss of resolution due to rounding. As long as the original data is preserved in the database along with the split multiplier, this isn't a problem. Similarly, as long as there are enough decimal places used in the adjusted data to hold the content of the prior, pre-split, time periods, then there isn't a problem with tick accuracy once you back out the pre-split prices. Unfortunately, some data vendors have done neither. Thanks for bringing up this topic for discussion.

Well, I registered for the report, and, to honest, I feel a bit taken by giving up my personal information compared to what I received from the content of this article.

First off, just having the two reported "exit rules" is insane. There is one of two exits: a large percentage of the stock's price or a conditional exit that will not exist for every bar in the trade. There is no continual exit for each bar in the trade other than the large initial stop. That is what the report claims for the exits anyway. This is a very bizarre system whose returns likely depended on huge bull market (the system is long only) to bail it out of a condition of either hitting the large initial exit or waiting for the conditional exit to actually trigger.

These exits are not related to Crabel's exit at all. Similarly, using the next day's open was a critical part of Crabel's ORB work, so not using the open is a rather odd path to choose in historical testing since it would be easy to do.

My expectation was that I would find a substantial portfolio analysis of using a Crabel-type ORB system applied to the Nasdaq100 stocks. Finding the odd exits coupled with simplistic “100 shareâ€

Yes you are correct, I could have entered using an offset off of the open but this would have not made the system truly tradable using end of day data. In fact I could create a better system if I wanted to get in off of the open. Using the true Crabel entry requires getting the open before placing the trades.

In terms of the bull market causing most of the returns, yes that is true but remember the Nasdaq was down 80% from the highs and this system held up very well during a period , it was actually profitable during a period in which most people lost their life savings trading the same stocks. I did not say this was the holy grail, but I though it was a good starting point system and better than most stock trading systems out their, and even better than some which are sold for a sizable amount of money.

In terms of the 100 lots, that only true on the first layer of testing, which is what I gave away in the report. We have money management, which can be set up by dividing your current money between the stocks and purchasing equal dollar amount of each, and reinvesting profits. This logic will allow you to do this on a 1-unit size and increase and decrease units based on available capitol. You can't do this type of analysis using split-adjusted data. In addition this also requires estimating how much money will become available as position close and new ones open. We developed logic to do these estimates in the money management side of this analysis, which was not presented in this report.

I guess my point is I though for an E-Mail address it was a nice give away. If I wanted $3000.00 then it was not. I fully disclosed a nice system which shows some good ideas for building your own. I can make more analysis available to register members of my site and will notify this forum when I post it.

In addition, TradersStudio addresses these issue for you and TradeStation and Metastock does not. You get a TradeByTrade report with actual entry and exit prices and even percent return reports are not available in these other products. The point is that many people might have thrown out a system which looked marginal on split adjusted data and was actually very profitable when handling the issues the correct way. I sure very few people realized these issues existed.

Well, my main point was in reference to the claimed exits for this system. The text says it is either the large % loss value or a conditional exit. In other words there is no profit capture mechanism other than a conditional exit. I would be very suspect of the results obtained from such an exit criteria.