I understand that TR on your website is what is more commonly known in the literature as True Range as defined by Wilder and ATR is Average True Range, again as defined by Wilder. You are using somewhat different names for these things which may cause a bit of confusion... If TR is not True Range (you can google this out), but some other range then please correct me.
Thanks for your reply. One more question: do you calculate these ranges using the daily trading session closing at 17:00 ET or some other time? I am asking this question as the exit is not at the daily session close, but an hour earlier and yet in your article you say at some point that the exit is at the close which might suggest that your ranges are calculated using 16:00 ET as the session close. Just want to clarify this point.
The idea they originally start with is based on market structure and quite valid; it is, essentially, a backdoor bet on volatility. What they do in the subsequent analysis, however, is a case study in how to data-mine your way out of a useful idea and into something that tests well but will under-perform going forward. A rule like "don't trade on Thursdays" is curve-fitting, plain and simple, and the fact that using it doubles the so-called expectancy should be setting off alarm bells. As should the use of a moving average, which, again, is curve-fitting. There are other problems. Two major ones: all the data tested against comes from a period undergoing a secular bear in volatility, and assuming consistently hitting entries and exits when one of the criteria is EOD price. Charting equity curves after psuedo-statistical tweaking of this kind is completely pointless: they will always show a nice, smooth, historical equity curve since that is the implicit optimization being made. Another warning bell, frankly. Most of the issues would have been uncovered if they had done proper out of sample testing. Yes, I know, they did something that kind of looks like it, but in reality it's not, the "out of sample" data is being used implicitly in the original setup. Again, the original well-known concept is a great starting point, unfortunately this article is not an example of how to get to the finish line. In one piece, anyway.
Random - would proper out of sample come from say 1999 or 1989 ie some period COMPLETELY difference from the secular vol bear that was being tested? Also re: Thursdays - doesn't the conclusion here basically become that the REASON the system is profitable is because they eliminated that day? If you allow for realistic slippage and execution issues the rest of it had very little true edge? Finally the most important discovery from data was the dynamic between tight stop and low expectancy return which confirms to me what I have always fekt intuitively that trading is an actuarial business at heart where you collect a lot of small policy premiums and suffer occasional death benefit and that the MOST important aspect of the game is to be as overleveraged as possible in order to absorb the inevitable risk.
This statement makes no sense: you cannot make a statement like that and pretend that it has any predictive power unless you specify the timeframe. What does it mean 'forward'? Months, years, decades? The performance is likely to fluctuate in the long run (years) depending on the market volatility as is the case with most sound systems and because of that using only one system is never a good idea if you want a steady return. Secondly, it seems that based on the past data you want to have something that would not under-perfom going forward and so you want to have something that would perform on a par with the past. Sorry, but even the biggest con artists in this business know that they cannot guarantee the future performance and no one of sound mind expects it. You slammed the system hard in the details and even if I can agree with some of your points (the Thursday effect is unlikely to hold any water in the long run and I would scratch it), you are missing a bigger picture and making unreasonable demands. Even with the Thursday effect neglected, you still have no guarantee that the future performance will be on a par with the past, so I really don't see your point. Systems like that have been around for a long time and they do work. Unless you expect them to NOT under-perform...
The statement that the secular bear has anything to do with the system performance is as suspicious to me as the Thursday effect. This is a volatility based intraday system. Are you suggesting that the long term (monthly) market trend can have impact on this system performance? I would find it very unlikely.
G. Loeb once wrote: "Something that is known to everybody ain't worth anything in the markets". [paraphrased] Keep on doodling along fellas nononsense
MOST important aspect of the game is to be as overleveraged as possible in order to absorb the That should read NOT to be overleveraged. In fact the point I am trying to make is that trading with penny and nickle ticks is probablly the single most important variable to determining your long term success.