Quote from bwolinsky:
It does appear you have no statistical experience, therefore it is doubtful that you're going to understand what I'm saying.
The APD stat is skewed to the left. This suggests its distribution is not normal. I'm of the mind that an APD above 0.1 to 0.2 is actually quite good. I'll expand on this. During the first six months, outlying APD's can be seen, these are ones above 0.4. Twelve months out, we'll see these decline to the middle of the skew to 0.2. When you go more than 1.5 years out there's strong evidence that you can see from the grid that most long term system converge on the 0.1 to 0.2 range. Over time these systems will stay there, so 0.1 to 0.2 appears to be decent long term sustainable results. Outlying systems above 0.3 will not be sustainable, and, since we know the APD stat distribution, we could go as far as saying that it is more likely that a high APD stat system is one to be avoided rather than subscribed to, and that should bake anyone's noodle to think about because this is actually "performance chasing."
You probably won't know you're chasing performance until after you subscribe, and, hence, why we get the "it stopped working" review so frequently on the site. This kind of review comes often enough that makes me know that the subscribers chase performance on regularly, because they haven't quite figured out how to analyze system's correctly, yet. I hope this post will help them out as they look at other systems.
Having no research based history behind this stat, the only conclusions that can be drawn about a portfolio manager's APD stat can be taken from collective2.com.
Any system with a higher APD stat, I believe, is just an outlier, and that at the core of all trading is "hold and hope." Were we to analyze Buffet I can tell you his APD would be around this average rate. He just took several years for his bets to play out.
Anyway, for me it's just the APD stat is skewed to the left, and that says enough about its distribution that I know that where the highest percentage of system lie, which is less than 0.2 is where nearly all of them are. So to sit there and say one system is better than others ignores length of time on c2. Over time, I believe all systems will converge to the middle and to the left of center on the bell curve somewhere betwee 0.1 and 0.2 ideally for profitable systems. You can watch either my own system or Mike's system if you want to watch our APD stats converge on this range, because that's where all "profitable" systems will be over long periods of time.
I think with as much experience watching the effects of trades on any system's APD stats suggests that no one will have high APD's if they stay long enough. The ones that do, we call outliers, but this says nothing about predicting future performance, and so APD is not actually an ideal indicator after all. Not that it doesn't tell us anything, but that to state 0.4 for a six month system means it's better than a 3 year old system with a 0.2 APD is not actually comparable. Only DD's and APR's can be compared across different time frames. If it were normally distributed we'd have more to say about its validity. Since we don't, all we can see is what happened in the past.
WL has the APD indicator now thanks to me coding it, but it will only work for stock based or ETF based systems. I've found very little evidence that the best systems on wl4.wealth-lab.com will have high APD's, and I found the negative skew just as we find it on collective2. This stat mainly is a statement of past profitability rather than a predictor of risk adjusted returns. For that, you need to go back, study finance, and use the Sharpe Ratio and some profit factor based filter.
