Or so you say, because there is, and for me, that’s contained in the distribution of the APD stats, which you seem to ignore and I think don’t comprehend what that means.
I don’t care about hedging. If you want to perfectly offset your trades, don’t hedge by not going long or short anything. Time it better. If you want to just hedge one of your positions, take a smaller position. Hedging isn’t realistic as an excuse for a low APD for me.
Beau:
I would not dismiss APD outright either. However, the data that you discussed so far don’t tell me anything. The mere fact that APD has some statistical distribution does not imply that it is a valid measure. Flawed statistics have distributions too. Also, the distribution itself, being no more than a simple fact, will never be enough to establish a normative bound. For this you will also need to evaluate the sampled systems by some criterion other than APD itself.
Examples of data that I would consider good evidence in favor of APD would be this:
1. Show that the percentage of systems that crash at time t + 1 is a monotone increasing function of their APD at time t.
2. Show that, across a large sample of systems, the return rate of systems between time t and t + 1 is positively correlated with their APD at time t, and that this correlation is significantly higher than the correlation with the Sharpe ratio at time t.
Others:
I see no reason to involve Palsun in this discussion, something which was started by Ross and not Futurm. The implicit reasoning applied here is some kind of ‘guilt by association’:
Palsun was a fool.
Palsun rejected APD.
Therefore all who reject APD are fools too.
Therefore APD must be good.
That is again a fallacy.
"Now, where is your evidence that there is no evidence? I don’t see your tests anywhere, just your opinion… "
No, you have the burden of proof. You claimed that APD is good and that there is overwhelming evidence for that, but you have NEVER shown any statistical test whatsoever, just your opinion.
"It does a very good job flagging systems that tend to run up large drawdowns compared to net profits"
Yes it does that, but it also appears to flag some systems that shouldn’t be flagged.
This is what I dont understand guys…
If APD signifies that a system allows losses to run, then how can a system have both a very low drawdown AND a low APD?
Pegasus
That can happen if the average profit is small. It can also happen if the drawdown of one trade is compensated by the profit of another, simultaneously open trade. In that case you won’t see the drawdown of the first trade in the equity curve, but it will affect the APD.
Beau,
Do you believe there is such a thing as money management? If you do, do you believe that money managment is intended to smooth out your equity curve and reduce your risk exposure? If you believe both of those then you have to believe that the Sharpe ratio captures the effects of money management because it is computed from the equity curve, not the trades.
But APD is not computed from the equity curve. It’s computed from the raw trades. It doesn’t capture the benefits of good money managment, or point out bad money management.
It doesn’t matter whether your money management strategy involves hedging, or limiting exposure in a group, or risking a small fixed percentage of equity on each trade, the intent is the same: a smoother equity curve. A worthwhile stat ought to recognize worthwhile money management.
Palsun is involved here, because FuturM decided to quote Palsun as a shining example of someone who had an early "assessment" on APD.
You need to take that up with FuturM
Who wrote the words is irrelevant. Only the words have meaning. You choose to discredit words based on who wrote them rather than on their substance.
No Ross, you are the one who started to talk about Palsun in this discussion, when three days ago you decided to write
"Palsun (from long ago) made it his agenda to rid C2 of APD - a person who often said a 95% drawdown in a trade did not matter unless it was "realized"."
I can only interprete that as an attempt to discredit our position by a fallacy.
That had nothing to do with why FuturM said this, but no matter, I am increasing my ignore list as we speak (FuturM, Jules)
I see rants and I see theory.
It is astounding to me that someone who has exhibited knowledge like yourself are so unaware of what happens to the max DDs of ANY system that averages down or holds & hope. LOOK AT THE MAX DDs FOR THESE TRADES. Is this hard for you to see what is happening? You need proof for what you already know and is screamingly obvious?
Do I have to prove to you that 3-3 is 0 ??
You need to stop putting on an air of superiority and acting clueless at the same time… If this is not screamingly obvious, then there is little hope for you. But this will be my last word. You can continue to pontificate, I really no longer care.
Keith, give me some credit for knowing what the Sharpe ratio does. The hedging you’re describing is not hedging, it’s reducing the position size.
Side note:
This is not directed at you, because I know you hedge your positions.
Only on this site do we get people who think that by putting on a short position while long in the exact same instrument are they “hedged.” That’s not hedging, and that’s what I thought you were referring to, because I’ve seen threads asking this same retarded question. You’re netting out your position or selling it. This is identical to flat and you’ll lose money on the commission. Good for brokers, bad for the trader.
I see the APD as a timing statistic, but it still captures that excess return above the drawdown.
As I described side by side definition of sharpe and definition of APD, we get:
Sharpe:= excess percentage return per unit of risk
I guess you want to make sure I know the definition:
Sharpe= E®-RfR/Std Dev of Returns
APD:=excess dollar return per dollar of drawdown.
=NetProfit/TotalDrawdown=(Gross Profit-Gross Loss)/(Total Number of trades)/(TotalDrawdown summed per trade)=(NetProfit/NumberofTrades)/(TotalDrawdown of each trade summed/number of trades) All are identically equivalent. (And the purpose of suggest backing out the Gross Loss was to have a recovery measure of the system. Big Profits can potentially make up for the large drawdowns, but, that is to be used in conjunction with the NetProfit used in the APD ratio. )
Not too high a difference in my opinion between the two. If it says anything to me it tells how good the timing is. Since no one has perfect timing, this is why I’m saying the APD will never be high long term.
It looks as if something over 0.2 to 0.3 is a decent APD. Using the Standard deviation as with the sharpe does account for the volatility “in both directions,” which I see as your logic for saying it doesn’t capture money mangement, when it does.
The average money manager, not here anyway, figures, well I’m looking 5 years out, or 10, whatever, it’ll probably be up by then. That average money manager might have a decent APD if he gets his timing correct. If he bought 2 years ago, chances are he had a lot of drawdown on all his positions.
Timing is a lot of what makes the systems work, so while I think there is room for interpretation into the statistic, it might be that we as vendors don’t feel like the rigid definition is applicable to every single case.
That’s not a bad thing, because the subscriber base probably can figure out what the insight is without having to ask the vendor. The vendor shouldn’t cuss or explain the uselessness of the APD straight off, because it does tell you how good the vendor’s timing is, which is a stat of vital importance to a prospective subscriber imho. Thus, why at the heart of the matter you shouldn’t discredit this statistic.
Trust me, I’m more concerned about doing more research into what a decent APD is rather than guess that 0.4 or 1 is good, which they are, because those aren’t long term stable numbers in any system I’ve backtested after implementing the ratio into my backtests.
There’s 1000’s more to test, and I seriously doubt I can find 1 with a 5 year APD above 1. I am positive it’s not to be found, and there isn’t a more diverse range of systems to test from than from wl4.wealth-lab.com.
If you’re not into its rigid definition, change your thinking of it as a timing statistic, rather than a risk statistic, though these are probably both the same thing.
"It is astounding to me that someone who has exhibited knowledge like yourself are so unaware of what happens to the max DDs of ANY system that averages down or holds & hope."
Here you use another fallacy: Making a nonsens allegations in the hope that something will stick in people’s memory. I never said that averaging down or hold & hope is good. I said your APD does not accurately measure it.
"You need proof for what you already know and is screamingly obvious?"
Your claims about the interpretation of a low APD are not obvious and yes I need proof for it. Even more so since you require proof of Jon and me, while you have NEVER given any proof for your own claims. Repeating the same unproven claim again and again is a mantra, not proof.
"You need to stop putting on an air of superiority and acting clueless at the same time… "
I’m sure Palsun agrees with you. See? I can make fallacies too.
The bottom line, if the APD flags hold and hope systems but at the same time is flagging other types of non hold and hope systems, then the APD is meaningless!
There is no evidence that I am aware of that tells us this is not happening. And until such time as this evidence is produced the APD should be removed. To not remove it is to take value away from productive venders.
Since the screaming obviousness of APD is lost to a few, I ran a simple test. I sorted on "The Grid" by APD. I only used systems with 3 months+. I also excluded systems over 1.0 or negative APDs to focus on the core.
Since there were only 7 in the first third, I also only did the first 7 in the other thirds.
APD 0.67 to 1
2.7
0.6
2
6.8
3.3
2.3
2.5
5.1
Average PF: 3.1625
Median PF: 2.6
APD 0.33 to 0.66
2.9
1.6
1.4
1.1
1.9
1.8
1.9
2.7
Average PF: 1.9125
Median PF: 1.85
APD 0 to 0.32
1
1.4
1.6
1.4
1.4
1.7
2.2
1.5
Average PF: 1.525
Median PF: 1.45
The direct relationship between APD and Profit Factor is quite obvious.
Of course, I expect someone to redo this and cherry pick an unflattering version of this. I did this honestly, but some just like to rant. But now I think I have them on ignore, so it means little to me if they continue ranting.
But when APD is high, you can be quite sure that it is not a h&h system. Therefore I would say it has meaning in some cases, and it should stay untill there is a better alternative. The only reason why I sometimes doubt even that, is that its inventor persists in abusing the statistic by claiming that a low APD is proof of h&h. If he continues to do so I might vote for removal. But this is fundamentally a problem of a different order, one that is related to the person rather than the statistic itself. I mean, suppose that Sharpe had abused the Sharpe ratio in a similar way (e.g. by claiming that a low Sharpe indicates high leverage) then that would a problem with the person Sharpe and not with the statistic Sharpe. Similarlly, here, the problem is not APD but the interpretation that is given to it. I guess we cannot change the mind of Ross (travelling back in time would be easier) but I hope we can change the explanatory text of "the trading system authority" C2.
Since you said you ignore me, I will ignore you too.
APD’s mouseover description is completely misleading. That said, I do like the statistic.
It is NOT a risk measurement tool. It IS a tool that generally identifies straightforward systems that exhibit good timing on entries and exits. By straightforward I mean traditional, naked-position systems not utilizing hedging of any type.
Don’t argue with Ross. Again, it’s pointless.
The Palsun quote WAS pretty funny though =) When I read it, I was thinking “what a lode of crap that paragraph was – it didn’t say a damn thing other than ‘water is wet’. Is this a Palsun quote???”
I would have to vote for icecream too.
APD is not a risk measure at all. The description, if it’s going to be changed, needs to be changed completely – removing the whole “risk” idea. '
It’s a timing measure, and also not useful for systems using true hedging.
It definitely does a good job of evaluating systems that trade a single (or a few highly correlated) symbols in the same direction. That was what prompted it’s creation, and why it was added to C2 – to identify the ‘Hold & Hope’ systems that were prevalent in the rankings a few years ago.
