"“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.”"
I agree. For goodness sake let’s not aim too high and ruin all the hard work to date, regardless of one’s views on the validity of APD as a system statistic until someone can show research that categorically can prove or disprove it’s worth we should stick to the far easier task of making the description as meaningful as possible so as to enlighten those otherwise unaware as to it’s limitations.
We had managed to have a reasonably civil discussion to arrive at the proposed new description, for those that missed it it’s the 3rd message in this thread. I understand people feel strongly enough to want APD removed entirely but I ask you to start with the assumption that APD won’t be removed and realise the next best course of action is to support the proposed new wording. It’s pretty simple, assuming APD is here to stay what wording would you prefer to accompany it:-
1. The existing wording
2. The new wording (third message in this thread)
"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. '"
In the proposed new description we have removed the references to risk as explained in the opening posts of this thread.
Yes, upon review I agree with the new description referenced above.
You got my vote.
I see here a lot of negative point of view about hold and hope systems.what do you mean about hold an hope ?
From my point of view hold and hope system could be a long term trend folowing system.
all the well known investment names over the last 30 years are all hold and hope systems.
Do you say you"ross" that they are all gabage systems?
APD is meaningless. (except for scalpers who are the less wellquoted in the real investment world)
Beau, I’m going to try one last time using your SuperBands system. I see that you have set the strategy to trade a number of different stocks on the same day, if the entry criteria is met. That is a form of money management. You are doing this to average out results over a number of trades (diversification) instead of just putting all your money on the first trade and maybe having a large loss. That’s a form of hedging in the same manner as going long a stock and short a different stock on the same day: your hedging against a large loss by trying to get the average return for the day.
What happens when you trade a number of stocks rather than one is that the average daily change in equity from day to day will be less than if you just trade one. That leads to a lower standard deviation and a higher Sharpe, all due to money management. In essence you’ve canceled out some of the daily drawdown of a single large loss with lower drawdown loses or outright wins. The Sharpe reflects that because it just looks at the equity changes, not the individual trades.
Profit factor has the same limitation as APD. It is not computed off the equity curve, but off the individual trades. If you look at systems with high APD, they should have high profit factor as well. Systems with low APD should have low profit factor. But systems with a high Sharpe, won’t necessarilly have a high APD or profit factor because good money management might be a large contribution to the high Sharpe.
The only systems that won’t be discriminated against by profit factor and APD are those that only trade one instrument at a time (like your pairs scalper). Those that trade more than one instrument at a time are using some form of money managment (usually diversification) and are being discriminated against, including your SuperBands system.
If APD were computed off the daily equity bar (profit for the day divided by the largest drawdown for the day), instead of the individual trades, I wouldn’t have a problem with it. In that case, the effects of money management are factored in and all systems (except C2 fantasy systems) can be equally compared. In essence it then becomes the same sort of metric as Sharpe. The only problem with computing APD on daily equity bars is that C2 can’t compute an accurate one because of limitations in time sampling.
Hold and hope is allowing open losses to run in the hope they recover to achieve relatively small profits, though not exclusive to them it is typically practiced by high win % systems and the APD does a good job of highlighting them.
Long term trend following systems do not hold and hope, as they cut losses short, let winners run and typically have low win %, but it is true the current description of APD would make you believe they do, and that is one of the concerns we have addressed with the proposed new description where we state "Other possible causes of low APD are the system applies a hedging strategy, or the system has many small losing trades and less frequent but large winning trades. These strategies do not necessarily imply a high risk."
"If APD were computed off the daily equity bar (profit for the day divided by the largest drawdown for the day), instead of the individual trades, I wouldn’t have a problem with it. … In essence it then becomes the same sort of metric as Sharpe."
Yes, that’s true. In an earlier thread (about a year ago) I did some simulations that indicate that equity-curve-APD and Sharpe have a monotone relationship when the returns have a normal distribution.
I am not sure that “money management” is the best word to describe the difference. Both APD and Sharpe are independent of leverage. I would rather say that timing and diversification are the difference, because Sharpe is sensitive to that and APD not.
Even though APD is only appropriate for a certain class of systems (single instrument), and although it might have some limitations even within that class, I prefer to keep it. You can still draw conclusions from it in some cases. For example,
- If APD is high then h&h is probably not often used by the system, timing is probably good, but there might be a lack of diversification
- If APD is low while Sharpe is high, then diversification may be an important part of the system and picking trades or changing the relative position sizes of trades might seriously detoriate the performance. However, the diversification can also mask a h&h strategy, and the trades should be investigated in detail to sort this out.
Therefore, I prefer to keep APD but change the mouseover text.
Anyhow, given that APD is here to stay, would you not agree that the text of Jon and me is better than the current mouseover text, and also better than the text of Ross? That’s what this thread is about.
Examples of data that I would consider good evidence in favor of APD would be this:
It’s possible, but it’s very time consuming.
How about clear “Buy&Hope” system stats? The stats below is for system that buys, holds position for 60 days and sell it. No stops. 100% of TC per trade. No averaging down (unfortunately).
2004
apd - 15.6194
Sharpe 3.47
Annualized Gain 185.88%
Net Profit 185.88%
Max DD 13.45%
Avg Profit/Loss % 31.06%
2005
apd - 1.9389
Sharpe 2.62
Annualized Gain 154.24%
Net Profit 543.09%
Max DD 25.11%
Avg Profit/Loss % 27.32%
2006
apd - 0.0702
Sharpe 2.05
Annualized Gain 95.81%
Net Profit 646.60%
Max DD 35.24%
Avg Profit/Loss % 19.88%
2007
apd - 0.0336
Sharpe 2.17
Annualized Gain 107.47%
Net Profit 1748.95%
Max DD 35.24%
Avg Profit/Loss % 21.62%
2008
apd - 0.1902
Sharpe 1.38
Annualized Gain 61.03%
Net Profit 900.05%
Max DD 54.02%
Avg Profit/Loss % 15.37%
—
Eu
Am I correct that the results of one year include the data of previous years? Because 2006 and 2007 have the same max DD.
In this case I would say that, albeit after two years, the APD correctly identifies it as a buy and hold system. Do you have a different conclusion?
Am I correct that the results of one year include the data of previous years?
Correct. The statistics is cumulative.
Because 2006 and 2007 have the same max DD.
Right. There wasn’t new Max DD in 2007.
In this case I would say that, albeit after two years, the APD correctly identifies it as a buy and hold system. Do you have a different conclusion?
And what is practical usage of the identification as “risk” indicator?
Sharp says the same story. Moreover Sharp continues degrade in 2008 when APD becomes optimistic again.
Highest “risk” APD shows for second best year (2007) of the system when Net Profit reached 1748.95%.
Accordingly APD we don’t trade 2004, because we need historical data. 2005 because APD starts degradation and we have to start looking on the system in 2008, because APD starts rising. In other words we shouldn’t trade it at all. It’s purpose of APD to convince somebody to not trade at all 
At the same time Sharps shows different picture.
So I’d like to repeat my question. What is practical usage of the identification as “risk” indicator? Sharp recognizes “Buy&Hope” systems as well and imho much better. Why do we need meaningless APD that has no studies/no statistical proof of usability/nothing/nada/niht/null.
Eu
Jules,
Hold and Hope systems are not a problem at C2. The problem is systems that add-on to losers. People have equated the two and they are not the same thing.
Since APD was claimed to be the sure fire way to find systems that add-on to losers, I’ll repeat my assertion that every one of Craig Benders systems had a high APD until it crashed. APD doesn’t sniff out those type of systems until after they crash. While just one glance at the trades the way Matthew has broken them out, clearly shows add-on-to-loser systems.
Why have a statistic that, “is only appropriate for a certain class of instruments (single instrument), and it might have some limitations even within that class” (your words), when the Sharpe ratio allows you to compare all classes of trading solutions.
I still have hopes we can get rid of the APD statistic and vote along with the other 5 or 6 others who said get rid of it.
I think I agree with you, but as it is, we know it’s a timing statistic with at least some marginal importance.
"I still have hopes we can get rid of the APD statistic and vote along with the other 5 or 6 others who said get rid of it."
This is like 2 days before the election between Obama and McCain saying you still want Hillary to win.
Getting rid of APD isn’t on the ballot, but a choice between a new wording or retaining the existing one is. What’s your vote?
"So I’d like to repeat my question. What is practical usage of the identification as “risk” indicator?"
There isn’t, and that’s why if you re-read the first 3 posts of this thread and the new proposed wording you will see we have removed all reference to it being a ‘risk’ indicator, therefore I’m going to assume this meets with your approval compared to the existing wording which says it measures risk.
Jon,
Everything you’re doing is informal. You decided on your own to reword the description and made that your “election”. Eu proposed voting on whether to keep APD or not. Neither one matters unless Matthew decides to do something. Since Matthew hasn’t said anything either way, I’m still hoping to get rid of APD. I’m voting on Eu’s ballot, not yours.
The purpose of APD is that it supposedly can hightlight a dangerous strategy in an earlier stage than Sharpe. Suppose that a systems has the habit of letting its losses run, hoping that they will recover. Suppose the system is also diversified. Most of the time the system will get away with this for a long time. At the moment that one trade is in a drawdown, others will be recovering. So the Sharpe can be high for a long time, because the strategy is masked by the diversification. Nevetheless there is a risk that suddenly all positions go down simultaneously, like in the current market. In that case the strategy of letting losses run is extremely risky. APD will highlight this risk in an earlier stage than Sharpe, namely at a moment when the risk is still being hidden and when Sharpe is still high. A low APD is no proof that the system is actually applying this strategy (herein I differ with Ross), nor a proof that such a crash will ever happen, but it is a reason to investigate the possibility.
In other words, I consider APD as a kind of worst-case-scenario Sharpe ratio. In may be too pessimistic in some cases, but then again, Sharpe ratio may be too optimistic in other cases.
For example, positive forex had once a very high Sharpe. The t-value (which can be viewed as an age-corrected Sharpe) was one of the two highest at C2. Even Ross concluded that the system had pretty much proven itself, despite the small APD. However, two weeks later the system collapsed totally. So in this case APD correctly indicated that there was a hidden risk before the collapse happened. Sharpe ratio indicated this only after the collapse happened.
This is only one example and therefore I do not consider this proof for anything except for the proposition that APD might work better than Sharpe in some cases. IMHO, someone who wants to remove APD should prove that it almost never works in any way and someone who wants to keep it without clear warnings should prove that it almost always works in at least one way. In both cases a large body of evidence of a large number of systems is needed. We lack both forms of evidence, and therefore I think we should keep it but warn people against limitations that are theoretically conceivable. This is what Jon and I propose.
"The AMD Ratio (“Average Moon phases to Drawdown”) measures how much Moon phases are correlated for each dollar of drawdown suffered while a trade is open.
A high AMD means that on average, the system makes trades with high correlation to Moon phases in comparison to drawdowns.
A low AMD can result from not following major Universe laws. This is considered a high risk strategy.
Other possible causes of low AMD are the system applies a Solar phases. These strategies do not necessarily imply a high risk.
In summary: a high AMD is good but a low AMD is not always bad. "
I’d like to replace APD to AMD. It’s much better indicator and it shows everything. My proof is my word and anybody who doesn’t agree is an idiot and cannot see obvious things.
If APD doesn’t have any practical meaning why we won’t introduce my AMD? It has better value than APD and in my opinion “trading God” opinion it will benefit everybody.
Eu
Actually that’s not true, but i’m not going to go back over what now must be well over 100 posts on this subject through various threads in the last few days, Matthew has been consulted on the issue and we set about developing a new wording in an open public forum rather than trying to hammer out a deal behind closed doors as we wanted to make it as democratic as possible in order to get the support of the majority of people so that it would stand more chance of being adopted. By all means try to pursue your agenda but I know others have been down the road of trying to remove it and it’s a dead end, as I stated previously I think you need to start from the point of respecting MK’s decision to publish it in the first place and assume it’s here to stay, if you move to that point then wouldn’t it make more sense to try and have it worded in terms that more accurately reflect what it does and how it should be interpreted? Or would you rather it stay as it is?
Need something about its use as a timing statistic.
Keith,
It is true that it has not been shown at C2 that hope and hold systems are a problem. That is another one of Ross’ unproven mantra’s. Nonetheless, I give him the credit that the idea is plausible at least. Even if that claim would be proven to be false, I can see some usage for APD. The fact is that ‘hope and hold’ versus ‘cut your losses’ are very distinct and opposed trading strategies. Regardless of the beliefs that people held about them, subscribers may be interested in sorting systems on this aspect.
Your example of Craig’s systems is not convincing for me. All his systems died young. All statistics are extremely unreliable on such short time frames, and this is no different for APD.
"Why have a statistic that … when the Sharpe ratio allows you to compare all classes of trading solutions."
As I tried to explain in my reply to Eu, APD is sensitive to signals of hidden risk that the Sharpe ratio ignores. I tend to consider it as a kind of worst-case-scenario index. It will sometimes be too pessimistic, but that’s the whole point of worst-case scenario’s. It is important though that people are aware that something is a worst-case scenario and not necessarily a reality (yet). That’s why Jon and I want to change the mouseover text that C2 displays at the green questionmark.
Compare this: A low APD is like a warning sign for a dangerous crosspoint, based on the physical properties of that crosspoint. It does not say that you will be hit by a truck on the instant you pass. It merely says that it may be difficult to see the truck coming if one is coming. A low Sharpe on the other hand, tells you that you were hit by a truck. That is more objective, but also too late.
