Sigh. All to which this can lead is that APD stays here AND that the current formulation is not changed because we can’t agree upon an alternative formulation. That is the worst possible outcome in the eyes of everybody except Ross.
Is it so hard to see for you that Jon’s formulation is better than the current one, especially when you think APD is meaningless? I can understand that Ross objects to the text, but if the anti-APD crowd objects to it too, then I’m lost. Like I said, after changing the text we can still discuss removal of APD, but if you vote against Jon’s text because you want to remove APD then the most probable outcome will be that nothing changes at all. Is that what you want?
There is no need to establish any bounds. The user can decide for himself whether a $1000 dd for $100 profit is something he likes or not.
There is also no need to prove a relationship with known performance metrics. We carefully deleted the suggestion that APD is a performance measure. Since we don’t claim it, we don’t have to prove it.
The main message of Jon’s text is that APD can be used to filter out certain trading habits. The text does not judge the risk of these habits. There is only one exception to this, namely it says that letting losses to run is considered a dangerous strategy. Do you really disagree with that?! Even so, the formulation “is considered” adequately describes the dominant opinion at the C2 forum. In the past three years I have seen only person defending the opposite position, namely Palsun.
OK, let me see if I can disect what you’re saying, as there’s an awful lot of definitives in there, you’re saying 'A higher apd WILL indicate superior timing… will it, or could it
I guarantee it will. By the fact that it will single out the holy grail, if it ever arrives, I think that’s a will. “Could it” is guaranteed. Jon, are you not thinking about what the APD is? Gross Profit, Gross Loss, and the variable total drawdown. Of course all of those variables will be higher, since you’ve got the two components of profit factor, and the sharpe contained in the APD stat.
And if a higher one indicates that then does a lower one indicate poor timing? Not necessarily, but people might conclude that if you definitively say high apd = good timing., just as the existing description currently implies bad apd = bad system, period. You might want to avoid that kind of rigid statement.
Jon, this gets to the context of what a high APD is, and on that I’ve found reasonable evidence to indicate it’s something above 0.2, which we appear to already be at. Most profitable systems exceed point two, so yeah, I’m saying below that I’m finding that those systems are probably trash. It’s a an unforgiving statistic that’ll follow you around, for sure. And I think that’s why people don’t like it. But the APD wil for sure indicate to you superior timing, which is at the heart of trading system development. Those without 0.2 probably are making too little profit with too much risk, and I agree that that is what the APD stat says. I know it hasn’t been seen that way up until this point, but it really took a lot of thinking to rationalize it outside of the realm of being a measure of risk. It is a measure of excess return, sure, but, more importantly, an indicator of good timing. I had an astronomically high APD of 1.2 6 months in. The PTQQS was definitely on its way. However, it was not long term sustainable, thus supporting what I have said that you might be looking back rather than forward. After that I decided I’d better wait to see what happens at the 9 month mark for most systems. Obviously that’s where the younger warning comes in. They haven’t been able to see how bad the losses can be.
‘over time a higher apd WILL exhibit both correspondingly higher Sharpe and PF’… will it? or could it?
and higher than what?
It’s an understood higher than what. Higher than the next comparable system with a lower APD, ceteris paribus. That’s a will, as Ross showed with his cursory check of the APD distribution of Profit Factors and Sharpe Ratios.
You’re basically saying if one is high all 3 WILL be high, but if that’s the case it would already show that in those statistics so there wouldn’t be any need to state it, would there? I’m not sure referencing other statistics is useful, I think it’s better to interpret the APD figure in itself without reference to other stats.
Personally I’m still not convinced a high APD indicates superior timing, all it indicates to me is that the profit is high relative to the drawdown, that is afterall exactly what’s being calculated, and what’s stated in the proposed new definition.
What don’t you get of its use as a timing statistic? Think of the holy grail. The absolute perfect timing system. It will have such low drawdown relative to its profits that the APD will be astronomical. It is that extreme case as in the case where there is no to little total drawdown and no to little gross loss that we see that APD is in fact, the statistic that will absolutely point out the holy grail to us in the sense that the system displays perfect timing. It’s all in the components, and, being the only one that’s ever described it this way until now, you can see as Mr. Lurker agreed with me that it is such a statistic. If you have large drawdowns for relatively small profits (ie: what Ross says the APD stat is), then the system’s timing is off. There’s not really any other way to put it. In fact, I’m so convinced after the weeks I’ve been studying it, and how it backtests on some of the older WL systems, that I am 100% certain that this statistic is more useful as a timing stat. If you want to equate that to risk, I don’t think it’s a stretch to go that route.
Why haven’t we seen it (ie:the holy grail)? No one has the holy grail yet, and this is why: In the extreme case of perfect timing, we will see extremely high APD’s, thus supporting its theoretical use as the Holy Grail Indicator.
Don’t you see that? The components of the statistic will indicate good timing, absolutely. If there’s all gross profit, little to no gross loss or total drawdown (wouldn’t that be nice), this APD stat will indicate it, and be extremely high, and so will the Sharpe Ratio, and, of course, our profit factors.
Writing this I see a nobel prize in the making for the statistic. The work supporting it can probably be garnered from the statistics of this website.
So, we’ve finally figured out what the APD stat is. It’s the holy grail indicator, but a good one is one above 0.2.
One of the major reasons I saw value in APD, is that assuming you have two systems that are stabilized, a system with APD of 0.6 theoretically allows three times as much leverage on average as an APD of 0.2.
In a sense, there seems some kind of relationship with Risk of Ruin, with APD.
The relationship Beau seems to be talking about is something I already felt. Many times in the past, I have improved a system’s APD, simply by testing different sized stop losses and different sized profit targets.
I found that some people who complain about their low APDs did not really grasp proper money management. The habit of scaling in and out of trades (making it difficult to figure out a system’s per-trade behavior), averaging down, scalping small wins, holding &hoping only obscure system behavior from subs who are trying to understand a system’s risk characteristics.
I find many of the arguments against APD as juvenile and unfounded. As if it needs to “explain” itself regarding different ways it can be used (I can “only use it for H&H or averaging down and not scalping because yadayada”), and it is useless unless I can prove something that is thoroughly obvious to anyone who spends 5 minutes observing or thinking about it.
Again, where are all these unfairly treated systems? I see almost nothing but rants and hypotheses.
The main message of Jon’s text is that APD can be used to filter out certain trading habits. The text does not judge the risk of these habits. There is only one exception to this, namely it says that letting losses to run is considered a dangerous strategy. Do you really disagree with that?! Even so, the formulation “is considered” adequately describes the dominant opinion at the C2 forum. In the past three years I have seen only person defending the opposite position, namely Palsun.
I’ll talk to you about this, Jules, and no, I don’t have you on ignore.
Of course letting losses to run is bad, if you’re debating that (not to you Jules) I think you belong in a looney bin. I don’t really think that’s the problem we have with the statement.
The statement “letting losses run” absolutely depends on the time frame involved. We need to “normalize” the time frames to arrive at a better comparison. Letting losses run on a scalping system, is not the same as letting losses run on a b&h system. They’re different time frames with different profiles, which I believe makes them uncomparable from system to system. So, on that, I think showing APD stats for short, medium, and long term systems for comparison in addition to the overall stats will give use a better context for our comparisons. Ross doesn’t appear to agree with this, but this gets back to the timing aspect of it.
APD tells you if you let losses run. It’s true, but it gets to “why did you let losses run?” Why is because your system has poor timing. That’s how I’ve managed to rationalize what he was saying. But the “letting losses run” part, I think is what you guys are referring to as the “misleading part” outside of the absolute definition of a poor system. Letting losses run relative to what? The size of your system’s profits. To me, that says more about the timing of the system, then the absolute value of the statistic.
I agree Jules we can’t just say below 0.2 shouldn’t be traded, but I’m finding that 0.2 really sucks. The systems are terrible below that, but the only outliers we have are with systems less than three months old, so I don’t consider that any indicator. It’s a better route to take to just say 0.2 is good, because there are the top 15% of systems right there.
(I don’t know if Matt ever saw the periodic APD stat post, but maybe I’ll just rehash that we should measure cumulatively and on a quarterly basis as part of the stats in a chart the APD of that period. Side note though.)
So it is all about “letting losses run”? Vendors denying the applicability of its stat to their system is kind of entertaining to me now, but it’s all about the timing.
If you knew for certain that gold was going to go up $43 tomorrow, and bought with a market order, how would you think you did if gold went down $150 before that happened? (Obviously an extreme case and you’ll probably never have info about a futures contract like that). Poor timing, and your apd stat would suffer, thus an example supporting its use as a timing stat.
I hope you don’t feel that way about my own analysis of it, Ross. The difference is in the context of “letting losses run” or holding and hoping. H&H is not the same on a short term system, as on a long term system. In fact the APD stats are going to vary by quite a bit, as I’ve seen in my backtests.
The Ulcer Performance Index (UPI) is somewhat similar to the Sortino index in the sense that it measures risk not by standard deviation but by a form of downside deviations. It is different from the Sortino ratio in that it is also sensitive to the duration of the drawdown. The funny thing is that, if I remember correctly, I had a subroutine for it in an early version of the advanced statistics module, but I removed it because there were already so many statistics in it and I did not want to defend why this one would be needed too.
The important point for the present discussion is that UPI is based on the equity curve only. I consider that an entirely different class than APD and PF, which are based on the trades. Indices that are based on the equity curve will not be sensitive to intratrade drawdowns that are compensated by other trades at the same moment. APD on the other hand, will be sensitive to such drawdowns. I consider that both an advantage and a disadvantage of APD. It is a disadvantage if you want to use APD as a performance metric while you are only interested in the equity curve, but it is an advantage if you want to use it as a kind of worst-case scenario that makes you aware of hidden risk.
For this reason I do not consider UPI as a good alternative for APD. But it might be a good idea to combine it with an earlier suggestion of me, namely to layout the trades in a kind of hypothetical equity curve. I could attempt to work that out, but frankly, given the fact that after 121 posts we are still unable to reach consensus about something so trivial as a mouseover text, I don’t see the point of such an excercise.
Long, long ago at the beginning of this conversation I stated that "“The description for APD currently says “Low APD means a system allows open losses to run in order to achieve relatively small profits (bad)” That is pretty clear, it’s not saying it COULD mean it does that, or that’s ONE of the interpretations, it’s basically saying if the APD is low then this system allows open losses to run. Period. No if’s and but’s, that’s it, no other way to look at it, bad APD = bad system. To me that is inaccurate and misleading and should be re-worded.”"
I thought we pretty much accepted that to be true but now you are telling me you can ‘guarantee’ everything in your statement should be phrased ‘will’ not ‘could’ etc…
You are effectively asking us to remove one definitive statement and replace it with at least two new ones. We’ve already established that there are perfectly good reasons why a system may have an APD of 0.20 or less and NOT be ‘trash’ as you call it, that is the whole point of this exercise, are we really back to square one?
Beau,
Are you aware that my post to which you replied was a reply to Eu?
I can see that APD has a relationship with timing. Someone who can perfectly catch the bottom will have no drawdowns and his APD will go through the roof. But is a low APD always the result of bad timing? I am not so sure that.
Remember that, as far as I am concerned, we are talking about the mouseover text. This text should be helpful, neutral, uncontroversial, and without unproven claims. This is what Jon and I tried to accomplish.
Now, if we would add your text about timing, we re-introduce the idea that APD is a performance measure, which is controversial to say the least. That is, the text would perhaps be more helpful, but it would certainly be more controversial. Therefore I think we should leave that part out untill there is more convincing evidence that APD is a proper measure of timing-adequacy.
Also, I think this process will never converge if we keep on adding new phrases of text. If we change the proposed text, we will have to vote again. It took us 120+ posts to get a few people so far that they voted in favor of Jon’s text. Do you really want us to do this all over?
The question is really very simple. Jon and I proposed a new mouseover text, and asked to vote about it. If you think the new text is better, say “yes”. If you think the old text is better, say “no”. Of course everyone may vote about anything else he wants, but obviously if we all make our own private ballot and disagree with all others, then nothing will change at all.
Well, I did say it is “better”, so it would be a yes.
There’s no doubt in my mind it is a timing statistic.
Thanks!
No one posted today, so I guess we may start counting votes?
I haven’t posted in these forums for a long time mainly because I got tired of all the personal attacks, mudslinging and the domination of a handful personalities, but I want to make a few observations.
There is absolutely nothing democratic about this forum. Don’t kid yourself that just because a handful people is discussing it here on a public forum make it democratic. There are probably 1000’s of C2 subscribers and only a very, very small fraction is posting on here and to think that this handul people represent the masses is extremely arrogant. Vast majority of people is not comfortable posting in public forums and simply not interested in getting involved in the mud slinging going on here. I know I am not and no one should be forced to make their vote public before they are heard.
If you really want to make this democratic, Matthew needs to create a poll with at least three options. These could be:
1. Leave APD as is
2. Change the description
3. Remove APD completely
People should be allowed to vote anonymously and obviously each registered user can only vote once.
For what it is worth, my vote would be for 3. If one have to add so many disclaimers and exceptions to a statistic to point out cases where it might or might not be valid, then it is useles. If a car has a broken engine, painting it a different color might make it look nice and shiny, but the car is still broken.
Instead of having a statistic to receive guidance, the viewer now have to try and figure out if the statistic is valid in a specific case or not. If someone is sophisticated enough to figure out if the statistic is valid in his specific case, then he do not need the statistic to begin with.
There is absolutely no scientific studies done on APD to show its validity. Every other generally accepted statistic out there is published and judged by peer groups. Ask yourself the question if APD is such a superior statistic, why is no one else using it and why aren’t there nobel prizes handed out?
Just another observation about something I have just noticed. Most of the posters in this forum is ignored by 30% to well over 50% of other people. How can you possibly think this is a democratic process if people are not even paying attention to you at all?
I was proposing to Ross that we do such a study. And we have plenty of data to support what he says from this site.
Anyway, I don’t think 3 will be on the ballot, because there’s too much evidence for it.
And, Fanus, if you’re wondering why you’re now ignored by 1% of those who use the ignore feature, it’s you just got 1 ignore from me.
Apparently there’s only 100 people who use it, and I’m 1% of that total, so it shows you’re ignored by 1% of those who use the ignore feature, not by everybody on the site.
Congrats. It take real effort to get me to do that on one of your first appearances with me.
For what it is worth, my vote would be for 3. If one have to add so many disclaimers and exceptions to a statistic to point out cases where it might or might not be valid, then it is useles. If a car has a broken engine, painting it a different color might make it look nice and shiny, but the car is still broken.
You’ve obviously not spent the 5 minutes it takes to think about the statistic.
And it appears your vote is biased, since you only have 2 systems above 0.2 on the APD that would make it as a reasonable option.
So I can see why you want to remove it, and that is something that obviously makes you uncomfortable to discuss.
And this is why I stopped posting to these forums. A minority of people who are ignored by a big portion think they get to call the shots and claims to be involved in productive and democratic discussions, but the moment someone disagree with them, they launch personal attacks.
Apparently there’s only 100 people who use it, and I’m 1% of that total, so it shows you’re ignored by 1% of those who use the ignore feature, not by everybody on the site.
So, using your argument, you are ignored by 28 people and you don’t think this is representative of everyone on the site, but yet you think a process to decide about changing the description and participated in by less than half the number of people ignoring you IS democratic and representative?
Thanks for reminding me with your post why I stopped using the forums.
I don’t care about APD as I have said, this is not based on any kind of scientfic study and is not accepted anywhere else. If you disagree, please point to the published studies and references about where else it is used.
Have you considered that my systems are maybe some of those exceptions where APD is not really relevant?
In this democratic process, is only people with a specific APD ratio allowed to vote and provide their opinion? Or are they allowed to provide their opinion if they have a low APD ratio, as long as they are in favor of it?
I am not interested in getting involved in who has the better system or launching personal attacks as you apparently are. Starting with personal insults are always a sure sign that one party is losing the arguement. If you can’t win the match, at least you can win the fight, right?
Well, you brought it up with ignore rate, and I guess you think that’s everyone on the site, because that’s what you implied?
1% equals one person. The rest of the community doesn’t ignore us.
There are no exceptions, as I have analyzed 100’s of backtests to see that they’re pretty consistent. I’ll take this time to analyze the grid of active systems.
You don’t need a study to explain APD, because it is what it is.
