# APD discussion

I thought we should probably move the APD conversation off of Spyder’s system into the general comments, I would like to see if others agree with my last post suggesting a potential compromise as follows:-

I’m glad Ross says he can see what we are saying, it seems we agree no stat is perfect, perhaps the problem here is the definition and how it is interpreted.

The description alongside APD is too definitive and rigid to me. It 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.

That is pretty definitive and I think that is part of the problem, most people don’t know APD and will want to know what it is and what it measures, this makes it pretty clear that that’s ALL it can mean, and that to me is wrong. Can we not agree on that?

So to compromise I think we could leave the calculation/measure of APD exactly as it is, but at a minimum we should change the description so as to give a fair basis for people to interpret it. What do people think?

I agree. Also I think that the formulation that Ross suggested in the old thread is an improvement.

You mean this -

""The best description would be to simply state what APD reflects, and how it might be used:

"Low APD means the sum of all maximum drawdowns per trade are very high compared to the sum of all net profits. This can reflect a system that averages down or “Hold&Hopes”"

I agree, it’s a start, it’s accurately stating what it’s calculated but it needs an additional way of interpreting it, there is an ‘OR’ at the end of that second sentence waiting to be written, just need to finish it off, as ‘reflecting that a system averages down or h&h’s’ is just one interpretation.

Low APD means the sum of all maximum drawdowns per trade are very high compared to the sum of all net profits. This will mostly result from either averaging down and holding onto losing positions in the hope they turn around, or alternatively from having many small losing trades and less frequent large winning trades.

With the above the reader now knows that those aren’t the only two interpretations but they are the most likely. They also then have the means to check the other statistics to determine which is the more likely for the system they are reviewing. If they see the system also has a avg win > than avg loss and/or a low win % they can easily conclude that is likely why they have a low APD and then if having a low win % is not something which is a concern to them they needn’t be concerned with APD being low.

Yes, that is the description that I meant and I agree that it is not complete. Instead of "This can reflect a system that averages down or “Hold&Hopes” I would suggest this:

"A high APD means that on average, the system makes trades with high returns in comparison to the drawdowns. A low APD, on the other hand, can have different causes. One possible cause is that the system frequently applies a ‘hope and hold’ strategy, letting its loosers run in the hope that they can be closed at small profit. This is considered a high risk strategy. Other possible causes of a low APD are that the system applies a hedging strategy, or that the system has many trades with both small drawdowns and small profits. These strategies do not necessarily imply a high risk. In summary: a high APD is good but a low APD is not always bad."

Your formulation is shorter than mine, which is probably what most people prefer. Two comments:

1. "many losing trades": this suggests that the trades are closed at a loss, in which case the system is loosing money. A low APD is also possible with many trades that have a small profit.

2. "will mostly result": That is what receives most attention, but I am not so sure that this is really the most frequent cause. Hedging and diversification influence APD too.

Fair points, I agree, mine is thinking a little too much of trend followers only, I also think you are right to define it with a High APD at the beginning, this statistic should (ideally) result in a positive number so we should first define it by what a High APD means not what a low one means, ie what you are trying to achieve. Your summary is good as well (that a high number is good but a low number isn’t necessarily bad) for the aforementioned reasons but may be too general for some, let’s see.

That is way too verbose. The original "can reflect" already just gives two possible negative alternatives that can (but not must) happen, without rambling on about every possible case, Other stats do not do this, they are simple.

"or that the system has many trades with both small drawdowns and small profits. These strategies do not necessarily imply a high risk."

I really do not think I agree with that at all. A system that has an APD of 0.10 is trash. A person awash in so many drawdowns seeking small profits, is little more than blindly getting in and out, and occasionally banking profits when it goes in his direction. Anyone can make a system that has a low APD and banks profits. That is what the whole point of averaging down and Hold&Hope do, even at the small scale level… This is not someone who outperforms the market. Again, seeking a lot of small profits is scalping, and we already know they tend to perform poorly in C2 autotrading and especially trying to do it manually.

If you saw a a lot of small drawdowns/losses seeking a LARGE profit, then I am more amenable than that. That is what some do and it makes sense. They often have low winning percentages but decent PFs…

"That is way too verbose."

Perhaps. But I am talking about the text that C2 displays when you move your mouse over the question mark next to APD in the basic system statistics. These texts are generally quite verbose, also for other statistics.

"A system that has an APD of 0.10 is trash."

That’s exactly the point on which we disagree. This implication is not proven and therefore it should not be part of the C2 description.

"A person awash in so many drawdowns seeking small profits, is little more than blindly getting in and out, and occasionally banking profits when it goes in his direction."

No, he may have an almost perfect timing in a trend following system with tight stop losses, a high probability of playing even (after commissions and slippage) and a small probability of a high profit. That would outperform the market.

"Again, seeking a lot of small profits is scalping, …"

Perhaps, but even so, scalping is not necessarily a high risk strategy. Yet the present C2 description says that low APD implies high risk. Furthermore, there are already other statistics that deal with slippage and commissions. Moreover, a low APD may actually cause less slippage because an initial dd means that the subscriber can open the trade at a better price.

"If you saw a a lot of small drawdowns/losses seeking a LARGE profit, then I am more amenable than that. That is what some do and it makes sense. They often have low winning percentages but decent PFs…"

Jules, what Ross says here is what I was alluding to when I was trying to allow for a system with ‘many small losses and occasional larger wins’ as these are trend following systems which are exactly as he suggests with low win % but decent PF’s. I don’t think mentioning a lot of small losses necessarily implies it’s a losing system.

"No, he may have an almost perfect timing in a trend following system with tight stop losses, a high probability of playing even (after commissions and slippage) and a small probability of a high profit. That would outperform the market."

I agree with this, but that’s not what it says, Ross is referring to the fact that is said ‘small profits’, which if by that you mean ‘small winning trades’ would be incorrect as trend followers will have infrequent large wins, and frequent small losses.

correct. a lot of small losses is supposed to be paired with occasional large trades. There is no point mentioning anything about small wins.

‘small losses’ is OK for me too.

I don’t think you’re giving enough credit to the subscriber base. They know what it means. I doubt any system will ever meet Ross’s 0.4 expectation over long periods of time. Search for systems over 2 years with APD’s of 0.4. Probably won’t be any.

I guess there were four out of the 6900+ systems we have now, which might be more survivorship bias than an indicator about future performance.

I think you’ve forgotten APD is something Ross created and was adopted by Matthew, so no-one outside of C2 can possibly know what it means without reading the description. That’s where they first discover what it means and learn how to interpret it, and that’s what we’re addressing because the current description is inaccurate.

I’d be fine with Net Profit/Total Drawdown. That would tell me everything I needed to know. Maybe I’m a bit more quantitative than most, but I think there would still be a lot of people that would get this definition immediately, espeicially if they’ve had theoretical exposure to excess return measures like the sharpe. That’s a very a specialized case and not the one your discussing, sure, but it does have something to do with what kind of subscriber you’re looking to attract. I would hope they can figure this out, and draw their own conclusions.

``````Long + Short	Long Only	Short Only	Buy &amp; Hold
``````

Starting Capital \$100,000.00

Ending Capital \$306,831.97

Net Profit \$206,831.97

Net Profit % 206.83%

Annualized Gain % 25.19%

Exposure 6.73%

Avg Profit/Loss \$31.52

Avg Bars Held 1.01

Winning % 64.17%

Gross Profit \$369,798.31

Avg Profit \$87.82

Avg Bars Held 1.01

Max Consecutive 171

Losing % 35.83%

Gross Loss \$-162,966.35

Avg Loss \$-69.32

Avg Bars Held 1.01

Max Consecutive 91

Max Drawdown \$-17,560.03

Max Drawdown Date 10/27/2008

Max Drawdown % -5.27%

Max Drawdown % Date 10/27/2008

APD 0.3427

Wealth-Lab Score 354.4838

RAR 374.2212

MAR 4.7751

Profit Factor 2.2692

Recovery Factor 11.7786

Sharpe Ratio 1.6238

Sortino Ratio 4.3340

Ulcer Index 0.8115

WL Error Term 4.8470

WL Reward Ratio 5.1961

Luck Coefficient 8.6290

Pessimistic Rate of Return 2.1891

Equity Drop Ratio 0.0000

K-Ratio 0.2402

Seykota Lake Ratio 0.0043

Expectancy 0.4895

Expectancy Score 626.9095

Congrats, pat on the back for me, and kudos go to Ross. That’s superbands backtest with an APD stat as a measure.

Now, assuming these statistics were possible, is that how you’d rate the most profitable chartscript ever written in Wealth Lab?

I believe I’m going to publish this in the wealth lab forums, and let people draw their own conclusions from it. I won’t go into the details because having this many statistics is a perfscript that I don’t look at the much. The way I coded the total dd was basically td:=td+abs(sharesPositionMAEPct/100PositionEntryprice)

It’s brilliant, and I know for a fact I’m the only vendor on this site that’d post that result.

It’s the most recent five years in there. I’ll post the full dataset going back to 1995 here for [LINKSYSTEM_35438029]

It’s an historic ocassion to be finally seeing a real world stat in the backtest.

And for all time, going back to 1/1/1996

Long + Short Long Only Short Only Buy & Hold

Starting Capital \$100,000.00 \$100,000.00 \$100,000.00

Ending Capital \$2,703,507.89 \$2,703,507.89 \$100,000.00

Net Profit \$2,603,507.89 \$2,603,507.89 \$0.00

Net Profit % 2,603.51% 2,603.51% 0.00%

Annualized Gain % 29.31% 29.31% 0.00%

Exposure 8.30% 8.30% 0.00%

Avg Profit/Loss \$90.63 \$90.63

Avg Bars Held 1.00 1.00

Winning % 60.06% 60.06%

Gross Profit \$5,492,539.71 \$5,492,539.71

Avg Profit \$318.37 \$318.37

Avg Bars Held 1.00 1.00

Max Consecutive 174 174

Losing % 39.94% 39.94%

Gross Loss \$-2,889,031.82 \$-2,889,031.82

Avg Loss \$-251.79 \$-251.79

Avg Bars Held 1.00 1.00

Max Consecutive 194 194

Max Drawdown \$-160,048.25 \$-160,048.25

Max Drawdown Date 10/27/2008 10/27/2008

Max Drawdown % -14.84% -14.84%

Max Drawdown % Date 4/17/2000 4/17/2000

APD 0.29750.2975

Wealth-Lab Score 300.5931 300.5931

RAR 352.9907 352.9907

MAR 1.9746 1.9746

Profit Factor 1.9012 1.9012

Recovery Factor 16.2670 16.2670

Sharpe Ratio 1.7330 1.7330

Sortino Ratio 2.6978 2.6978

Ulcer Index 2.8587 2.8587

WL Error Term 17.7820 17.7820

WL Reward Ratio 1.6484 1.6484

Luck Coefficient 21.6134 21.6134

Pessimistic Rate of Return 1.8692 1.8692

Equity Drop Ratio 0.1234 0.1234

K-Ratio 0.1772 0.1772

Seykota Lake Ratio 0.0088 0.0088

Expectancy 0.2786 0.2786

Expectancy Score 600.9768 600.9768

Max Losers Held 226 226

Max Winners Held 205 205

Everything’s twice because the long trades only section is identical to the backtest.