In response to customer reaction, I have changed the name and description of the Hold and Hope Statistic to the “Average Profit To Drawdown Ratio” (APD) and removed references to one single C2 user’s metrics as to which systems are tradable.
Even so, I would like to publicly thank Ross for urging me to include this statistic on the site. I do find it very valuable, and I appreciate Ross’ help in developing it.
Matthew
Matthew
Thank you for making the adjustment.
I think the way it is named and described as another statistic to evaluate systems, is fair and neutral. Do I sound like Fox News? 
Regards
- Fanus
Thanks for the responsiveness, Matthew. This name and description are much better, precise and neutral. I think it is indeed very convenient to have this statistic, as it captures a characteristic that cannot be derived from the other statistics.
I also think that it is wise to keep some reference points in the description, as you do now. A new statistic is often difficult to interpret if you don’t know which value is “normal”. It’s a smart move to derive the reference points from the distribution at C2.
I can imagine that it would be a little easier to understand for newbies if you present the outcome as a percentage (i.e. 0.35 would become 35%), like most other statistics in the orange/grey table. It is of course mathematically equivalent, but I believe that many people are more used to think in percentages than in fractions.
It can still be worthwhile if Ross writes his original metric on the forum (I mean his personal reference points and the judgemental labels that were deleted from the grey box) . After all, he has the most experience with it on this moment. This is still better than the previous solution, because there will be a clear separation between facts and opinions, and others will be able to comment on it like equals.
One substantive point: Cheetah pointed out that there can be positive and negative numbers in the drawdowns that are reported in the trade details. Will the algorithm deal with that properly?
Another question for MK is this:
What happens with trades where the DD is noted as being "N/A"? Is the trade eliminated from the calculation? I was looking at BIG CAT and noticed that there were a number of trades that had N/A as the DD figure. I think that the description of the indicator would need to tell us what happens with these trades in the calculation. Beyond that…how can trades have a DD that is N/A? Why is that and what does that mean?
Final thoughts on this:
1) MK Thanks for your willingness to listen to all. I especially like your giving actual stats at the 0.2, 0.4, and 1.0 levels.
– I suggest APD hs no meaning to readers. Can you just squeeze "Profit-to-DD" in the original window? At least it has some idea what it is.
– And in the popup, maybe add a sentence of what the indicator is This indicator is a ratio of 2 columns in track record: sum of all values in max DD (without "no calc") under sum of all their corresponding net profits
2) I would like to say that I have found this the most useful value when I assess systems. Some vendors work a good equity curve via averaging down, holding on until it turns profitable, scalping methods, or never closing trades.
Excepting for a few methods (hedging, etc.) smaller numbers indicate increased punishment is inflicted on subscribers for the same amount of net profits. Someone with APD of 0.2 inflicts about 5 times as much per-trade drawdown as someone at APD of 1.0. And one must ask whether someone with a pretty low number even offers any value. It is likely random entries, picking off some profit target, and randomly exiting might do as well…
3) I requested MK add this in, as I thought others would find value in it and it saves hand calculating it. MK asked for a description to accompany it. I gave precisely the purpose and parameters I found in putting many systems through it, and seeing what resulted. Credit was never sought nor requested.
About the drawdown stats: People have asked why some numbers were displayed as negative, and some as positive. The answer is that there are, in fact, two separate software routines for calculating and displaying drawdowns - one for forex, one for everything else. Due to sloppiness on the part of one of my key programmers (me), the result was that some drawdown numbers were listed as negative, and others positive. However, all numbers were stored in the database identically (and correctly). I have since reprimanded the programmer and fixed this inconsistency, and now all numbers are displayed as negative.
Cheetah asked why some drawdown stats have “n/a” next to them. This happens either when (1) a trade was entered and exited faster than the granularity of my historical data (i.e. the trade took place for only 4 minutes, but I have only 5-minute bars available to study), or (2) the historical data from my third-party data provider has too many bad ticks in the time period in question, and is thus unreliable.
While we strive for accuracy with all data, it’s not possible 100% of the time. The idea, though, is that there will be enough other, clean drawdown data points to allow accurate measurements, and that errors, when they do appear, will get lost in the wash.
Finally, to answer your specific question about how the Average Profit to Drawdown ratio is calculated: yes, only those trades with valid drawdown stats are used in the calculation.
Matthew
MK,
Thanks for the reply.
Another question–Would it ever be possible to have the trade data somehow be downloadable in some type of spreadsheet format so that all trades can be downloaded at once instead of cutting and pasting 25 trades at a time?
Yes. That is functionality that will be added in the near term.
Thanks, Matthew. If anything your APD explanation text is a bit overly kind to systems (like mine) which manage risk by multiple positions. It would probably be more accurate to say that for such systems it is unlikely that all trades would be in drawdown simultaneously. We all know the market can do whatever it wants. 
On a related subject… would it be possible to get the APD Ratio stat in ‘My Analyst’ pages? The ideal would be for each C2 to be able to customize their personal ‘My Analyst’ view to show the stats they want to see including Jules ‘advanced’ stats.
Thanks for your great work and responsiveness.
> Cheetah asked why some drawdown stats have “n/a” next to them. This happens either when (1) a trade was entered and exited faster than the granularity of my historical data (i.e. the trade took place for only 4 minutes, but I have only 5-minute bars available to study)…
This also explains (to some extent) the higher than exact DD stats for Cheetah. If you check the 786.10 1/26/07 10:31 ($2,320) DD
(ID # 24771607) you will see on one minute data the exit is @ 10:31 @ 786.10, but the *5-minute high @ 10:32-10:33 (AFTER the trade is exited) is 786.40.
Nonetheless, C2 shows 786.80 as the “worst price”. Bad tick? Neither TS nor IB show 786.80 within the 5 minute range. They do show 786.90 after 10:39…but 786.90 still isn’t 786.80 ;-), and neither price should apply to the trade’s DD.
May I ask who the data vendor is? Are you using bid / ask or actual trade data? Thank you.
"It would probably be more accurate to say that for such systems it is unlikely that all trades would be in drawdown simultaneously."
Keep in mind, this has nothing to do with managing drawdown or system drawdown. This has everything to do with how much drawdown a system averages for a given net profit. In other words, many systems with low ADP have no predictive value. They simply make profit by averaging down, hanging on until trade turns around, etc…
This tries to see if a system has timing potential, or is generally better than random
>>Keep in mind, this has nothing to do with managing drawdown or system drawdown. <<
Exactly the point MK’s text makes quite clearly.
Ross, it seems to me that given the stated purpose of your APD ratio, it is best applied only to closed trades. If the market isn’t moving much, trend follwoing systems like mine “let winners run” by holding positions until it does. Perhaps a % confidence indicator should accompany the ADP ratio based on the percentage of closed trades a system has.
For example, the system Big Calls has APD Ratio 3.13 while no closed trades are shown. Well, the statistics for this system are remarkable:
Trades 3
Profitable 2
Losses 1
Win % 66.7%
APD Ratio 3.13
Avg Trade Length 394.2 days
Compound Annual % 258% over 394 days
Max Drawdown 1480.65% (20051231 to 20061215)
How can a Win % and APD ratio be shown if there are no closed trades? (Unless I looked exactly in the 5 hour delay after they were closed). How a system can trade after a drawdown of 1480% still amazes me.
Jules: The max drawdown stat for that system was incorrect (I cleaned a garbage data point on its equity chart), but everything else is correct. C2 includes open positions in all stats, including winning % and APD.
>it seems to me that given the stated purpose of your APD ratio, it is best applied only to closed trades.
In fact, all of the statistics (and equity curves) should be based on closed trades. The faster C2 reach that solution, the faster its credibility will be restored…
I think the new name, explanation and stats as to where the current systems fall in are much improved.
Yes, much improved, and I thank MK for his responsiveness.
I wonder what the AVERAGE ADP rating is for all systems. Considering that only 30% are above .20. I fear it may not be a pretty picture.
In the future it might benefit MK if he passed new ideas through a few experienced traders to collect views and insights before going public. It may help to avoid hanging out all the dirty laundry we’ve just witnessed.
The idea here seems to be to pass new ideas through everyone interested in contributing. And it seems to work very well.
MK tends to sit back and wait till the dust settles a little, and then steps in with his take on it all and makes the final discision.
Thats a good point… makes you wonder if the really high ratings are statiscal outliers and will eventually revert back to the mean… Providing an everage ADP figure makes sense to see where you are, compared to balace of the group. If we find that .18 is the mean, then someone with a .30 sytem may actually have a very acceptable model. There will always be those on the extremes, but often the high readins are those models that are running at extremely effecient levels that tend to fall off in time, especially when multi-year track records are reviewed (if there is no style drift in the model).
> In fact, all of the statistics (and equity curves) should be based on closed trades. The faster C2 reach that solution, the faster its credibility will be restored…
Situational ethics again. You didn’t like the ADP because it wasn’t SOP,
but now you want C2 to track equity and stats based on closed trades
only when the entire Brokerage / System Testing / CTA / Hedge Fund universe tracks open position equity and drawdowns.
You can’t have it both ways.
