I am posting this as a new thread because no one with answers seems to address this issue.
The issue is the erroneous data that is being reported in the Drawdown column for individual trades that is then being utilized to calculate the ADP ratio. Over the last few weeks, probably about 1/2 of the DD’s reported on my system’s trades have been overstated. Some have not even been close! I just checked the DD’s reported on two trades for Friday and both were overstated. The prices and the times of the largest drawdown amount and time period are BOTH incorrect. Why is this? Does every single vendor have to spend the time checking these amounts every day so that C2 has to then spend the time to adjust this? This is absolutely silly that these numbers can be so consistently wrong at least over the last few weeks and at least in regards to the ER2.
What’s up?
Cheetah:
I simply don’t agree that half of all dradwown stats are inaccurate.
As to why there are any errors: Drawdown stats are calulated using minute-by-minute historical data. They are calculated after trades close. If there’s an incorrect tick in the historical data provided by the data vendor, then of course this will throw off the drawdown stat. I have tried to create different levels of filters to throw out obviously bad data, but it’s not as simple as it sounds. Some instruments are quite volatile, which means that a spike up or down may be real data and not just a garbage tick.
But I do try to adjust the filters as I get more experience and see which symbols are prone to bad ticks. The point is: Please report any bad drawdown data when you see it. I know it’s a hassle, but this will allow me to (1) fix the cases that are incorrect, (2) see which instruments are prone to bad ticks, and work with our data vendors to insure cleaner data, and (3) establish more meaningful software filters on bad ticks.
I may also work on some sort of automated checking of drawdown data, to spot and correct problems in the background. But for now – yes, please take the time to look at your drawdown data from time to time to spot any glaring problems.
In the meantime, I will continue to improve the software that operates behind the scenes. I know that this is a source of frustration for system vendors, and I will continue working on it.
Drawdown data is vitally important to analyzing system performance, and so I think it’s important to keep drawdown data functionality even as we clean up bad data.
MK,
Thanks for the reply. To clarify, I didn’t say that 1/2 of drawdown stats are inaccurate. I was only referring to the ER2 stats on my system over the past few weeks. In fact, seven of the last eleven trades had incorrect DDs on my system. As I have noted, perhaps this is just an ER2 data problem and perhaps it has just occurred over the last few weeks. I cant take the time to check other instruments data. I am only concerned with the ER2.
I believe you have said that five minute data is used to determine the DDs. Out of curiousity, I checked another system on C2 that trades the ER2 exclusively to see if there were errors. I didnt have to look too far. The last trade posted on that system was a short sale on 1/25 at 11:58 at 792.00. The DD price is posted at being 793.80 at 12:00 (using 5 minute data I assume). But there is a problem with using the 5 minute bar for drawdowns. That means you are using data that occurred prior to the execution of the trade (11:55 to 11:57). The DD price of 793.80 (my data actually shows 793.60 should be the DD price) actually occurred on that five minute bar to the fill of the trade at 792.00. In this case, the sell was a very good trade and from looking at the one minute bars, it would appear as though the real DD was more like .30 instead of the 1.8 calculated by C2. The number of contracts traded was 4. Actual DD is more like $120 and C2 posted a DD of $720. This is rather significant considering it was only four contracts and the difference is because the DD is calculated using five minute bars which allows price prior to the actual fill to be used in the DD calculation.
I would suggest that if C2 is going to use five minute bars, then the first five minute bar in the calculation should be the bar that occurs after the trade is filled. Then when the trade is closed, the last five minute bar in the calculation should be the bar prior to the bar that contains the trade being closed. To make the DD more accurate, I would recommend using one minute bars with the same criteria if that is feasible. Some systems might be in and out of a trade in less than 30 minutes, so using 5 minute bars does not seem like the best way to go.
Obviously there are problems in DDs, and it is useful to continuously purge problems. But I agree that even if problems, they offer more information. And I would suspect when looking at the total, some may be under and some may be over the true drawdownm and for a summation, this might even up a bit…
One thing you might want to consider - the USD/CAD system had a $225,000 drawdown, which seemed orders of magnitude impossible. Things like this also happen for the net profit column. I once had a $17,000 loss in a day, for one contract of Rough Rice!!!
Maybe you should have serious outliers (DDs that are well outside the last 20 or 50 or whole set of DDs, or that are some percentage beyond the value of the underlying instrument) auto-report to you for examination, and keep it "No calc" until it is verified?
> I would suggest that if C2 is going to use five minute bars, then the first five minute bar in the calculation should be the bar that occurs after the trade is filled. Then when the trade is closed, the last five minute bar in the calculation should be the bar prior to the bar that contains the trade being closed.
NO! That is just as unacceptable, inaccurate, and arbitrary as the current method, but it simply favors the vendor.
> To make the DD more accurate, I would recommend using one minute bars with the same criteria if that is feasible. Some systems might be in and out of a trade in less than 30 minutes, so using 5 minute bars does not seem like the best way to go.
Agreed. Or better yet tick data.
