|
|
|
| |||||
|
|
| Interest Rate adjusted Data | ||||
To trade futures on a long-term basis anybody
needs a long track of correct data. But the most of the active
lifetime of the well known futures contracts are about 3
months long from "Roll-Over-Day" to
"Roll-Over-Day" (ROD to ROD). Putting those together
in one database to satisfy your analytical and charting needs, you are
getting a terrible "noise" - the gaps in prices between two (old
and new) contracts. 
The value of gap
is different from time to time and varies according market
environment. To eliminate this "noise" we adjust
actual data. There are many ways to adjust futures data into
the long term history. We use so named INTEREST RATE
ADJUSTMENT method. With this method the value of adjustment is
a function of current price level, number of days until contract's expire
day and current Federal Reserve Interest rate.
|
As a result we have the most correct data of
a Continued futures Contract (CC). It looks like a new
indicator that closely follows spot market. More over, it is built into the actual futures data.
We have inserted it into the last field of a standard
field's set - instead of OI (Open Interest). As you know,
the OI field is usually empty if we use intra day futures
data.
|
|||
| What is CC Data file exactly | ||||
| The intraday (1 minute) Data is in the universal ASCII text (csv) format that is compatible with the most of charting packages, such as Trade Station (Omega Research), MetaStock (Equis) and other. MS Excel works with CSV files with no problem, simply double click file name and Excel will open it as a sheet. To turn our 1 minute data into the 3,5,10 and so on min data somebody have to use our CC Data Converter - it is free to download and install.. | Our data looks like "2 in 1". User
will get traditional actual open-high-low-close-volume data and adjusted
close data in the same file: |
|||
| Understanding Data accuracy | ||||
|
Why it is important to use most accurate data in system
developing? All data providers have data that may differ from data CME has itself. With the sample below the cumulative effect of the 5*0.25 on a seven bars was reflected in a triggered Buy signal on a CME data one bar earlier than on a eSignal data. Signal generation one bar earlier/later may bring additional win/loss on a short time period. With a long time trading win/loss as a result of difference in data will offset each other. But if you use data from different data vendors you might get strange result in track during system testing |
We can see differences in close prices of eSignal and CME 15 min
data (11/13/2002): == eSignal close === CME close === difference = 1330, 00882.75 || 1330, 00882.75 || 0.00 1345, 00878.50 || 1345, 00878.75 || +0.25 1400, 00873.25 || 1400, 00873.50 || +0.25 1415, 00874.50 || 1415, 00874.75 || +0.25 1430, 00876.75 || 1430, 00877.00 || +0.25 1445, 00878.50 || 1445, 00878.50 || 0.00 1500, 00882.25 || 1500, 00882.50 || +0.25 1515, 00886.00 || 1515, 00885.75 || -0.25 The reason is a delay in communication mainly while a strong trades flow. |
|||
| Data makes a difference | ||||
|
"...Even if all vendors reported prices the
same way, differences in their data still would exist. Data are
processed and reformatted several times between the pit and trader's
computer, and errors can occur along way.
|
The system itself is a simple version of the popular
volatility-adjusted breakout strategy. It is always in the market,
either long or short. This equity curves highlight the variability
in the system test result based on the data differences. The
test period was 2 years long (from Jan. 97 to Dec. 98.) The results varied greatly, with net profit ranging from $95,275 with Historybank.com data to $127,500 for CSI, and drawdowns of $25,375 for CSI to $45,475 for Bridge. When trading a system in real time, you must trigger your trades with the same kind of prices used for testing or the results will not match the test..." From the article:
Sheldon
Knight. |
|||
|
|
||||
| Understanding Roll Over Date | Understanding Length of data history | |||
|
If you have traded futures before you know that futures
contracts have expiration dates. If you have an open positions
and this expiration date draws near it would be a wise idea to
roll over to the next active month in order to avoid delivery
obligations. Typically traders like to roll when the
volume and open interest in the current month (the one we have
the position in) drops below the volume and open interest of the
next active month. On the e-mini S&P 500 and Nasdaq 100 futures roll over date comes 10 days before expiration date. There is a last active trading day of the current contract on Wednesday (a week before expiration week) and there is a first active trading day of the next contract on Thursday (a week before expiration week).
|
How long data history must be to test system? There is no only answer this question. The common rule is - the data must include uptrend and downtrend periods. So, having three last years of index data history, you risk to develop system that wins on the declining market only, but what about system's behavior when market goes up. Fore - five years of historical data is an optimal period for e-mini system testing. |
|||
© Copyright 2000-2006 ANFutures.com |
