Can anyone name a few reasons why a strategy would perform differently on a 300 second chart compared with a 5 minute chart?
I am not using barssinceentry or barinterval. The sessions are identical.
thanks
300 second chart vs. a 5 minute chart
-
- Posts: 223
- Joined: 16 Oct 2011
- Has thanked: 66 times
- Been thanked: 20 times
- TJ
- Posts: 7744
- Joined: 29 Aug 2006
- Location: Global Citizen
- Has thanked: 1033 times
- Been thanked: 2223 times
Re: 300 second chart vs. a 5 minute chart
1. starting times are not the same.Can anyone name a few reasons why a strategy would perform differently on a 300 second chart compared with a 5 minute chart?
I am not using barssinceentry or barinterval. The sessions are identical.
thanks
2. ending prices might not be the same.
-
- Posts: 223
- Joined: 16 Oct 2011
- Has thanked: 66 times
- Been thanked: 20 times
Re: 300 second chart vs. a 5 minute chart
Would you please elaborate TJ?
why would a resolution change (seconds to min) that is equivalent (300s = 5 min) be any different?
why would a resolution change (seconds to min) that is equivalent (300s = 5 min) be any different?
- TJ
- Posts: 7744
- Joined: 29 Aug 2006
- Location: Global Citizen
- Has thanked: 1033 times
- Been thanked: 2223 times
Re: 300 second chart vs. a 5 minute chart
Would you please elaborate TJ?
why would a resolution change (seconds to min) that is equivalent (300s = 5 min) be any different?
What would you like to elaborate ?
You can do a export from your charts to see if the data are the same.
The problem is not in MultiCharts, it is a simple mechanical process.
The problem is in how/what your dataprovider streams the data to you,
and the time stamp when they are mated up to form the bars.
Theoretically they should be the same.
-
- Posts: 401
- Joined: 19 Jan 2011
- Location: Netherlands
- Has thanked: 85 times
- Been thanked: 125 times
Re: 300 second chart vs. a 5 minute chart
In realtime, there can be a difference between the closing of the bar. With tick charts the timing of the closing of the bar depends on the number of ticks that are processed. With minute charts the timing of the close of the bar is on a exact time. The close of the tick and minute bar will probably differ most of the time. That is what TJ already said about the mechanical process.
I don't know if you use IB data. But the following could be also causing a difference with backtesting. If you compare historical data with a 300sec or 5min chart.
From my understanding for IB data (and maybe other datafeeds?), MC is using tick data to create minute data. IB does not provide realtime minute data. But they do provide historical minute data. The minute data is fetched from historical data the next time you load the chart and replaces the minute data MC has created with the tick data. This will result in different behaviour of your strategy in realtime and backtest if you use minute data. And it will also create difference between a 300sec and 5min chart in a backtest with collected realtime data.
There is a entry in the PM for this:
http://www.multicharts.com/pm/viewissue ... _no=MC-763
I don't know if you use IB data. But the following could be also causing a difference with backtesting. If you compare historical data with a 300sec or 5min chart.
From my understanding for IB data (and maybe other datafeeds?), MC is using tick data to create minute data. IB does not provide realtime minute data. But they do provide historical minute data. The minute data is fetched from historical data the next time you load the chart and replaces the minute data MC has created with the tick data. This will result in different behaviour of your strategy in realtime and backtest if you use minute data. And it will also create difference between a 300sec and 5min chart in a backtest with collected realtime data.
There is a entry in the PM for this:
http://www.multicharts.com/pm/viewissue ... _no=MC-763
Re: 300 second chart vs. a 5 minute chart
Did you guys test with what data for 5min & 300sec Multicharts runs the strategy ? Did you test it in real-time and on historical data ?
I assumed that <MIN> data are build by a fixed amount of time for a full cycle of minimum 1min beginning with 0 at every new minute. That seems to be the case.
So overall on historical data 5min and 300sec charts are equal. At least here on my end ..... In the case of realtime data I have not one single difference today between both ..... I do not see it as a data feed problem either, as even IB delivers the tick in relatime which changes the price ....
Kind Regards
Ben
I assumed that <MIN> data are build by a fixed amount of time for a full cycle of minimum 1min beginning with 0 at every new minute. That seems to be the case.
Further I assumed that calculations for <SEC> charts are starting with the time of the day the first tick/trade was made. Additionally I thought and believed every new bar was set with the time of the 1st tick/trade after the previous bar ran out of time as 1st tick time_s + 300sec ....data calculation begins at 8:00 no matter when the 1st price/tick was made till 8:05 ... > 8:05 > 8:10 .....
This is obviously not the case. Not in real-time nor w/historical data. Further I am surprised that even <SEC> charts start their cycle at round numbers no matter when the first tick of the session was made.data calculation begins at 8:00.18 ... > +300sec ... bar closed with last price/tick before or equal to 8:05.18 ... > new 1st price/tick after 8:05.18 traded at 8:05.32 ... > +300sec ... closing of the bar at 8:10.32 .... and so on
So overall on historical data 5min and 300sec charts are equal. At least here on my end ..... In the case of realtime data I have not one single difference today between both ..... I do not see it as a data feed problem either, as even IB delivers the tick in relatime which changes the price ....
Kind Regards
Ben
-
- Posts: 223
- Joined: 16 Oct 2011
- Has thanked: 66 times
- Been thanked: 20 times
Re: 300 second chart vs. a 5 minute chart
Thanks for the replies-
I have tried using fixed ASCII data, and IQfeed data. The data does seem to match up.
In both circumstances the MC platform is a bit finicky on loading more than 2 years of data when using seconds resolution.
To get back to my original question, I wonder what other key words, etc. would affect results when transitioning from minutes to seconds
I have tried using fixed ASCII data, and IQfeed data. The data does seem to match up.
In both circumstances the MC platform is a bit finicky on loading more than 2 years of data when using seconds resolution.
To get back to my original question, I wonder what other key words, etc. would affect results when transitioning from minutes to seconds
- TJ
- Posts: 7744
- Joined: 29 Aug 2006
- Location: Global Citizen
- Has thanked: 1033 times
- Been thanked: 2223 times
Re: 300 second chart vs. a 5 minute chart
Thanks for the replies-
I have tried using fixed ASCII data, and IQfeed data. The data does seem to match up.
In both circumstances the MC platform is a bit finicky on loading more than 2 years of data when using seconds resolution.
To get back to my original question, I wonder what other key words, etc. would affect results when transitioning from minutes to seconds
Because Easylanguage strategies are bar specific,
if the bars are the same, the strategy should behave the same.
Can't tell you more without specific info and code.
-
- Posts: 223
- Joined: 16 Oct 2011
- Has thanked: 66 times
- Been thanked: 20 times
Re: 300 second chart vs. a 5 minute chart
For example, using barinterval doesn't work. Barinterval returns "300" for a 300 sec. chart and "5" for a 5 minute
- TJ
- Posts: 7744
- Joined: 29 Aug 2006
- Location: Global Citizen
- Has thanked: 1033 times
- Been thanked: 2223 times
Re: 300 second chart vs. a 5 minute chart
For example, using barinterval doesn't work. Barinterval returns "300" for a 300 sec. chart and "5" for a 5 minute
For code questions, please post your code.
Yes, for those type of measurements, you cannot cross the denominator and expect the same result.
That is just simple logic.