TESTING: BackTesting & Autotrading

Questions about MultiCharts and user contributed studies.
User avatar
danilo
Posts: 170
Joined: 02 Jan 2007
Location: Italy
Has thanked: 4 times
Been thanked: 9 times

TESTING: BackTesting & Autotrading

Postby danilo » 14 Oct 2008

I have tested an intraday strategy on 7 futures.
I have attached the results, the meaning of columns should be clear, the only one that needs an explanation are:

Method= Async Autotrading with timeout = 30sec
Trades = the numbers of round-trip (Buy + Sell)
Market = the Market's P/L on TWS (true executions)
Theo = the Theoretical P/L on MC
Theo (Reload) = the Theoretical P/L on MC after closing & reloading the ws

Some comments:

1) The reliability of autotrading has been improved all the positons was flat at the end of day

2) There is a good agrement for OMX and NQ

3) For SPMIB Market=Theo but after reloading the ws the P/L is > 0

4) For CL and EMD Market & Theo are quite near but Theo(Reload) is very different

5) For ES there is a good agreement for Market & Theo(Reload) but Theo is very different

In Summary: the MC autotrading has been improved, but whitout a reliable backtesting it doesn't make sense spent a lot of efforts to develop "intraday strategy", assuming that the problem could be solved with the "look-inside" bar implementation I cannot explain the comment n.5, so I think that more tests are necessary.

Regards,
Danilo
Attachments
Test_Results.png
BackTesting & Autotrading Test Results
Test_Results.png (4.65 KiB) Viewed 300 times

Moulinex
Posts: 10
Joined: 26 Sep 2008

Postby Moulinex » 15 Oct 2008

I still repeat that it doesn't make sense implement an intraday strategy without IOG...
How could be truthful a backtesting if in realtime orders are sent in intrabar manner and backtesting not???
I relate especially to bars with more than 20 tick...
Have you any idea how much the values are faked??

Bravo Danilo!!

SUPER
Posts: 624
Joined: 03 Mar 2007
Has thanked: 99 times
Been thanked: 80 times

Postby SUPER » 15 Oct 2008

Danilo,

What type of orders did you try, stop, limit or market?

Just shows how important it is to have look-inside bar feature.

Regards
Super

User avatar
danilo
Posts: 170
Joined: 02 Jan 2007
Location: Italy
Has thanked: 4 times
Been thanked: 9 times

Postby danilo » 15 Oct 2008

SUPER, only stop orders

Regards,
Danilo

User avatar
Marina Pashkova
Posts: 2758
Joined: 27 Jul 2007

Re: TESTING: BackTesting & Autotrading

Postby Marina Pashkova » 22 Oct 2008

I have tested an intraday strategy on 7 futures.
I have attached the results, the meaning of columns should be clear, the only one that needs an explanation are:

Method= Async Autotrading with timeout = 30sec
Trades = the numbers of round-trip (Buy + Sell)
Market = the Market's P/L on TWS (true executions)
Theo = the Theoretical P/L on MC
Theo (Reload) = the Theoretical P/L on MC after closing & reloading the ws

Some comments:

1) The reliability of autotrading has been improved all the positons was flat at the end of day

2) There is a good agrement for OMX and NQ

3) For SPMIB Market=Theo but after reloading the ws the P/L is > 0

4) For CL and EMD Market & Theo are quite near but Theo(Reload) is very different

5) For ES there is a good agreement for Market & Theo(Reload) but Theo is very different

In Summary: the MC autotrading has been improved, but whitout a reliable backtesting it doesn't make sense spent a lot of efforts to develop "intraday strategy", assuming that the problem could be solved with the "look-inside" bar implementation I cannot explain the comment n.5, so I think that more tests are necessary.

Regards,
Danilo


Hi Danilo.

I apologize for not getting back to your earlier. I have asked our engineers to analyze your post and I'll get back to your with the answer on Friday.

Regards.

User avatar
Marina Pashkova
Posts: 2758
Joined: 27 Jul 2007

Re: TESTING: BackTesting & Autotrading

Postby Marina Pashkova » 28 Oct 2008

I have tested an intraday strategy on 7 futures.
I have attached the results, the meaning of columns should be clear, the only one that needs an explanation are:

Method= Async Autotrading with timeout = 30sec
Trades = the numbers of round-trip (Buy + Sell)
Market = the Market's P/L on TWS (true executions)
Theo = the Theoretical P/L on MC
Theo (Reload) = the Theoretical P/L on MC after closing & reloading the ws

Some comments:

1) The reliability of autotrading has been improved all the positons was flat at the end of day

2) There is a good agrement for OMX and NQ

3) For SPMIB Market=Theo but after reloading the ws the P/L is > 0

4) For CL and EMD Market & Theo are quite near but Theo(Reload) is very different

5) For ES there is a good agreement for Market & Theo(Reload) but Theo is very different

In Summary: the MC autotrading has been improved, but whitout a reliable backtesting it doesn't make sense spent a lot of efforts to develop "intraday strategy", assuming that the problem could be solved with the "look-inside" bar implementation I cannot explain the comment n.5, so I think that more tests are necessary.

Regards,
Danilo


Hi Danilo,

I am sorry it took me so long to get back to you.

Look-inside-bar is planned for future versions of MultiCharts.

However, there are other factors that could result in the difference between the theoretical P/L and the theoretical P/L after re-opening.
For example: desynchronization or changing of the barsback request (the starting point could have changed).

Theoretical P/L and real P/L in the asynchronous mode is natural. Prices for the theoretical P/L are based on the chart prices, while the real P/L is based on bids and asks. Also, desynchronization can contribute to the differences.

Regards.

User avatar
danilo
Posts: 170
Joined: 02 Jan 2007
Location: Italy
Has thanked: 4 times
Been thanked: 9 times

Postby danilo » 29 Oct 2008

Marina,

When the WS is reload for each bar are known only the OHLC values, MC guess the path that connect the OHLC, during real time MC know the exact path and then is able to execute correctly the stop-loss and stop-profit orders. That's the main reason why the real-time P/L could be different from the reload WS P/L. The solution to this problem is the implementation of the look-inside bar. Implementation that we are waiting for a long time...

Regards,
Danilo


Return to “MultiCharts”