SmartQuant Discussion
http://www.smartquant.com/forums/

Paper vs Backtest and quote vs 1 min data (stat perf)
http://www.smartquant.com/forums/viewtopic.php?f=44&t=10277
Page 1 of 1

Author:  drolles [ Sat Jul 28, 2012 1:32 pm ]
Post subject:  Paper vs Backtest and quote vs 1 min data (stat perf)

Hi,

Could you please help me explain something or suggest some areas for possible examination?

I’ve been back testing a strategy that performs really well with 1 min data. I took that strategy and live tested it using OQ’s built in Paper trading facility (with the code handling 1 min signals). The Paper trading results were in line with the backtesting results based on 1 min data.

I then wanted to drive the level of granularity down to quote data to ensure I was getting the best fills possible. So I’ve written the strategy to handle quote data.

Once I did that the performance of the strategy deteriorated beyond recognition.

What do we thing is happening here? I understand that the 1 min backtests were not taking into account slippage or spread. However, once I was running it in Paper mode with good performance, wouldn’t that prove that the strategy is solid? That performance would include the spread (doesn’t it)?

Any pointers would be greatly appreciated.

Thanks and regards,

drolles

Author:  Dr. Anton Fokin [ Tue Jul 31, 2012 9:36 pm ]
Post subject:  Re: Paper vs Backtest and quote vs 1 min data (stat perf)

Hi,

if you mean OQ paper mode (not for example IB paper trading account), then OQ execution simulator uses market data that you pass to the strategy. Indeed results (fills) would be quite different if you use bars, trades or quotes for execution simulation. Execution simulation with quotes is quite close to real life, while simulation with trades doesn't take into account bid ask spread in many cases and simulation with bars is the worst possible approximation to real fills. Actually we have tested many strategies that performed extrimely well with trades but then pnl dropped like a rock when we switched to quotes.

This indeed depends on the nature of your strategy, time scale and frequency of trades.

Regards,
Anton

Author:  drolles [ Tue Aug 07, 2012 1:14 am ]
Post subject:  Re: Paper vs Backtest and quote vs 1 min data (stat perf)

Anton,

Thanks very much for your reply.

Yes, I’m referring to OQ’s Paper functionality. I’m currently trying to set-up paper trading with a Currenex broker to test the strategy. However, I’m having some challenges. Alexei has been extremely helpful in getting this resolved.

If OQ’s Paper Trading facility does use the price feed & thereby includes the spread (which is the sole cost of trading in FX) then I’m assuming the performance in the paper account was relatively robust?

I was running it against the live market using IQFeed on a group of FX instruments, they take their feed from FXCM.

Thanks and regards,

Dan

Author:  Dr. Anton Fokin [ Fri Aug 17, 2012 4:49 pm ]
Post subject:  Re: Paper vs Backtest and quote vs 1 min data (stat perf)

Well, yes, I guess if you test in the paper mode of OQ and use a good quote feed (and check that OQ simulator uses only quotes, not something else like trades or bars (i.e. FillOnBar/FillOnTrade = false), then you should be pretty close to reality, unless you are sending orders which can move the market in real life, but this is another story.

Regards,
Anton

Page 1 of 1 All times are UTC + 3 hours
Powered by phpBB® Forum Software © phpBB Group
https://www.phpbb.com/