Completely Different Results in Backtesting @ wealthlab and @!!!

Discussion in 'Strategy Development' started by nighttrader_wal, Apr 15, 2006.

  1. Good afternoon!

    I work with wealthlab and have developed a trading strategy for the DJI.

    I refined it a several times and now it runs very good - on the paper yet.
    The backtesting results for 8 years are:
    + 900.000 $, max drawdown 16.000 $ and profitable trades : around 80%.
    First I did not believe these good results, even because it is a very simple strategy that buys / sells using a limit order and closes the position at the end of day.

    I looked at the chart that the wealthlab software draws and could not find any technical errors that could have resulted from false programming.

    Here's my Problem: Before trading with real money, I tried to backtest the strategy manually at, a german website with a good charting tool (for free).
    For example in wealthlab there was the daily high @ 11000, and in it was @ 11030. Another day, there was a black hammer instead of a white doji at tradesignal.

    I looked after the problem but could not find an answer. I did not change the DJI with its future or something like this. I just have two different data sources , it seems.

    My trading system runs not that good @ tradesignal, of course, because it is a intraday strategy (basing on a limit order to open and an order for the closing price of the day to close).

    ===>>> Which data can i trust in? Which one is wrong? Could the reason be that tradesignal is a german site with a chart that is convertetd from $ to EUR?? (althougt there is USD written at the left side). What should I do?? The rrading system runs super in wealthlab, but does it because of wrong data?
  2. Get another source of data. If you're using Yahoo data you'll get nailed...their data is practically made up. Sign up for a free trial with QCharts, ESignal, etc and get some better data (not that it's all that wonderful from those vendors either, but it's better than Yahoo).

  3. Have you at least tested your system on something that is tradeable? It makes no sense to spend all this time optimizing on an index, only to find that you cannot reproduce it using DIA, or YM, or whatever. I can't tell you the number of systems I have seen with amazing results using SPX that cannot be reproduced using SPY or any derivative.
  4. Spelunk


    The DJI is a particularly poor index when you ask for cheap data. I've read that Yahoo data for instance (at least in the past) was known for calculating it at the end of the day. They just looked at all the DJI components, find their highs for the day and used that to calculate the high for the DJI not caring that they may not all be at their highs at the same time. So it's not based on the real world but fantasy.
  5. Agree.

    Although I only manually back test my pattern signals along with real trading small size testing...

    It's not uncommon for me to get pattern signals in the DJIA and not see them in the DIA (ETF) nor YM (futures) or vice versa.

    The pattern signals in either DIA or YM carry more weight with me because this is where the actual trading occurs (not the index) although I will trade YM or DIA that doesn't have a pattern signal while there is a pattern signal in the DJIA due to the fact the overall price action and its reasons are the same (DJIA, YM, DIA) along with being much more important than the pattern signal itself.

  6. I$land


    Unreliable data... I had the same problem a while back. Check out the post I started back them :

    I finally decided to go with Metasock Real-time with QuoteCenter because they have Reuters data.

    Like you, I develop my trading systems in Wealth-Lab but I import the Reuters data into it. Unfortunately, the ASCII format they use is not standard and you cannot import the data directly into Wealh-Lab. I had to write a file converter (C++) so I could then import the data files.

    It's not the fastest route to get good data but it's the best one I have found so far.

  7. hektor


    The tradesignal data are from Reuters, so they should be ok.

    In the newest version TS5 you can import your own data, so you could try to import your wealth-lab data into tradesignal and compare the results with WL for the same datafile.
  8. I just downloaded data from QAccess, using Ticker SEarch.

    It has the same good results as before in wealth lab.

    Still, I dont have something to trade my strategy with. I think, the dow future will be good for this, but will differ a little bit from the DJI itself. Do you have an idea where to get data from the DOW Future - if possible it shouldnt cost something, because I usually dont spend much money in this, because I dont trade yet, but will begin as soon as i know the system will work - otherwise I wont. So, a price of 150 $ , that they usually take for this, is too expensive for me.

    best regards
  9. But the results from QAccess are still good - still not the same as in tradesignal, where the data is from reuters.

    Does it mean that i can forget qaccess?
  10. Does tradesignal have a way to export the data? If you can export it to an ASCII file you can then run it in Wealthlab and see.

    Also are there slippage settings in tradesignal? There definitely are in wealthlab, but are they turned on? What about commissions?

    #10     Apr 16, 2006