Source of "Clean" EOD Data?

Discussion in 'Data Sets and Feeds' started by HotTip, Feb 9, 2011.

  1. HotTip


    I'm currently backtesting an equity trading strategy that is based on the high/low of EOD data to determine whether or not I get into a trade. However, there are tons of examples where just looking at the OHLC for the day I'm led to believe that it was a volatile day, but when I look at the chart, I find that the high or low was the result of an errant trade of just a few hundred shares. It's a tick spike that's obviously erroneous on an intraday chart, but not at all obvious on the OHLC data.

    So, instead of having to examine every intraday chart of every backtested trade (I have over a thousand of them) is there a data provider who sells EOD stock data cleaned up of these spikes?

    Thanks for any help!
  2. Bob111


    it is impossible to backtest anything accurately on daily OHLC.
    i'm storing tick data every single day for my stocks,that i'm trading. even based on tick data my real and backested results are different. in fact the difference can be as wide as 50%+,if you trade something less than 200K shares a day on average volume. and i count trade as valid in my backtest if there is at least 3 trades @ my size and price are traded. some 100 shares lot is not enough for me.thousand trades imo-no big deal to backtest,even on tick data

    this is the only way,if you want accurate results.
  3. HotTip


    Thanks Bob. Yeah, I'm thinking I might have to just plow through it as well.
  4. Agrd w/ above. If you want help getting TAQ data PM me.
  5. Hard to find clean data. You will be amazed how many vendors provide data where even the close is outsise the high-low range. However, you can never know if a low is due to a faulty spike or due to an actual spike.

    I believe Metastock in the past had an option in the downloader to check for faulty data. I use a program to find patterns that checks for faulty data. Very useful thing. I am more interested in cases where the open or low is outside the high-low range. Should not be hard to write a small VB program to do that.
  6. has good EOD data across most asset classses
  7. Using the "=N" function on qcharts or esignal for NYSE stocks will really help clean everything up. It only gives the prints on the NYSE exchange.

    Another solution is to write a function that says a price is only included if enough volume trades at that particular price.
  8. lindq


    I base a lot of my backtesting on daily lows, and lows are always suspect on daily bars.

    A somewhat helpful solution I've developed is to run my backtests on X portfolio for X time period, then create a list those stocks that signaled based on my rules.

    I'll then download 5 min data for only those stocks, and run another backtest using only the close of those 5 min bars, thereby avoiding the spiked EOD low or EOD high on daily bars.

    This is not at all a perfect solution, but it gives me a general idea of the soundness of the strategy before looking further.
  9. HotTip


    All great suggestions. Thanks everyone!
  10. spindr0


    Yeh, the TEST feature is still in the Downloader. They used to have a DOS stand alone utility called the Metastock Utility (MSU) which did TEST, RANK, ALPHABETIZE, CONVERT, etc. I don't know if they provide a Windows version now.
    #10     Feb 9, 2011