Backtesting expectation

Discussion in 'Strategy Building' started by Engine99, Nov 19, 2007.

  1. This is my first post here, so please don't bash the thread right away if this is a really stupid question.

    Over the last couple month I've tried multiple strategies (from books, wealthlab, internet) but when I take the strategy and add slippage (2 cents),commisions (9$ per trade) and spreads (depends on the symbol 1-15 cents) and test them with different stock symbols (all symbols > 1m shares volume) I never come out ahead.

    They work great for a subset of stocks, for a year, two years but as soon as I backtest them to 1999 and expand the range of stocks, they all fail.

    What am I missing?

    I'm almost back to the random walk theory.

    http://www.amazon.com/Random-Walk-Down-Wall-Street/dp/0393062457
     
  2. (1) Assume >>> 90% random walk.

    (2) Flat cost/trade sub-optimal in today's fragmented market.

    (3) If you are such a poor trader that you have bid-ask "slippage"... quit right now.
    You must profit from the spread to be competitive...
    That means > 90% limit orders.

    (4) Focus on a well-defined niche... and become an expert in that niche.
    That means stay away from the high-profile, liquid stuff that every loser is trying to trade.

    (5) You MUST have solid profit margins... at least 30-40-50% after costs...
    Unless you are a high volume Bot.

    (6) Many strategies will work for a few years...
    And then will no longer be viable due to changing market conditions.
     
  3. You can always open an account with AG Edwards and maybe try the systems on pork bellies, OJ, or feeder cattle.

    You want a system that gives no more than 1/2% total drawdown. That was my goal and ninja gave it to me with a little work and lots of studying.

    The system sucks on old stocks...stocks get stagnant when they are old, you need new stocks to test with new systems.
     
  4. HoundDogOne has given some very solid advice there.

    All I would add is - try not to get too disheartened. It's a cliche but the markets really are very hard to beat. The fact that you have started out testing ideas rather than just trading puts you ahead of many on the curve.

    Also, try to think outside the box a bit. 90% of what I have read in books has proved to be garbage when I back-tested it. Another 9% or so didn't work but gave me the inspiration to try a few things that did have value. I've left another 1% as I am sure there is a gem of knowledge in a book out there somewhere....
     
  5. lindq

    lindq

    From your post I assume you're backtesting with daily data.

    It's very unlikely that you're going to find a system that works consistently in every market using only daily data. Over the past 5 years I've probably thrown away 10-15 systems that backtested wonderfully on daily data, traded nicely, then stopped working.

    It wasn't until I began to access and backtest intraday data that I became consistently profitable. And that is my suggestion to you. Even if you are looking to hold for a few days, or longer, access to intraday data will help you to better define your entry points and make necessary adjustments. You'll also have multiple timeframes to work with, and that's a key to trading equities. (What's happening daily? What's happening intraday?)

    And I have to take exception to the comment that you need to trade with limit orders. It all depends on the strategy, and for mine, limit orders are not helpful at all.

    Also understand that backtesting, and developing a strategy from backtesting, will be most helpful if you think of it as getting you in the general ballpark, and helping you to appreciate the logic of what's happening. It is NOT an end in itself.

    To become consistently profitable, you'll need a LOT of real-time experience to begin to truly understand price action, and how to work with it.

    If you are a couple months into this process, you've only just begun.
     
  6. Agreed.
     
  7. A solid post overall, and I agree with most of what is being said here.

    Number 4 especially.

    But i must disagree with point number 3. But I think this depends for the type of system you are trading. With long term trend following systems, like the weekly system I am trading, I just buy market price.

    William O'neill said:

    Novice investors like to put price limits on their buy-and-sell orders. They rarely place market orders. This procedure is poor because the investor is quibbling for eighths and quarters of a point, rather than emphasizing the more important and larger overall movement. Limit orders eventually result in your completely missing the market and not getting out of stocks that should be sold to avoid substantial losses.

    http://members.aon.at/tips/citation.html

    Cheers.

    Nizar.