Evaluation of system stats anyone?

Discussion in 'Automated Trading' started by sluzbenik, Oct 23, 2005.

  1. Could anyone take a look at these two systems and have a look? One has an additional filter that is cutting profit, but does do a good job of staying out of some hairy times. Check out the equity curves. The system trades a basket of 1400 US equities with sufficient liquidty (not very strict actually since I'm a little fish.)

    I am getting a bit better at broad market analysis and am tempted to switch between them depending on my outlook for the broad markets.

    I will not discuss the core of the system, but I will talk about the money management, which is nothing more than an ATR ratchet max risk stop at 2%, plus a cap on positions as a percentage of equity. The system does a bit better with just a risk-based stop at 1% but I'm not comfortable with that. It does look like it might work better with just a percentage of equity, but I've seen risk-based stops work better quite often so I'm probably going with that. Returns are also better without the equity cap, but gap downs happen. The system does not use margin.

    Thank you.
  2. sluz,

    Nice system. 80% annual return in a bear market is awesome. Especially without leverage.

    1. Could you have similar results with a system that trades for holding periods of say 5-10 X your average days held of 10.5? Effectively, it would become a long term system.

    2. what is system 2? it seems the same except less drawdown.


  3. System2 is the same, just with an extra filter on entries.

    I couldn't change the time frame it uses. Why?

    We'll see how she forward tests, if the markets go where I think it's going it'll give it a good kick in the pants before the filters kick in. I have tried very long and hard, and getting out ahead of a bear is very tough. The best I can do is get out a month or so after it's started. Then the equity curve flattens out after drawdown, which is what is happening now.
  4. I would prefer to trade the second system as well. Since this is a WLD backtest I would suggest you make sure things like position sizing and total possible trades vs actual trades taken are under control -- mostly I say this since you are using 1400 candidates. Those two variables can take a very nice backtested curve and turn it into a trashy real performance.


  5. Ya i tried using a signal on an index before entry and forcing exits on my system during a correction or crash, but it doesnt help much. Its more a token thing than anything. I dont think theres a EOD system out there that if it existed in '87 didnt lose 30% in two days.

    The reason I asked about lengthening the timeframe is to see if its still profitable using longer term signals ( but maybe you you price bars, and therefore this isnt possible).
  6. well, actually I have, but results over longer time frames things are much more difficult. The actual character of the stock itself comes into play, but if you are swing trading, it doesn't really matter what it is as long as it's not in a downtrend.

    I backtested it just on the Naz 100 last night and it was profitable but not so much so. Too efficient, too many traders making trading difficult, I think. Much better on the S&P500, there are a lot of stocks people just don't give a crap about in there.