I have now the first version in place, without money management and still without candlestick testing according to the LMT manual. I am still testing it. I hope to upload the EA by end of the week for further testing, probably including candles but still without money mgmt.
I had a few losses this week with LMT, mainly due to choppy market conditions. This disclosed the major wekaness of LMT, i.e. the trend is not tested with regard to its strength. The trend indicator is fine, but does not give any indication about its strength. So I will try to enhance LMT with other elements, e.g. with the Fx Ed strategy by Ed Ponsi (from his book Forex Patterns & Probabilities). He checks for the 10 day MA as a support over the last 10 days, for the 10 day MA > 20 day MA, 20 day MA > 50 day MA and 50 day MA > 200 day MA (reverse for short). I found that this system also gives good signals, with strong trends. Obviously there are not too may signals per year and pair (maybe 4-5), but either I extend the system to LMT (without the 200 day MA which is maybe too restrictive) or to all 28 pairs (of the 8 major currencies). This would probably give enough signals even if we subtract the flat pairs such as AUDNZD and EURCHF. I will see next week which way will be the best.
The thing that I find a little frustrating about LMT is that the trigger can be deceptive.
You may be on an uptrend, and after three red bars you see a green bar, so you enter the trade.... thing is, sometimes this green candle slides back and turns back into a red bar!
I think your EA will need to operate precisely on the close and opening of a new bar - every four hours on H4, and every day on the Dailies.
Perhaps the candlestick testing can be optimized per pair/time-chart. I guess that candlestick size that filters out trades that retrace is different per currency and time frame.
The trigger indicator is repainting as long as the candle stays open. The EA will check for new opens after the close of a candle, so this will not happen. But as LMT does not check for trend strength, it happens quite often that the next or over next bar will be down again. SO additional filters could help.
The EA will contain parameters which can be set for each pair/tf. But the size of false entries/candles currently is measured as percentage of ATR (as described in the manual). This could be changed if required.
Here is a first version of the EA, I did some basic tests. Feedback is highly appreciated.
Last edited by Rolf; 06-25-2009 at 17:37. Reason: Fixed a bug in opening trades based on moving averages.
For testing or trading the pairs based on moving averages, the following indicator gives an overview over the status of the 28 pairs. gray means neutral, red bearish, green bullish.
MA10: MA10 acts as support/resistance
MA20: MA10 > MA20 (or < for downtrends)
MA50: MA20 > MA50
MA200: MA50 > MA200
it looks like it cant be backtested, so only forward tests are possible ??
Backtests are possible as well. However, my own database only covers a few pairs, and not all 28 which could be traded theoretically. My own backtests give better results for the moving avg approach than for LMT, and here not all pairs are similoar. The most trending pairs like EURUSD, GBPUSD or GBPJPY are better than e.g. USDCHF. I think additional filters are required as the system should work also with others. For example, the MA approach triggered a buy for AUDCAD 2 days ago, which brought 200 pips so far.
then there was problem with my computer as it took too much time processing, I will try again to backtest it