My auto day-trade strategy shows good profits in backtesting but is the opposite in live trading and I need help to find out if I have done something wrong one way or another. With ten times leverage, the strategy trading a portfolio of stock CFDs produces a return of over 500% in 6 months, so I think it is worth the while for interested parties to help explore this a bit. I would either share the code of the strategy or to compensate in other forms with those who help to look into the matter.
My auto day-trade strategy is to sell at double tops and buy at double bottoms, with stop orders to take losses and the reversal signals of Parabolic SAR to take profits. The code is written in EasyLanguage. Portfolio Trader in MultiCharts is the trading platform, with data supplied by IQFeed and Interactive Brokers serving as the broker. I expected the strategy to be profitable, which it is, in backtesting. Right now I have tested a portfolio of US stock CFDs with the strategy. The strategy can be used to trade also a mixed portfolio including FOREX, stock indexes and commodity futures, or one can trade these instruments in separate portfolios.
For me, out of a pool of over 900 US stocks (from the component stocks in S&P 500, S&P 400 Mid-Cap and NASDAQ 100 indexes), I have selected a portfolio of 149 stocks, based on 2 selection criteria: Profit Factor of 3 or above, and profit % of 35% or above, through genetic optimization in MultiCharts. I did not try to over-optimize or did anything like that. I just ran the stocks through the strategy and took those that fit the 2 criteria and discarded those that didnt.
Understandably, different time frames produce different numbers of stocks that fit the criteria. For the 6-month time frame, I got 149 stocks; 3-month, 104 stocks; and 2-month, 76 stocks. The three portfolios all did similarly well in backtesting, with the equity curves trending upward uniformly. Because stock CFDs generally carry ten times leverage, the return was over 500% in 6 months for the 149 CFD portfolio; 140% in 3 months for the 104 CFD portfolio; and 220% in 2 months for the 76 CFD portfolio.
One thing is that there are not too many trades with this day-trade strategy. For the 149 CFD portfolio, it averaged about 25 trades per day; the 76 CFD portfolio averaged about 13 trades.
The strategy looks good in different time frames as well as with different portfolios, but once it is tested in live trading, the result is the opposite. As far as fills are concerned, about 95% of backtesting trades match those of live trading. The entry prices for both backtesting and live trading trades are matched quite competitively. The fills for exit trades are mostly worse in live trading, but within reasonable range.
Somehow, though, in backtesting, hardly were there two losing days in a row, but in live trading, almost all days were losing days for the past 15 trading days. Another thing is, in backtesting, the daily profit % for trades is between 40% to over 50%, but in live trading, it is between 30% to 0% (some days all trades were losing trades).
Other than luck, where else should I begin to look for bugs? Any help would be most appreciated. Thanks!