I backtested EminTrader's Strategy using 2 year's worth of 1 minute data, using my Wealth Lab testing environment. The data used was retrieved using TDAmeritrade's API facility (they only permit up to 2 years of intraday data to be retrived). The data looked pretty clean on brief inspection. ---------------- Assumptions: ---------------- (1) 1 Minute Data used from 5/1/2007 to 4/22/2009 (almost 2 years); (2) I assumed IB trading commission of $ 0.005/share; (3) Since I don't know how L5 is calculated precisely, I used the current values for each stock (specified by EminTrader) for the entire 2-year duration; (4) I assumed a fixed dollar amount of $ 5,000 per trade was made. Since the dollar amount per trade is fairly low, a rather high commission % is being assessed here. Trading a higher dollar amount per trade may result better profit; (5) I ran test for each of the 5 stocks that EminTrader has been tracking; (6) I only tested the LONG strategy, as originally specified; (7) I did not account for slippage; --------------------------- Here were my results: --------------------------- HCBK: L5 = 2.0992 % # Trades: 156 # Winning Trades: 82 % Winning Trades: 52.56 % Avg Profit/Loss: $ 3.58 Avg Profit/Loss %: 0.07 % Net Profit: $ 559.16 Net Profit %: 2.24 % MRVL: L5 = 2.9668 % # Trades: 160 # Winning Trades: 75 % Winning Trades: 46.88 % Avg Profit/Loss: $ 4.52 Avg Profit/Loss %: 0.09 % Net Profit: $ 723.67 Net Profit %: 2.89 % RIMM: L5 = 2.493 % # Trades: 210 # Winning Trades: 105 % Winning Trades: 50 % Avg Profit/Loss: $ 6.43 Avg Profit/Loss %: 0.13 % Net Profit: $ 1,349.95 Net Profit %: 5.40 % JNPR: L5 = 2.2844 % # Trades: 189 # Winning Trades: 94 % Winning Trades: 49.74 % Avg Profit/Loss: $ 3.63 Avg Profit/Loss %: 0.07 % Net Profit: $ 685.22 Net Profit %: 2.74 % SCHW: L5 = 2.2177 % # Trades: 194 # Winning Trades: 101 % Winning Trades: 52.06 % Avg Profit/Loss: $ 6.22 Avg Profit/Loss %: 0.12 % Net Profit: $ 1,206.29 Net Profit %: 4.83 % ======================= OVERALL $ GAIN: $ 4,524.29 ANNUALIZED $ GAIN: $ 2,262.15 OVERALL % Gain: 18.10 % ANNUALIZED GAIN: 9.05 % ======================= When I find some time this week, I am going to play around with some optimization stuff, using ATR, and other things.
I suspect the L5 values are being calculated using some volatility measure such as ATR. I've seen this thread for a few weeks now but never actually looked at it. It seems interesting. A long-only system is particularly interesting for trading in an IRA for instance.
I don't want to spoil the party here, since I think you're system is actually very good. There is also always the potential to improve a system by for example using principal component analysis vectors instead of the time series data. So these are just preliminary results. Only point I do want to make is that if you want us to take you serious, please stop talking about behavioural analysis. What you do is just some form of time series pattern recognition/rule based analysis. As long as you only use the data of the original time series, I cannot call this behavioural analysis. If you score the sentiment by reading the newspapers everyday, recording values, importing those into your system, or use other forms of recording real human behaviour, that could be a start. The fact that you backtest using a cripled systems as bad as PROTREALTIME proves that the latter is not the case, you cannot even import external time series data in this system. The real power in this system is the good use of money management. Using a portfolio of some stocks and being very picky with the signals.
Hi hayman, good job but sorry i don't see utility of your work with fixed L5% values while they are not fixed Hi, stevegee58 NO AGREE. Be patient.
I manually went through the charts and OHLC data for SKF from 03/12/09 till present . During that time SKF traded down from a 03/12/09 open of 173.98 to a 04/24/09 close of 56.70 . A loss of 67.4% . Applying emintraders system (long only) during that period yielded approx +4% gain.
heres the thing... 3 months ago he started this at 50%. Today hes at 133%. Almost, 3 times what he started with. So, is everyone saying hes on a hot streak? I am confused here..... A guy makes 80% plus... In the next 2 months, if he goes to +200% is there still something wrong or is he still on a hot streak? I would appreciate any input here.
Thats no good doing it that way. You would have to take a 1 minute chart and manually go back and see the moves throughout the day and heres why: You trigger in long at 50. Profit is 50.50. Stock goes to 49, stop loss hit. Then later on, goes upto 51. So in your calculations, you have a winner when in reality, you have about a break even trade. 1 trade stopped out and 1 winner. So, using soley OHLC is no good.
I actually did go through each days intraday chart , to make sure what you warned about did'nt happen . Sorry for any confusion caused .
That was only iteration # 1, EminTrader. Granted it doesn't mimic the fluctations in daily L5 values. My next task is to optimize on my current trading algorithm, to come up with an optimal solution. In fairness, my rather crude results did paint your system in a favorable light.....with optimization, it has the potential to become much better. I'll keep you posted on progress.