EminTrader, A couple of obvious problems here: 1) Open Prices for ProRealTime obviously include some pre-market data here. You need to use a more reliable data source for OHLC; 2) As another poster pointed out, you need to ideally backtest with TIC data, but I'm sure you will get 95 % + reliability with 1-min bars. Backtesting this algorithm using OHLC data will certainly be very distorted, especially during volatile trading days; 3) Your stipulated L5 values are not posted correctly in many instances. I have spreadsheeted all of HCBL and some of MRVL, and there are obvious posting problems (I state these just for the benefit of anyone who is trading this system): a) MRVL - starting on 2/24/09, you have a % gain of 2.7273 %, and you've hit buy and target prices. Yet you state that L5 is 2.8037; b) HCBK - starting on 3/12/09, you have % gain as 2.0561 %, and you've hit buy and target prices. Yet you have L5 stated as 2.0992; c) The list goes on and on.....many transcribed errors. Doesn't your spreadsheet for these mock trades calculate L5 for you ???????? If you need some 1-min data for backtesting, please PM me.
The real problems with this system will start when the markets go really volatile...hitting stoplosses and triggering random trades.
Been tracking FSLR since 3/12/09 approx 14% gain Bracket trader seems to work with this system , has a flatten on close feature in case of neither bracketed orders being hit .
Results : HCBK : - 1.499% MRVL : RIMM : JNPR : SCHW : - 1.21% 2009-04-29 : HCBK : L5 = 2,0992% MRVL: L5 = 2,6144% RIMM : L5 = 2,493% JNPR : L5 = 2,2844% SCHW : L5 = 2,2117% Have good trades... Overall %result : + 126,861% FSLR: L5 = 2,8512% SKF : L5 = 4,4556%
Sometimes, my L5% values can change between time T and T+1 only because of prorealtime datas...I don't understand why (i asked them but no response). So i do with that in my trading ; differences are weak
How do you calculate the overall % result? I notice Monday 04-27-09 05:40 PM you report an overall % result of + 129,85%. I calculate 129.85 - 1.499 -1.21 = 127.141 You calculate 126.861.
I calculate two buy signals today, HCBK at $ 12.58 / share and SCHW at $ 17.84 / share. I attach both stock charts. I do not calculate any sell signals during the day so I sell both positions at the close. I calculate a buy signal for HCBK appears about 10:30 AM USA Eastern Daylight Savings Time (EDST) and the buy signal for SCHW appears about 12:30 EDST. You write (04-25-09 08:43 AM) : "no specific money management rules (i use all my capital at each order)". If I use all my capital to purchase the first stock then there is no capital available to purchase the second stock. I suspect your overall % result values are not showing the actual experience of a trader using this method. Not all the trades can actually be executed. The daily profit or loss may depend on if the first trade is a big winner or loser, not the total result of all trades shown.
This system has not been backtested properly. Long-term backtesting cannot be accomplished with daily values, stated fluctations of intraday of L5 cannot be happening, clean data and proper OHLC values from a reliable datasource are required, proper money management rules need to be assumed and applied, and the list goes on. I've come up with a reasonable, but not exact, facsimile of L5, and testing under 2 years' worth of 1-min bars for the 5 stocks results in an overall loss of close to 10 %. If EminTrader wants to prove me wrong on this, he can either: 1) Supply me with L5 formula, which I will apply to my backtesting algorithm....I will supply WL code to verify that we are on the same page with algorithm; 2) I can furnish EminTrader with 1 min data for the 5 stocks, for 2 years (sorry, that's all I have), and he can migrate his algorithm to a real backtesting platform (WL, Tradestation, etc.), and generate results for himself. Other than that, way too many holes, assumptions, and mistakes to make this thread of value any more. It has been a fun exercise to test though.