i backtested a simple strategy (using 1000shares per trade) and the P&L is shown below. by looking at the pattern of P&L alone, what comments would you have? Daily PL 0 500 0 60 140 60 70 80 -80 200 480 130 -700 70 -50 0 70 360 -150 570 360 0 290 0 -60 330 150 -180 -120 50 70 290 120 0 0 0 0 400 0 0 180 0 0 70 0 130 70 0 0 0 -150 50 150 0 80 0 370 0 -30 0 300 0 70 0 0 0 80 500 80 0 0 0 210 0 210 90 380 0 0 0 730 60 80 0 0 0 230 720 90 290 430 70 0 0 0 0 100 0 210 0 0 240 370 0 -320 0 500 0 0 0 0 -390 0 60 0 60 160 0 140 0
The strategy trades only 68 days out of 120. I presume it does only 1 trade on most days when it actually trades at all. It has very high Sharpe ratio of about 7. Given the small number of trades I would assume the strategy is over-optimized.
thanks for the comment. Yes, when it does trade, it only does 1 or 2 trades a day (1000 shares per trade). Only one parameter is calculated from the sample data and other parameters are from outside of the data. The strategy itself does not use any optimization. I come up with the strategy by basic economic reasoning first without looking at any data, then used some historical data to calibrate one parameter and started backtesting. I backtested a 1-month period using out-of-sample data, similar performance. The % of winning trades is about 70%-80%.
I think at this point you should start testing it live but with a smaller amount of shares, say 100 or so till its proven profitable in the real markets.
I agree I think that is a good way to start out. I'm curious, what do you use to backtest your strategy? I've been coming up with some new strategies and want to be able to backtest them.
Please tell me how many years of data did you use? If it is only 120 days or so then this doesn't say anything. So starting with live trading, even if it is a very small number would be very stupid. I suggest you test at least from 2001. Then you will have a number of different market environments and you need that to have a valid strategy. Second are you 100% sure your backtest program is right and does the same thing you have in mind? Check every type of trade line by line if the prices are correct when it is trading and if the pnl is correct.
Congrats BayStCaptial! Very impressive. Have you had a chance to test this on some of the contracts like ES, NQ, YM, 6E, FDAX? If your strat works there, you're golden.
The main comment is, that backtests rarely have much value. Paper traders will often throw up different ideas and "backtest" them. But it is usually void of Trading Metrics (Sharpe, PF, DD, etc.), and they tend to cherry pick promising runs. They curve fit them. etc. A proper backtest runs over multiple market eras (bear, bull, choppy, sideways, etc.). It is multiyear, with many many trades. And the backtester has divided the past into at least an in and out of test period. And then they walk forward in the real market to get a real idea of how it works. Then of course, there is usibng real money testing it. THEN you might have something of a little value.