Quote from B.Willis:
Neoticker would need almost 3 minutes to backtest on a single stock 10 years of daily data.
This can´t be correct.
Dear Mr. Willis,
Neoticker would not need as much as 3 min. to backtest 10 years of daily data. Below, are the calculations showing how:
1. You have considered backtesting 10 years of daily data, which gives us ~ 2500 bars (there are about 2500 trading days in 10 years, that is , 2500 daily bars)
2. In our optimization we used 170,000 bars. This is 68 times more than what you used: 170,000/2500 = 68.
3. In our tests, we used 378 simulations. Backtesting involves 1 simulation only.
4. Summarizing the above we have:
68 times more bars in our test than in yours
378 our simulations vs. your 1 simulation
11,740 seconds - our test results
Which yields:
5. 11,740 sec/ 68 / 378 = 0.4567
That means that a single simulation for 2500 bars takes ~ 0.5 sec.
Your calculations might have involved one of the following erros:
1. 378 simulations might have been overlooked.
2. optimization and backtesting are two different things. Backtesting involves a single simulation, while optimization involves multiple simulations - each for a particular set of parameters. So when you compare backtesting and optimization the number of simulations must be taken into consideration.
