backtesting is stupid
- Thread starter jonp
- Start date
I have nothing better to do with my time than wait for my inevitable death, passing the time until then with interminable backtests, so I'll debate you. Your move.
I'll be here, if I don't die first. Will it take you that long for you to devise your opening move? I kind of feel sorry for you already, you being at an intellectual disadvantage and all. I have to warn you that I am no dummy. I spent three years mastering the ninth grade. Got it down solid. I'm going to go take my old man nap now so I'll be sharp for your killer next post.
Here's an example of how backtesting can be useful.
Why spend years trying to decipher Jack Hershey's "teachings" when you can easily test those aspects of his "methods" that are quantifiable and see that they're unprofitable?
For example, Jack unsuccessfully tried to impose static parameters on stocks in "Tomorrow's Newspaper Today" (attached). I tested buying the "0 to 7 turn" of Jack's "P,V relation" per page 8 of his paper on 1000 stocks from 2000 to 2005 -- a total of 5000 stock-years -- using spydertrader's code for the scoring and exiting 5 days later and got the equity curve below.
Why spend years trying to decipher Jack Hershey's "teachings" when you can easily test those aspects of his "methods" that are quantifiable and see that they're unprofitable?
For example, Jack unsuccessfully tried to impose static parameters on stocks in "Tomorrow's Newspaper Today" (attached). I tested buying the "0 to 7 turn" of Jack's "P,V relation" per page 8 of his paper on 1000 stocks from 2000 to 2005 -- a total of 5000 stock-years -- using spydertrader's code for the scoring and exiting 5 days later and got the equity curve below.
Well let's say I have an idea about an edge. It seems to work when I look at a chart during the day. But, am I really seeing what is happening or what I expect to happen?
The ability to code up the idea and run it on past market action will, if I properly coded my idea, show whether the edge existed in the past. This assumes proper allowance for slippage, etc., etc.
If the backtest fails, my edge was illusory as my brain was mis-interpreting the chart. If it shows a good win percentage or whatever metric I'm looking for, it confirms my idea. However, backtesting does not guarantee that the edge still exists. I think any discussion should address how the market constantly changes and how to handle the change.
The ability to code up the idea and run it on past market action will, if I properly coded my idea, show whether the edge existed in the past. This assumes proper allowance for slippage, etc., etc.
If the backtest fails, my edge was illusory as my brain was mis-interpreting the chart. If it shows a good win percentage or whatever metric I'm looking for, it confirms my idea. However, backtesting does not guarantee that the edge still exists. I think any discussion should address how the market constantly changes and how to handle the change.
Brilliantly put. Memory fools fools' memories.