Kevin_in_GA
4,599 posts
msg #115514
- Ignore Kevin_in_GA |
9/24/2013 1:25:50 PM
The values used to set these boundaries are from Wilson's work, but are by no means carved in stone. I'll run an optimization with a large number of possible variants against the entire S&P 500, from 2005 - 2010, then look at how well the best systems did against new OOS data from then until now.
As always, I'll make all of the results freely available here on SF and post the relevant code.
kevin
|
jimmyjazz
102 posts
msg #115522
- Ignore jimmyjazz |
9/24/2013 5:37:46 PM
Much appreciated, Kevin. Frankly, I find Wilson's writing a bit obtuse. It's a battle to get through his articles and separate the wheat from the chaff.
|
Kevin_in_GA
4,599 posts
msg #115524
- Ignore Kevin_in_GA |
9/24/2013 9:25:19 PM
Well, it seems that this approach looks great on a chart but actually backtests fairly poorly, regardless of the specific settings for NZU. I looked at several thousand variants and none really did well from 2005-2010. Worse than that, the top 5 failed to beat the S&P performance for the OOS period of 2010-2013. Why trade if it doesn't beat Buy and Hold?
Chalk this up to "fun coding it but don't trade it."
|
jimmyjazz
102 posts
msg #115528
- Ignore jimmyjazz |
9/25/2013 1:43:38 AM
Ha! I'll trust your results. Interesting. This kind of jibes with my intuition -- he promotes and obfuscates in equal measures. Make the readers feel they're stupid, sell books, profit. Yay.
|
davesaint86
725 posts
msg #115531
- Ignore davesaint86 |
9/25/2013 4:00:26 PM
Kevin,
I was wondering if you could run a backtest where buys cross above from the lower zone and sells crossing from above down through the upper zone. I think the results will be better.
|