The ROC indicator has been doing well of late, which has of course generated some reader interest. I’ve ran various backtests of the indicator and will dump the statistics into this post.
Michael Stokes of MarketSci wrote a helpful piece about this indicator: Woodshedder’s Long-Term Indicator.
ROC means rate-of-change.
The ROC indicator works by going long when the ROC252 (252 day ROC) has been above the ROC5 (5 day ROC) for one day and will again close above the ROC5 and going short when the ROC5 has been above the ROC252 for one day and will again close above the ROC252.
All trades are made at the close and the account is assumed to have started with $10,000.
From 1.1.1994 – 7.14.2012
Equity Curve and Drawdowns:
Okay, so the indicator has done well in backtesting. How about in real-time?
I first wrote about this indicator on August 28th, 2011. Let’s see how it has performed in out-of-sample, real-time performance since that date.
From 8.29.2011 – 7.14.2012
Equity Curve and Drawdowns:
Real-time results show a net gain of 15.93%, which is annualized at 18.42%. Note that the short side has resulted in a negative average trade.
While the indicator is designed to catch long-term trends (note in the stats that a few LARGE winners provide the bulk of the gains), the indicator has been handling the whipsaws of the past year quite well.
A SPY buy-n-hold since 8.29.11 has yield a net profit of 11.86% and a max drawdown of -10.01%. The ROC indicator, in out-of-sample performance has yielded almost 4% more than buy-n-hold with a slightly lower maximum drawdown.
If we make the system long-only, real-time performance is even better.
Some of you have asked how the indicator works with individual stocks. I’ll look at that next. I will be happy to entertain any other questions in the comments section.
13 Responses to ROC Indicator – Statistics Dump, Backtested and Real-time Results
Very interesting. My binary strategy also tests out to have an improved performance when going long-only/cash.
Also interesting is how system performance appears to have a lot more variability in return and drawdowns since 2008.
Cool story about Tweedy and Uncle Tupelo last night, btw.
The variability/volatility I think is due to the performance increase/decrease of mean-reversion type trades over the past 4 years. Put simply, this indicator, when it gets whipsawed, is a mean-reversion indicator. At least that is my current theory!
As for Tweedy, when I was at their practice pad, he told me they were breaking up. They were getting ready to go on tour to support Anodyne. I was of course devastated but he gave me a business card and wrote on the back of it to let me backstage at the date for their show in my hometown, with his signature (the fuckers didn’t let me backstage, and they KEPT the business card when I handed it to security.)
I will write more about that meeting with Tweedy as it comes to me. I haven’t thought about it in some time. I remember two stacks of empty pizza boxes about 4 feet high each and a lot of empty diet coke cans.
I have a set list off the stage from the first Wilco tour. Coomer, the drummer was the only one that signed it for me. Of course he later sued Tweedy.
I didn’t get into them until much later, but hearing the Uncle Tupelo is breaking up from Jeff Tweedy had to be pretty surreal. I’d say that’s the horses mouth. I don’t know if it’s my favorite, but I probably listen to Anodyne more than any of their other albums.
I love all of them….but No Depression may be my favorite, with March 16-20 coming in 2nd. No Depression to me is the ultimate in cow punk. “Whiskey bottle over Jesus…not forever, just for now…”
Do you like Brian Henneman and the Bottlerockets? If so, find online the recording of Henneman playing with Farrar and Tweedy as “Coffee Creek” in St. Louis. They do some great fucking classic country covers. I think it is 3 discs. The recording is live from the soundboard.
I could also tell the story about when John Stirratt, after the breakup of UT, then playing with Tweedy and Wilco, fucked my best friend’s girlfriend (and now wife) after the show. It was surreal, the Wilco bus picking up Stirratt outside her apt. the next morning, navigating the tiny streets of downtown Charleston, S.C. We had all hung out after the show in a local bar called ACs. Stirratt was playing footsie under the table with my friends girlfriend. My friend got pissed off and walked home. Big mistake. Of course he still married her but when I told him what happened, he didn’t believe me and our relationship was never the same.
Yeah, the sound of No Depression is so much different from the later albums, which I got into first, so it took a long time to grow on me. It’s a lot like Slanted & Enchanted (if you are into Pavement) in that way. Some of the songs on March 16-20 are so heartbreaking.
Never heard of Henneman…will check him/them out.
Re: Stirratt, ouch. The lesson, as always: don’t leave your girl at a bar with someone in a popular band.
The Bottlerockets, Henneman’s group, has some produce some good stuff but it may not be hard enough for you. Definitely more alt. country, rockabilly, rock and roll, than anything else, imho.
The heartbreaking stuff on 16-20 imo is meant to echo/illustrate the life of the factory/manufacturing employee in the midwest factory belt.
As for Stirratt, I will never forget that. Which reminds me, his sister played bass for a band called Blue Mountain. Great band from Mississippi, and the guitarist does a great Neal Young. Saw him channel Young one time live while covering Cowgirl in the Sand. They are long since broken up, I believe, but if you want to pursue the cow punk genre, worth checking out. They had a minor pop hit back in the nineties call Blue Canoe.
We were in Belleville visiting mom’s side of the family over Christmas time. My uncle asks if I like Uncle Tupelo…of course I was a rabid fan, and knew they were from Belleville. So right then and there we get in the car and drive over to their practice pad. No phone call, no nothing. Being Steve Tweedy’s good friend, I guess he didn’t need to get permission from the younger brother Jeff to just drop in. And now, from Wikipedia….
“In January 1994, Farrar called manager Tony Margherita to inform him of his decision to leave the band.”
If my timeline is correct, and I believe it is, the news of their breakup was likely very fresh, which makes Jeff’s statement to me all that more historic.
All of my best performing systems/indicators have seen increased performance variability since 2008… its as if the nature of the stock market has changed, (perhaps due to HFT and/or other algorithmic computer based systems).
Do you have a system (some proprietary mixture of indicators) that you’ve found outperforms??
dyer, I agree, I’ve seen the same thing. I think equity curves that don’t show the variability have been curve fit in backtesting. I do believe the simplest explanation is that the nature of the market has changed.
I do not have any proprietary mixture of indicators that are immune to the variability.
Interesting that paired switching of SPY with VUSTX (every 13 weeks on Friday, by SPY if did better than VUSTX in the 12 week period ending on prior Friday, otherwise buy VUSTX) also returns 12.4% annually. Since 8/29/2011 this strategy has yielded 22%.
VUSTX is giving that system an extra kick. I could try having the ROC system rotate into VUSTX instead of going short. Might be interesting.
You can actually predict ahead of time when this system is going to cross. The formula for ROC(n) is close/close[n]-1, so your system is equivalent to
which is equivalent to
which is equivalent to
So you get a 5 day warning on when the system will cross. It doesn't depend on the close the day it happens. This is basically a ROC(247) crossover system with the trades delayed by 5 days. The five days provide time for a pullback to a more favorable trade price, since the initial crossover will tend to reverse temporarily. It could probably be improved by using a short term indicator to time the entry, instead of just having a static five day lag.