Sunday, June 3, 2012

CCI Stats and Profit Curves

Stats for the new(ish) CCI system are pretty much complete. I will be posting them over the next few days.

Updated 6.22.12

Here are the CCI Stats for SPY.

Compound Annual Growth Rate: 26.55%
Profit Factor (2008-Present): 1.52
Win/Loss record: 123 wins, 52 losses (70.90%)
Average return per Trade: 0.54%
Median return per Trade: 0.96%
Top 3 Max Gains: 17.95%, 10.17%, 8.50%
Top 3 Worst Losses: -22.69%, -12.71%, -11.77%

Profit curves section:
I should mention up front that the system performed blah in 2009 for some reason. I need to examine that further to see why and possibly find a fix. I mentioned the other day I was going to look into partial positions to account for trends, but that was blah, too. Too much whipsawing. I did stumble onto something, but that's for another post.

The system has 218 trades since mid 2008. It comes out to about 60 trades per year. Just over 1 per week, so this is a nice & easy swing trading system which requires little time and effort to follow. It favors mean-reversion and only very short term trends. The most recent short trade was a great example. Strong, steady, powerful moves in 1 direction are definitely its Achilles' heel.

If you can't stomach big swings, please avoid using leverage! :) You'll see what I mean when looking at the TNA curve. SPY offers a nice steady gain over time.

That's all for right now...Like I said, more coming soon... Happy Sunday!


  1. J, thanks for the stats. Two comments/suggestions:

    1. You really need to run that back-testing longer, though. Your entire sampling period has been a time of financial turmoil and irregularity where the markets have been largely forced by liquidity injections. How about going back to 1997 for 15 years, and multiple bull/bear mkts? 4 years is really not that much. I presume you're running this on something like eSignal, so, going back 10+years is just a clickie.

    2.) I'm a bit surprised by that profit curve (on either TNA or SPY). For TNA, nearly all the profits are in the last 30% of the trades, after roughly 2.5 years of no net account gain. For SPY, from trades 25 to 150, there is again, essentially no net gain, as well. So, put yourself into this system back in time. Pretend you do not know the future. You start trading this system in 2008… At first, you make a little money… then you get into a rut of 6 months, then 12 months, then 18 months, then TWO YEARS (!!) of no net gain…. Or if you’re trading the SPY, imagine having 125 trades in a row with no net account gain. After all that abuse, do you honestly think you're going to still have the confidence in a system that generates zero net return for so long that you will still be sticking with it to be able to benefit from the blast-off that occurs in the last stretch?

    1. 1) No eSignal, just Excel.

      2) Agreed. 2009 was not very kind to this system. In fact, that was the motivation for consolidating the CCI with the original JTrader system. They work much better as a team, as evidenced by the vastly superior SPY profit curve (of the Composite System).

  2. J, first, thanks for the profit curves. And JKH, thanks for the excellent comments.

    I agree with JKH and only want to add one thing. J, you say that you want to go back to 2009, and see if you can fix the flatness of the profit curve. I urge you to be cautious when doing this, because this sounds like a classic case of curve-fitting. A profit curve can always be fixed retroacively by further optimization, but, paradoxically, such fixes are usually bad for the system in the long run. You probably already know this, J, and are taking appropriate measures, but I thought I'd say it anyway.

    1. Yup, I hear you loud and clear. Rather than curve fit, I think I did the right thing by just merging the 2 systems. Results were better than I expected, and certainly better than either model on its own.