Tuesday, May 31, 2011

It's hard to ignore (pending) green ... / Connors' TPS Strategy & EdgeRater


  • We're seeing some thawing of the bear ice regarding the pricing model slopes, e.g., the 5d EMA is above the 8d and the 8d is above the 13d.  This is bullish on a short term basis only.
  • The remainder of the longer pricing model EMAs, e.g., those longer than 13d, are all negative.  We have to watch the "slope of the slope" to see if these are going to transition positive or remain negative.
  • The VTI short-term timer, which is a whole-market short term timer, fired LONG with the close of business Friday, May 23rd.  I have placed a buy stop loss at $69.38 due to the potential gap up this morning (e.g., it will fill if the gap fills).
  • The Intermediate-Termed Elder Force Index timer is in CASH, and given the numbers, is most likely there for another day or two.
  • The Long-Cash Ratio model slopes are thawing regarding bear ice, e.g., the 5d and 8d slopes *just* moved positive on Friday.  This indicates that on a short-term basis that more stocks are flowing to the "LONG" side than those moving to the "CASH" side.  This is positive for us -- we want to invest when the number of stocks that are outperforming historical averages are increasing, not increasing.
Conclusion:  We are approaching a new bull leg on a short-term basis.  Many indicators are still in bearish territory, but if you're nimble, you should be able to move in and out quickly and make a profit.  Note that on an intermediate-termed basis you should be in CASH.


Slope Models

The GGT Pricing Slope model is becoming more bullish:

As with all my images, right-click then open/save as desired.

Volume is in the toilet, which simply means that either money is sitting on the sidelines or it's being obligated in the options/futures markets, and not in the general equity markets.  

Looking above at the left side of the presentation, we see two consecutive days of the 5d EMA and the 8d EMA showing positive values.  This means that these two indicators have moved upward into positive territory -- they are gaining in $/day on these time frames.

A look to the right side of the presentation shows why the left side is doing what it is.  Out of the last 9 trading days, 7 of these days have had an upward-pointing slope.  It should not be a surprise to you that we're seeing some form of bullish signal as this occurs.

What is most important here is that we experience some form of follow through.

Here's the Long-Cash Ratio dashboard:

Again, much of the same presentation as above.

What is encouraging here is that we now have the LCR system indicating that on a short-term basis -- 5d and 8d -- that we are experiencing a net inflow of stocks with a LONG recommendation to those with a CASH recommendation.  Today's action is sure to add more green to the left side of the presentation, increasing the universe of stocks we can choose from.

Conclusions:  I'm dipping my toes into the bullish waters, as it appears that we can do so with some partial wind at our backs ...


ETF Trading Bandit and EdgeRater

One of my GGT colleagues in the Dallas area, Bob Wilson, poked me to re-evaluate / use some software that I've had which has been written by Chris White.  ETF Trading Bandit is a package that was originally written with Larry Connor's strategies in mind, but any list can be processed using the package and what results are recommendations based on mean reversion strategies.  While the "Bandit" is based on daily closing prices, EdgeRater is a backtesting system, and it's usefulness is growing within my cadre of tools.

One of the first things I did was set up EdgeRater to backtest Larry Connor's published strategies in his book.    I'm fond of a strategy called "TPS", which is an abbreviation for a strategy that means Time-Price-Scale In.  The basic philosophy is to first invest a 10% position in the equity meeting the requirements, then 20% on the subsequent day if conditions are met, 30% on the next, and finally, 40% on the last day.  Of course you could go further, but then you're exposure would be greater than 100% of a position size.

Connors did his work with unleveraged long ETFs; here are his results:

You see that on average, you're pocketing 1.45% on each trade over 6 days, and 89% of the time the trade works in your favor.  Note that this was buying/selling AT THE CLOSE on the day of the signal.

Those are published results -- if EdgeRater can do what it says, it should be able to duplicate these results.

After setting up the program (rather easy if you watch Chris' videos), here are the results I attained:

So in the compare/contrast, here is what we have:
  1. Connors listed 1159 trades, EdgeRater listed 1071.  The difference in statistical significance is negligible.
  2. Connors listed 88.78% winners, EdgeRater found 87.86%.  Close enough.
  3. Connors listed 1.45% average gain per trade, EdgeRater found 1.46% per trade.  Again, close enough.
At first glance, we have excellent correlation, at least for published data through 12/31/08.

One thing that should catch your eye:  the average profit per winning trade is $56.59 on an amount invested of $10,000 (the $10K isn't shown).  This isn't a great deal of profit per trade, and the amount that you take home will be very sensitive to commissions.  It is absolutely critical that you trade Connors' strategies in a low-commission account!  

One very important graph that EdgeRater provides is the statistical distribution of the results.  This graph can give you confidence at the repeatability of the strategy, as well as giving you some view of the significance of the outliers (the tails -- how bad or good can it get over the backtested period).

Here's the chart:

  • The x-axis is the % gain/loss on 1% increments.
  • The y-axis is the number of occurrences of the strategy within a given bar interval (here, 1% bars). 
  • The "S" curve is the normal (bell-curve) probability distribution of the sample -- we would expect that at 50% (y-axis, right side) that we would have the largest "n" occurrences of our data.
  • To have an "edge", we want the 50% S-curve to be as far to the right as possible.  In the upper right corner of the graph, the average of the dataset is +1.45%.
  • More subtle, but equally important, is the other data in the upper right corner of the graph.  Here, we see that the median is 1.4545 (%), and that the standard deviation is 1.9951 (%).  When the mean and the median are almost equal, this means the dataset is "well behaved" and we can have confidence in the strategy.  The standard deviation means that we have a 66% chance of the results of a given strategy falling between 1.4506 - 1.9915 = -0.54% and 1.4506 + 1.9915 = 3.442%, and since more lies on the positive side than negative side, we can expect positive returns with the strategy.
Another thing to test is the ability to invest end-of-day, say in the last 5 minutes of the trading day, vs. at the open the next morning.  Connors' published works all are end-of-day entry/exits, which is difficult for many people with day jobs.

To test this, EdgeRater allows to choose "Event Bar on Close" (end of day entry/exit) or "Next Bar on Open" (next morning at the open).  The results are as follows:

Here, we see a drop of nearly 5% overall in Net Profit %, or a reduction in -0.18% per trade in the Average Profit per trade.  This is significant, and indicates that we should attempt to enter the trade at the closing price OR BETTER (even if it is the next morning).  This can be accomplished with a LIMIT order that can be placed in the overnight hours.

Having established that we have good correlation from the beginning of Connor's time to 12/31/08, what about since then?  EdgeRater makes easy play of this effort:

Here is how I interpret the above table:
  1. There have been a total of 388 trades since 1/1/09, and nearly 94% of them have been profitable.
  2. Average profit is 1.73%.  Part of this can be attributed to the bull that we've been experiencing since 3/09.
  3. Note that Net Profit % would be less than shown because of the impact of commissions.
EdgeRater give you more info than a table; here's the equity curve:

The top graph is the equity curve, starting at $100K on 1/1/09.  The bottom part of the graph is the required open positions within the TPS strategy to achieve the equity curve shown above.  It should be of interest to you that the period leading up into May 2009 has little to no trades using the strategy -- Connor's has a built-in safety mechanism that the equity has to be trading above the 200d MA in order for the strategy to be valid.

While it may be hard to see, around 10/23/09 we had ~18 positions open.  This is a conundrum -- if our starting equity was $100K and I was investing $10K/position, how can I commit $180K?  Obviously, you cannot without the use of margin.  I'll research this more with Chris White, the creator of EdgeRater, and will get back to you on this.

What this does point out is that you will need simultaneous positions opened in order to achieve the theoretical  equity curve due to past performance.  The results above, with 20 ETFs, there were market situations where they *all* (or the vast majority) could be invested.  This suggests that on a $100K portfolio that your maximum position size is $5K, not $10K, and as a result, the amount of capital deployed and put to use to generate income is far less than what we see above.

As a test of this, I reduced the "Dollars per Trade" to $5K and reran the test.  As you might expect, the Net Profit % dropped to 8.53% (which excludes commissions).  I then doubled the "Dollars per Trade" to $20K and reran the test -- the result was a Net Profit % of 34.79%.  To be accurate, the program should cap your ability to move into a position as long as the amount of available equity (cash) is greater than 0.  Again, I'll talk to Chris about this.

Bob Wilson and I are doing further tests with EdgeRater and I'll report back here as we move forward.  We're validating each other's work so that our concepts and findings are confirmed with another set of eyeballs, at least between the two of us.

In general, I like EdgeRater.  Furthermore, use of the 20 Connor's ETFs, combined with a good backtesting package like EdgeRater, can give you confidence to move into the markets when the equity price is falling yet the equity is in an uptrend (which is what a mean-reversion strategy accomplishes).  I'd like to see some nuances addressed, but for the most part, this is a solid piece of software that can give you (me) confidence to use the tested strategy.


Effective Volume

I am blogging daily at http://forums.effectivevolume.com.  I have my own forum there, and it is easier than blogger to have multiple threads/conversations in the running.  

Please register (for free) at the EV site and submit a request to join the GGT forum if you would like to see how I use GGT + EV for my investment decisions.


Remember, you are responsible for your own investment decisions, and I am not.  Please do your own work, and please take ownership for your actions.