Qs Linear Regression Crosses

Here's another snippet of the Qs mean reversion signal I've been working on. In much the same manner as the 8/16 moving average crossover this algorithm focuses on the reliability of using the crossover of 2 linear regressions to signal impending reversals of momentum.

In coming posts I'll detail my reasons for favoring linear regression and linear regression channels over other forecasting models. For the near term I'll just mention that linear regressions channels are unique (IMHO) in their ability to capture the trend of both mean reversion and random walk price patterns.

Below is the optimization report for a close range of inputs. What's interesting here is the robustness of the crossover, almost regardless of the settings. I've optimized for max consecutive losers (1), just because risk avoidance is more important to me than maximizing equity output.
Other traders may not be as risk adverse and may choose other input options.
The (9,13,10) model also produces nice returns.

The TS2000i code is very simple and, as with previous MR studies, I'm just looking at the short side for now. I've used a Fixed Bar exit for the study in lieu of a signal reversal. This tactic gains a few more bucks on the equity curve and knocks down the risk exposure time considerably.
The fixed bar exit is (no surprise) 9 days.

Weekly Update

The VIX/Price cross indicator proved its worth this week and one of my short term goals is to code the cross into TS and test it out on a variety of ETFs. Based on my little ETF basket, the VIX cross looks very encouraging. This rather unusual application of the VIX is not focused on the VIX's behavior above and below moving averages or in conjunction with bollinger bands, but is really a reflection of a zero line volatility cross in the same spirit as the standard deviation study of Friday. The odd hiccup on Thursday wherein the VIX and markets fell together can probably be chalked up to pre-expiration shenanigans or ? (send in your idea).
While the VIX cross will never get you in at the top or bottom, it will let you ride the sweet spot of the support/resistance cycle with very little risk exposure. It's difficult (for my old feeble mind) to imagine a daytrading application of the VIX cross, but on daily bars as a swing trading indicator there are several interesting possibilities including pairs arbitrage based on the sequential timing (firing) of the VIX crosses.
Yet to explored is a reliable exit strategy for the VIX crosses as waiting for a recross will invariably sacrifice most if not all of the accumulated trade gain. As usual, I'll look at fixed bar exits and several other momentum reversal signals profiled already in the Qs Dirty Dozen.

I'm not going to spend a lot of time on the 4LRs study this week. After this past week's carnage all the technicals in the basket are solidly oversold. IWM , the Qs and XLE look poised for a rally, currently riding the lower LR30 channel and, in the case of IWM and XLE, below the LR75.
While XLF's swoon has been dismal, if not downright pitiful, the LR4 chart suggests more downside action is likely as XLF sits on the LR30 mean.
We're overdue for a rally here, but when (and if) it comes next week, the first target levels of resistance in IWM, XLE and the Qs will be the LR30 means.
Easy does it.

Qs SD Signal Line

This is a continuing snippet of research based on last weekend's stated goal of refining a robust mean reversion signal using standard deviations of a floating price point, the pivots range mean reversion and a linear regression component.

This little algorithm was inspired by Jeff Pietsch's z-score analysis of ETF pairs trading opportunities. Now Jeff's work is considerably more complex than this system could ever hope to be, and I've stripped off the log functions to make it even simpler as they have no impact when not comparing 2 different ETFs. My goal here is rather basic and requires a conviction that price does indeed revert to the mean (eventually).
I've written before about the advantages of balancing probability theory (mean reversion) versus uncertainty (chaos theory, random walk, nonlinear logic) to build a Kit of Parts trading system of non-correlated factors and today's work focuses on some of the mean reversion aspects of that equation.

A little closer look at the TS code below reveals that we're actually looking for reversals above or below a zero line.
If the Len average price equals the current close, the result is 0, showing no + or - bias.
If the Len average price is less than the current close, the result is a + number, reflecting a price expansion above the current range.
If the Len average price is more than the current close, the result is a - number, reflecting price expansion below the current range.
If we then divide the +/- resultant by the standard deviation of the same Len range, we end up (hopefully) with a simple mean reversion to the zero line indicator.
I've just looked at the short side for now. . . the final product will incorporate both sides of the equation not only to generate BUY signals but, just as importantly, to confirm the bias of other mean reversion signal components.
As with many of the other short side systems I've profiled in the Qs Dirty Dozen, the Fixed Bar exit optimizes to 9, adding further support to this time frame as a reliable Qs short cycle.

Qs Short Setup

I thought I'd take a little break from the flurry of recent Qs system posts, so here's a picture perfect short at yesterday's open that alert traders should have picked up on.
With the NYAD running at .07 for most of Tuesday, Wednesday's uber pop to something like R25 was a dead give away that the open was going to retrace.
The subsequent downslope in the corresponding 2 minute TICK provided further confirmation of failing enthusiasm for the opening pop.
The parabolics actually gave us 2 quick short entries and once the NYAD and TICK leveled off a few minutes after 10:00 it looked like time to cover. With the Qs performing a quick penetration of S1, and no follow through, the cover signal was confirmed.
This quick trade produced a low exposure $ .40 in just 30 minutes, or a little more than $.01/minute, which is the level of return that I target for these 20-60 minute positions.

Qs RSI Lag

Continuing with the new Qs Dirty Dozen, the RSI Lag system looks like a worthy component.
Trading both long and short with a good balance of performance, the system incorporates a 2 high, 2 low signal with an RSI momentum confirmation. The long side has a few more consecutive losers than I'm typically comfortable with, but the relative drawdown in pretty muted. The system only holds for a few days (which I like), minimizing risk exposure.
The equity curve is attractive given the inclusion of both long and short signals. The short side only curve is even better looking. TS2000i code is posted below:

Short Qicker

As long as I was off on a tangent looking at the relative performance of certain candlestick patterns like the inside day and engulfing bar, I decided to look at a pattern that Steve Bigalow calls the most powerful pattern. . .the "kicker". This is basically a reversal bar that forecasts a change in market direction. Now Steve's definition of the short side "kicker" pattern requires the trigger bar to open at or below the previous day's low, which (no surprise) limits the number of incidents significantly.
In typical fashion I've revised the setup to loosen up entry conditions , I've added a little momentum filter and I've turned on pyramiding to take advantage of those times when the markets decide to do a little head-fake or retrace before going in the original direction forecast by the kicker . . . the result is the "Qicker".
I've written the TS2000i code as a series of conditions and,as usual, I've used a simple fixed bar exit. It's interesting that the fixed bar exits on the short side for the Qs system typically optimize in the 7-9 day range, so that encourages me to be alert for other Qs setups that mirror this timing cycle.

Weekly Update + Pivots MR

Above, the daily line charts of my 4 ETF basket overlaid by the VIX and the new moving LR7, which I described a couple weeks ago. I'm particularly interested in those instances where the VIX crosses the ETF price as in most cases this has resulted in a tradable 1 to 2 week price surge in the underlying ETF. This is clearly a longer term tell and. reiterating a tactical approach mentioned last week, the best (safest) strategy is to bide your time, employing the patience of Job, and waiting the signals to set up.

As part of my on-going systems refinement, my plan is to merge this VIX divergence signal with the pivots mean regression (MR) signal discussed below to produce a more robust forecaster of trend and momentum.
The lower chart here is the 4 ETF basket again with my usual 4LRs study. All the LR30s are downslope except the XLE, which continues to display a nice lateral pattern supporting my butterfly 47 position that has worked well for the past 2 months.

Some divergence in the RSI and the 4 LRs on all the ETFs, except the XLFs, which continue to look dismal.

My attitude for the coming week is one of caution . . . HEY!, that's the same one I've had for the past 18 months. With expiration this week and the full impact of the bailout plan yet to be factored into the market, I'm keeping my powder dry and sticking to my stable of daytrades to scalp off a few bucks until the trend signals get a bit more confirmation.

Below is the update of the Qs weekly pivot study that I began last weekend. After last week's decidedly negative results, which followed the Qs mean reversion (MR) chart's signal (bottom chart), I am clearly encouraged to pursue the forecasting possibilities of this signal a lot further. Now this chart just reflects the Qs behavior, so a bit of caution is warranted. Nevertheless, in the market's constant flux of mean reversion and random walk, the arguments favoring a mean reversion bias are increasingly supported by the results of the 2 Dirty Dozen system portfolios I've posted.

The results I've posted so far are based on weekly bars and I've yet to apply the algorithm to daily bars or even shorter time frames. It all takes time . . . a lot of time.

My intent for the short term is to refine a standard deviation signal line that was inspired by some of the truly stellar work by Jeff Pietsch at Market Rewind and combine that with the Qs MR signal line and the VIX/ETF crosses mnetioned above to yield a unique forecasting signal.

This endeavor is still in it's infancy and may ultimately become another Kit Of Parts project with a variety of adaptive plug and play Conditions that reflect short term trend and volatility. For the near term my focus will remain on the Qs as I complete the Qs Dirty Dozen components and update the Qs KOP, which continues to yield impressive (IMHO) returns with minimal drawdown exposure.

Qs Engulfing System

This is the inverse of my previous inside days study and reflect the Qs behavior subsequent to engulfing bars. TradeStation already has a resident algorithm for engulfing bars (plus other candle patterns), so the coding doesn't require much effort. Nevertheless, I've combined the bull and bear patterns and added my no-brainer fixed bar exits just to see if any pattern emerges.
While nobody's gonna get rich playing the long side here, the short side is an eye opener, with 100% success (this version has pyramiding turned on.) When used as a complement to the inside days system, it helps confirm short term momentum.
Exposure time is a comfortable 5 days (for me).
Not a lot of trades over the past 16 months but, as with most of the Dirty Dozen systems, the trick is to just sit tight until the systems fire before putting any dollars at risk. Now that's just my own tactical approach to risk management and each trader should review their own unique comfort level (and trading plan) before initiating any new positions.

TS2000i code is shown below:

Qs Happy Trails

Last July I introduced Happy Trails for the IWM.
This is an updated version for the Qs as I continue to build the Qs Dirty Dozen.
The top report has pyramiding turned on and requires a maximum 24x position.
The results speak for themselves.
I'm including this system in the Dozen for a number of reason. First, it doesn't use the RSI(2), and as such enhances the reliability of signals based on non-correlated factors (systems). Second, if you study this thing on your own TS chart you'll see how the sequence of the pyramiding trades help define the strength of momentum in either direction (homework for the True Believers). That whole momentum building concept is a study in itself that may see further investigation in my (very limited) spare time.
The report below is the basic system. . . so there's not a lot of activity based on 70 day cycles within a 16 month test period. That's why pyramiding was turned on for the top report.

Those who followed the first version of Happy Trails will notice that I've added a new input to reflect the variable momentum in the long and short cycle. While optimizing with the added input yields only marginal gains, at the same time the similar of the inputs reflects the reliability of the system. For the Qs test I turned off the stops previously engaged. TS users can fiddle with their favorite stops to tweak performance and manage risk to their own comfort level.

The inputs for Happy Trails with pyramiding turned on are a bit different, to reflect the increased activity. . .(60,70,13) so adjust accordingly when performing your own testing.

Cheaphooker Meets the Qs

I originally mentioned the Cheaphooker system and it's evolution last year. Here's the version that profiles the Qs as part of my on-going this project to complete the Qs Dirty Dozen systems and update the Qs KOP. As mentioned in the original post this system is a bit unusual in that the prime entry signals are triggered by the day of week (1=Monday, 2=Tuesday, etc.). Also, a bit different from the bulk of my system exits, which are at the Close, these positions exit at the Market, which is the next day's open.

As with many of the systems I've profiled recently that have both a long and short side, the short side outperforms the long side. . .in this case significantly so. . .reflecting the negative trend of the market over the backtest period. Trading once a week, with almost equal long and short signals, the long side gains are actually acceptable from my risk/reward criteria, while the short side trades provide exceptionally well. A 17:1 consec win/loss ratio and minimal intraday drawdown are great odds for just selling Thursday's close.

The short term MA filter that triggers the sell strength/buy weakness signals enhances system performance considerably.

Qs Pivot Bands

This is some continuing work on the weekly QQQQ pivots that I began last Sept 2nd. A little thumbnail explanation of what I was trying to accomplish with these studies is shown here. If this is something that interests you, I'd advise rewinding back to my Sept 2nd post and reading some of the subsequent postings and weekly updates to the pivot file. Probably 75% of my daytrading gains are achieved using only the daily pivots, the NYAD, the parabolics and the 8/16 MA cross on 2 and 5 minute bars, so the pivots are really fundamental to my bottom line and I'm always looking for ways to squeeze a little more juice out of them.

The chart at the top is simply a line diagram of the weekly R4-S4 pivots from 7/3 until the present. The lower excel matrix is a continuing work in progress to configure some forecasting metrics based on the pivot dynamics. If you take the time to review the previous pivot related posts you'll notice the slope convergence/divergence posts as indicators of market momentum and those turned put to be modestly correct.

Above is a simple weekly bar line chart of the PP and R1-S1 progressive delta for the same time frame. The calculations are simply the result of dividing the previous week's PP (series 1) and R1-S1 delta (series 2) by the current values. mean values therefore =1, reflecting no change and no volatility. The indicator is currently (no surprise) oversold. This is clearly a work in progress but preliminary studies have encouraged me to pursue system applications yet to come.

Qs RSI(2) Zippy

With the Qs in seriously uber-bought territory as mentioned in the weekend update, I though it might be fun to employ a very simple (short only) RSI(2) reversal system to the Qs and test the results. Turns out simple works pretty good. . . a theme I've explored before in depth.

I call this system Zippy, named after my first cat and also reflecting the way the exit signals trigger. In this case I've used multiple exits, either a fixed bar (6) or a close less than the 5 bar moving average. The entries are premised on a long term declining market close (less than the MA 150) and an RSI(2) cross above 62.

Now these conditions don't reflect the current Qs scenario laid out this weekend or reflect the RSI(2) current 90+ readings. Battle weary traders know from experience that a high probability tactic for the Qs now is to wait until the RSI(2) crosses below the "magic number" 90 (or whatever you've predetermined as the target reversal threshold) before initiating shorts.
Nevertheless, patience is the watch word here, and when the Zippy setup does occur down the road, this is a great low risk approach to harvesting some gains.

The max consecutive losers is only 1 and the intraday drawdown is minimal, reflecting the type of risk control I prefer. Zippy will become one of the BZB Qs Dirty Dozen systems, to compliment the IWM Dirty Dozen. . .see right panel of blog for details.
The current rolling over of Zippy's equity curve may be an indication of a changing in momentum in the Qs as suggested by the 4 linear regression study this past weekend.
As usual, TS 2000i code for Zippy is shown below.

And. . .as has happened before with these systems, there's a little counter intuitive logic here.
The first question a quick glance at this code would typically generate is "what's with the RSI 62 number. That seems way too low". But you've got to keep in mind that we're selling weakness here, not strength. As the Qs in a downward trend try to recover, Zippy finds that a reading of RSI62 (isn't that a fib number?) provides a great tell for a failed rally.

The second question a quick glance at the code should generate is "what's with the exit signal that requires a close less than the Average. Shouldn't that read "more than?""
Nope, and the same answer applies. Most of the gain from this system derives from the 6 bar fixed exit, but we do pick up a few extra bucks and increase our trade probability considerably if we confirm the downward trend with this little condition.
And you thought trading was easy.

Weekly Update

Here's my ETF basket update using the new 4 linear regression study (75,30,11,M7) detailed last weekend.

The Qs have led the way up for the past 3 weeks and are now in overbought territory. As in the past, overbought and oversold conditions can continue for quite some time, so this situation alone should not be a trigger to bet against the Qs. A more mean regression oriented theme would be to expect the other sector ETFs to gain more ground proportionate to the Qs, before a reversal kicks in. The midpanel technicals in the IWM, XLF and XLE reinforce the probability of a bullish surge with short term targets of IWM-50, XLF-12 and XLE-51.

The current market strength has been widely attributed to the wonderful new (emerging) bail-out plan and may in fact be the catalyst for a market to kiss the channels good-bye to the upside. These moves would, of course, be driven by hope and hype, more so than actual fundamentals since that $800B or $1T has to come from somewhere, and mostly that's yours and my pockets, so there's a dark side to the coin, whose implications are seldom mentioned by the talking heads (sorry, got off a little tangent there).

Below, the VIX and T2100 (advance/decline line) with the VIX reflecting a neutral to bullish stance hugging the zero line while the daily T2100 is strongly bullish for the short term (next week). The weekly T2100 is at the upper band of the LR30, but the mean of the LR75, so it's some got the potential for a 5 point move up before becoming seriously overbought.

More QQQQ 3 Finger Setups

Here are the results for yesterday's Qs 3 finger reverse (short only) system using the same settings (54,16,12,10) but with pyramiding turned off. This change reduces the capital requirements considerably at the sacrifice of more than 50% of the net gain. Comparing yesterday's breakdown with this one, an alternate strategy would simply be to double your initial entry position size, which would yield comparable results to a pyramid approach.

Below, on the opposite side of the coin is the 3 finger lead (3FL)system, included in the BZB Dirty Dozen and reflecting long only trades. The test results of the system for the Qs are clearly less impressive than the 3 finger reverse, probably reflecting the overall downtrend of the market during the test period and the resultant lack of upside trade opportunities during this time frame. This suspicion is confirmed by the optimized 3FL backtest, which yields settings of (30,8,10,8). The look back period has been compressed from 54 days for the short side only 3 finger reverse trades to a 30 day look back in order to capture profitable trades.

Above is the test study with pyramiding turned off, below is the study with pyramiding turned on. Once again, an alternate risk management approach is simply to double initial entry position size to capture more gains in lieu of trading the pyramiding model which requires more capital.

Qs 3 Finger Reverse

Here's the newest addition to the updated Qs KOP, the 3 finger reverse (3FR) system component. . . and I really like the way this one turned out. See yesterday's post for the link to the original 3 finger reverse post, which includes the TS 2000i coding. The settings for the Qs (with pyramiding turned on) are (54,16,12,10).

As with previous 3 finger lead and 3 finger reverse system studies, the goal is to identify sector or ETF component leaders and then ride the momentum of the outliers to pick up the lag in momentum in the larger sector or ETF.
Based on the previously posted risk map of the Qs, we can see that the largest market cap components of the Qs are ORCL, CSCO, MSFT, INTC, GOOG and AAPL. Picking the more volatile of these stocks suggests AAPL and GOOG and I've included RIMM as the third outlier just because I know from tracking RIMM for a few years that it's a great tell for short term Qs moves (a little nonlinear logic there).
I'll no doubt test out other outliers but the results of this study, with only 1 consecutive loser is very low risk. The max intraday drawdown is also very low, so from a risk exposure viewpoint this is about as good as it gets.
As with previous new QsKOP studies, we only looked back to 9/07, which works out well in capturing the 11/07 swoon when we consider the 54 bar look back that characterizes this study.

The second mouse gets the cheese

This is a continuing study of my 3 finger reverse short only system (3FR), in this case using the IYT ishares Transports ETF (no 2x 0r 3x to my knowledge). Pyramiding is turned on, which results in multiple entries with a fixed bar cumulative exit based on the first entry. Turning on the pyramiding function also requires a considerably larger capital commitment that many traders may not be comfortable with. An obvious alternative tactic might be to just sell calls or buy puts since the average hold time is only 6 days. The bad news here is that the IYT options are very thinly traded and the spreads approach the comical at .15 to .35 for the ATMs. Turning off pyramiding results in only 8 trades for the same time period, with only 1 consecutive loser, a 75% success rate and approximately 1/2 the net gain with pyramiding. The basic system trades about every 6 weeks so lots of other opportunities to deploy your capital while waiting for the setups. Best used in conjunction with the 3 finger lead system (see BZB Dirty Dozen) for a full range of long & short setups.

Revised 3FR settings for the IYT are (50,16,10,6).

As with the earlier 3FR studies, the goal is to find sector (or ETF) leaders and use their price action to predict the reaction in a larger and slower ETF. The leaders (first mice) don't actually have to be components of the ETF (most often they are), but the trick is to find highly correlated but more volatile mice to lead the way into the trap and absorb the risk so that the probability path of the second mouse (diluted volatility ETF) is much higher and less prone to the uncertainty of surprise.

TS is not generating the correct net gains, which I'm trying to resolve. Clearly no one would trade this system for such marginal gains. The posted system gain/loss numbers need to be multiplied by 1000 (or 100) (or whatever the position size you choose) The results produced by this report present 1 contract (share). You can see the actual impressive gains that this system produced in the trade breakdown above.

Dollar / NYAD mean reversion

Here's a trade setup I'm exploring that looks similar to the GE/NYAD setup but, in this case we're looking at the dollar as reflected by the UUP, the Proshares double dollar bullish ETF. The big difference between GE and UUP is, of course, that the dollar typically trends opposite the major market indices, so a chart of the UUP looks a lot more like the VIX than it does the SPY.
In the above chart, the UUP is shown (colored candles) with the NYAD in the shadow candles.
The trading box in orange describes the divergence and reversion setup.
The lower chart annex is the true NYAD overlaid with the 8/16 MAs and a 5,20,3 MACD signal line and a classic 12,26,9 MACD histogram in the lower study panel.
Notice how the zero line crosses on the MACD histogram have provided solid confirmation of either the 8/16 MA crosses or the parabolic SAR signals.
The technique of using the MACD signal line in conjunction with the MACD histogram is something I've been fiddling with lately and in the process developed this pairing which produces fairly consistent results when used in conjunction with the NYAD. Whether this pairing can deliver similar results with other indicators and/or equities will be a subject for future post explorations.
As with the GE/NYAD trades, the UUP/NYAD trades involve a lot of waiting around. On trending days no trades typically develop, while on volatile range days, 2-4 setup opportunities may present. This type of trade can test the resolve of the more hyper traders, but, as usual you have to let the trades come to you and avoid the perils of seeing a setup (and trading) when no setup and no trading edge exists.

XLE closeup

This is a closer look at the XLE Spyder energy sector ETF, one of my 4 ETF basket that I regularly trade. The interesting thing about the chart configuration of the XLE is the flat profile of the LR75 channel(blue), with the mean yielding a clear reversion target of 47.
In December we put on the Jan 47 butterfly with great results, and we are currently in the Feb 47 butterfly, having entered on Jan 21st as the XLE touched 47. The great things about the XLE (depending on your viewpoint) are the highly liquid options and the penny spreads, although the spreads open up a bit as the strikes move away from the ATMs.For those who prefer to just trade the ETF and not play the premium decay game, the XLE triple, ERX, generates more bang for the buck (assuming you're right), as evidenced by this little backtest of the past 500 bars using a simple long/short RSI system on the ERX and XLE on 5, 10, 60 minute and daily bars (ERX has only been up since 11/6/08, so there's not really 500 days of data). Options on the ERX are thin and the spreads are humongous, so don't even think about it.

Finally, an impressive piece of work from Market Rewind providing a neat little snapshot of ETF performance metrics for a variety of time periods. Thanks Jeff.

Weekly Update - new template

First of all, here's a new thumbnail look at my little 4 ETF basket using line charts of the ETFs and the VIX and overlaid with the new Telechart moving linear regression (7) study (described below). This is a little visual lesson in mean regression theory and is designed to help forecast when the ETF/VIX divergence is about to mean revert.

For this week's secondary study (below) I've returned to the VIX daily and weekly bars and the T2100 advance/decline line daily and weekly bars.
While Don Worden's comment at Friday's close was in support of an oversold environment, the technical studies presented in these 3 charts suggest a grimmer prognosis as all technicals are essentially downtrend at the zero line.

The big change this week is a reconfiguration of the 3LRs study (3 linear regressions) to a new 4LRs study. Previously, I was using periods of 30,11 and 3 to define the LR channels. I've now revised the LR periods to 75,30,11. I've added the LR75 to help better detect potential butterfly candidates, a strategy which has worked out well so far. Telechart has also added a moving average linear regression (MLR) study to their inventory of indicators and after playing with various settings, I've settled on a MLR7 to replace the LR3, which was the least useful LR of the study (IMHO) in forecasting momentum and direction. Note how the MALR7 mirrors the cycle of the RSI2 and stochastics 10,2,2 in the middle technical pane of the charts. I've also reconfigured the MACD 5,20,3 indicator to display as a histogram for a somewhat clearer view of momentum and zero line reversals.

Finally, here's an interesting site using a sophisticated algorithm to time the markets. They provide a free weekly forecast of the SPY low. . .you've got to spend some money if you want the rest of the stats. They've been pretty right on for the past month, although this week's forecast was not a rousing success (HEY!, everybody misses a few). Good for confirmations of market turn ups, put sales and potential short term longs.

Qs Inside Days

Here's another addition to the new KOP Qs using the inside day system I profiled earlier this month. Not a lot of trades, about 1 every 3 weeks, but the results are pretty impressive, with marginal drawdown. Refer to the original inside day post for system concept details and open code posted as Comment #1. Keep in mind our test period is only 9/07 to present.

TS2000i code is below: