Monday, August 31, 2009
It's been almost a year since the original post and the equity curve is still chugging along with a couple real zingers on the downside.
Trades 100-105 were, eh, terrible and one good argument for some type of calamity stop loss, although the equity curve recovered to parity by trade 110.
Equity growth has been zippo since trade 130 and I intend to review each trade in turn to determine possible root problems for that situation. Breaking news as I figure them out.
Part of the problem with the KOP is the lack of robustness on the long side entries. The short side entries demonstrate a lot more robustness and as part of future refinements to the KOP I intend to add more non-correlated entry signals fro the double 33s, big gulp and % VIX advantage systems profiled recently.
Friday, August 28, 2009
The performance report is shown below..................
Thursday, August 27, 2009
Wednesday, August 26, 2009
The hairy bottom resolved above the PP at 9:43, accompanied by a parabolic BUY signal.
The trade got some additional confirmation from a VXN cross at 9:45.
The NYAD MAs are upslope from the open and the parabolics fire a BUY at 9:46.
I've warned previously against putting too much stock in NYAD parabolic signals until after the first 30-45 minutes, but in this case I liked the way the NYAD was riding above the double 8s channel, and entered the trade.
This was a real quickie as GE shot for R1 at 10:00, at which point all the technicals started to deteriorate. . and I closed the trade.
Total time in trade was 18 minutes and the net gain was only $.12, but I followed my trading plan signals so I've no regrets.
Tuesday, August 25, 2009
The net result of the $4 breakeven floor stop is that we get a lot of trades that go nowhere. . . as reflected in the flat period of the equity curve shown below. Nevertheless, there are a number of simple conditions that can easily tone down the trade frequency and reduce the equity chatter.
Monday, August 24, 2009
In the spirit of Mark Fisher's The Logical Trader I wait for the first hour and then run fib lines from the current low to high. The fibs help sort out the odds for momentum moves when the pivots have not been hit and can support breakout/breakdown moves from the pivots.
In today's scenario the first hour range defined by the white vertical line set the 0% line fib about 30% above the PP pivot.
#1. . at 11:18 the parabolics fire SELL off both the R1 and 100% fib line.
#2. . the NYAD also fires a parabolics SELL
#3. . the MAs slope of both the Qs and the NYAD continue downslope
#4. . Qs reach and breach the 0% fib line, BUT ARE STILL 30% ABOVE PP.
#5. . With the NYAD still downslope, this is my confirmation to stay the trade and not cut and run.
#6. . Qs barely kiss the PP at 13:30 (not shown) and I'm out of the trade.
In summary, a smooth trade with a clear short entry and an exit at the target pivot. A $.38 move in 132 minutes. . . longer than I like to be exposed, but it looked like a trend day and the odds favored riding it for a while.
Saturday, August 22, 2009
This one turned out a bit better than the Tuesday trade.
Today we're looking at 2 minute bars, which is actually the time frame you need to monitor to catch the VXN crosses in real time.
Looking at the VXN crosses in hindsight alters their relative position on the chart due to a change in scale dynamics as time progresses: the greater the time frame lookback, the greater the distortion of the actual cross.
#1 . . At 11:50 the Qs are sitting mid point (actually dead on 62% fib line)
#2 . . The parabolics are on a BUY and the VXN crosses at 11:50
#3 . . The 3 MAs are upslope and the MACD histogram is above the zero line and rising
#4 . . This is a low risk entry given the Qs relationship to the pivots and I enter long with the PP as my first target (line of resistance).
#5 . . The Qs blow through PP in only 14 minutes, run a little spill over and the parabolics fire a SELL (cover) at 12:18, at which point I exit the position.
This was a quick little trade, just 28 minutes. I actually added another BUY lot to the trade at 11:58 since it's my experience that this type of momentum (greater than 60 degree slope) and close proximity to the next pivot will drive price to that level.
I exited both positions at 12:18 and am off for the rest of the day.
Friday, August 21, 2009
In this case I've applied the system to the IWM for our demonstration backtest period..
.....A little aside here.....
About 10 years ago when I first got into systems trading INTC use to be the poster child for technical alignment. (also see CSS Analytics). TradeStation used INTC heavily to promote the "reliability' of their systems and to demonstrate how easy is was to make money using simple algorithms. Times have changed. And I personally find IWM to be a superior index for testing trading concepts in lieu of the DIA, SPY , Qs and any individual stocks. So, the next logical question is.."Why not just trade the IWM?"
The answer is that the Qs offer certain features that I prefer and I actually find it easier to make money with the Qs than the IWM. That doesn't mean I ignore the IWM. In fact, I watch it with the same 2 minute scrutiny that I watch the Qs and I'll explain why and in future posts.
For the time being, an application of the Double 3s yields a respectable equity curve. I've used a 6 period fixed bar exit BUT, and this is a big but, there is no "MarketPosition=0" in the code so the system will stop and reverse if a contrary signal fires ahead of the 6 day fixed exit. The net result is that the average number of bars for winners is only 2. The other net result is that this is a very active system with 201 days in 29 months, or about 1 trade every 6.5 days.
As mentioned in the original 3x3 post one thing I look for is the max consecutive losers both long and short, and this system generates some acceptable returns.
The max intraday drawdown is twice as stiff for the longs and the shorts, possibly reflected the short bias for most of the test period and the tendency of the market to drop faster than it rises.
The trade breakdown shown below shows that a simple stop loss of $200 could improve net return and cull out the losing outliers.
Keep in mind that this version of the double 3s is a work in progress and just part of a larger trading concept that I've been exploring since May of 2008, that is, using a small basket of systems to trade a small basket of ETFs.
And while the post may generate as many questions and answers. . . to me that's what its all about.
Thursday, August 20, 2009
This was followed by a number of subsequent blog posts discussing blogger cooperation, information sharing and the value of critique versus respect. I'm not going to posts all the links. If you're a regular visitor to this site all you have to do is flick over to the Sites of Interest on the right side of the blog and flip through the archives of a few quant sites to know what I'm talking about.
That's just me.
Wednesday, August 19, 2009
This is Tuesday's chart of the Qs 5 minute bars and the green vertical line is our long trigger.
#1 . . the NYAD is showing a little breakout to the upside while the 3 MA signal line (3,7,14) remains solidly upslope.
#2 . . the parabolics are on a BUY on the NAD and we're well into the day at this point (11:12 pst) so the first hour caution about following the parabolics on the NYAD has been passed.
#3 . . the VXN has just crossed through the Qs price chart. These are what I call VIXEN trades and they have a very high success rate, although seldom picking absolute short term tops or bottoms.
#4 . . Qs are just breaking the R1 pivot resistance. I actually waiting until 11:15 take the trade as I wanted to see even more conformation that the Qs were going up.
#5 . . Qs are on a parabolic BUY signal
#6 . . Qs are riding nicely my double 8s channel (8 MA of the high, 8MA of the low) this is very bullish.
FOR THE EXIT..................
Overall market momentum was a bit iffy Tuesday so I was prepared to bail on this trade at the first sign of failure. That occurred at the red vertical line when. .
#7 . . Both the NYAD and the Qs parabolics fire SELL (or cover)
#8 . . Both the NYAD and Qs prices fall into the double 8s channel
This turned out to be a case of premature exit and only netted a lousy .06 so I should have stuck with it until 11:48 when I could have netted another .08.
It's a learning process.
But . . I followed the trading plan signals and the next time I may give the trade a little more room to breath. Maybe not.
Tuesday, August 18, 2009
One of the problems with relying on daytrading to provide the bulk of your nut is that you've got to ride momentum and carve out whatever you can on the intraday swings. This tactical approach to trading can produce a kind of trigger-happy attitude that really needs to be contained in order to avoid the otherwise frustrating whipsaws that can test a trader's patience and sanity.
And then there are days like yesterday . . . which I consider a high-risk throw away for day traders. In the chat room the guys are always on the lookout for the breaks, either up or down, and the goal is to get multiple technical confirmations before jumping into the tar pit.
But this wasn't a down trend day. . .this was just a down day.
And the clues were all there to show that this was going to be a rough one for any buyers.
#1 . . the NYAD. It never got above .12. Without this prime indicator showing some strength there was literally no hope for a rally. The NYAD is my chief forecaster for market momentum and I watch it closely on 1 and 5 minute bars all day long.
#2 . . the VIX. Shown as an overlay on the Qs chart, there was never any time during the day when the VIXEN cross over pattern (see right blog panel) was even a possibility. These are low risk trades and without some momentum towards a cross momentum has to be regarded as dead.
#3 . . the underlying technicals. The 3 moving average signal was flat as a pancake both in the Qs and the NYAD (actually the Qs were slightly downslope most of the day).
#4 . . the ambiguity of the Qs parabolics signal. When you see this type of ebb and flow parabolic signal the safe course of action is to stand back until it gets resolved. The trick is you never know which way it will get resolved so patience is the best risk management tool you've got. Use it!
Monday, August 17, 2009
I've reset the inputs to 3,14,3,2,10. And, when I tested these inputs on the original test period the return was actually as good as the March 4th settings.
Interestingly, this new expanded timeframe backtest actually confirms my previous 3,14,3 MACD settings. Sometimes I surprise myself!
Equity curve is shown below and while the returns aren't dramatic, keep in mind that I'm actually more concerned with maximum intrady drawdown and max consecutive losers. In that respect the system has some attractive performance features.
Pyramiding has not been turned on. . .I'll get to that in a later post.
Below is the distribution of the actual trades and it should be immediately apparent that system performance can be improved by culling out the outliers losers and a good percentage of other losers, by simply setting a $125 stop loss/trade. The performance report is based on 100 share positions of the Qs, whose value has fluctuated during the test period, so an alternate approach would be to use a 4% stop loss.
Friday, August 14, 2009
While I'm off for the rest of day on some family business, here are three articles to tease your trading brain and share some of the best current thinking on dynamic equity curve switching, mean reversion and implied statistical correlations.
These are all themes that I've (very) briefly touched on in recent posts, but the collective output of this little brain trust provides insights and perspectives well beyond the abilities of this old trader and should pique' the interest of every inquisitive mind. Check em' out.
Jeff Pietsch - Equity Curve Switching
Michael Stokes - Short-Term Mean Reversion
Don Fishback - Statistical Correlation
Thursday, August 13, 2009
Wednesday, August 12, 2009
I got a little sidetracked over the past few days on the VIX % Change system and the subsequent Topticker post, but now I'm headed back to the barn and saddling up the Qs as I get down to business with refining (hopefully) this 3x3 system.
Bear in mind these upcoming posts are a work in progress and may end up a complete fiasco so as long as your expectations are as modest as mine we may all come out of this OK.
Per the above Qs performance report I've optimized the Qs Double 7s for the same arbitrary time period used for the SPX on Monday. The optimized settings returned are 3,3,5,9. Now, don't send me nasty emails or comments saying I handpicked the time frame. . .I didn't. . .and in future posts I'll suggest some signals that optimization has become curve fitting and that input variables need to be reconsidered.
Keep in mind that our ultimate objective is a system with only a few variables that can utilized as a timing model not only for swing traders but for day traders as well using a variety of intraday time frames.
For the time being we'll focus on the daily bars . . . and I find it somewhat remarkable that the first 2 inputs defining the higher highs and lower lows entry thresholds are the same as the SPX : 3 and 3. This leads me to suspect a higher correlation between the SPX and the Qs than I had considered earlier. At the same time, the fixed exit bars for the Qs optimize at 5 & 9, whereas for the SPX were 9 & 10.
Given that the SPX system made 174 trades and the Qs made 197 trades in the same time period this was somewhat to be expected.
The other interesting data generated by this study was the optimized period of 9 days for a short position in the Qs.
The 9 day fixed bar short position has been generated by over half of the Qs Dirty Dozen systems that I've profiled even though they are based on a non-correlated basket of entry conditions, so I think this is a valuable number to file away when trading the Qs short.
In addition to the system development questions mentioned on Monday two more of extreme importance to me include:
What is the time exposure of each trade and how much time is capital at risk?
For my own account, the larger the size of the capital pool at risk, the shorter the targeted time frame I prefer. Which is one of the reasons that daytrading is my preferred MO and where I focus my risk and why my swing trading is focused on lower risk premium decay and hedged strategies. Again . . . just my comfort level. . .you may have a completely different (and completely legitimate) risk tolerance level.
Is there a balance between long and short trades?
If the system trades 80% long then I'm likely to question why and look closely to see what happens in down markets and why no corresponding short signals are generated. This long/short balance will come into play in future posts as we explore how periodic resetting of the system variables can produce a dynamic market timing model without compromising risk exposure. . . and those last 4 words are the key.
Tuesday, August 11, 2009
Normally I don't do reader requests, my workload is quite full already, thank you very much. But I'm a practical and inquisitive kind of guy, a Capricorn by birth, and my first response to any such query is "show me the code!'.
Now I understand that some bloggers are reluctant to post code, either because they think somebody will steal it and make a gilzillion dollars and they won't get any residuals or any number of other reasons. I have no such reservations. On the downside I have noted with some dismay that some of my dirty dozen systems have been offered on a number of sites as monthly fee based systems. Of course, I was never contacted, offered any revenue split or acknowledged as the author of the code. That's the world we live in. Some eagles and some bottom feeders.
Sorry for the digression.
Nevertheless, I wanted to test TopTicker's hypothesis and ran a backtest of the exact same time frame using only the SPX and the entry TopTicker provided.
I used the same exit signals and the same inputs.
The results are shown above.
Now, to give the system a fair chance, I re-optimized the inputs, which yielded a 9,2 setting as opposed to the %VIX settings of 9,5.
Those results are shown above and reflect a much improved net return.
The equity curve shown below reflects the re-optimized 9,2 inputs, and does amazingly well until trade 70, which corresponds with Feb, 4 ,2009.
Obviously there's been a basic change in the trading paradigm subsequent to that period as 14 of the last 17 trades have been losers.
Keep in mind that this is an incredibly simple system and we could obviously control the drawdown with a variety of stops and checks.
That being said, a simple "do the opposite" approach to the SPX does appear to merit further investigation . . . which at this juncure I will leave to TopTicker, Michael and others to investigate.
I also highly recommend checking out Jeff Pietsch's spin the DV-2. Jeff is a thinking man's quant whom I am honored to call a friend. He also runs a chat room frequented by a few other notable bloggers and his fee based products IMHO are a exceptional value at 5x the price.
The TopTicker TradeStation 2001 code language is shown below for those that want to fiddle with it some more. Let me know how it goes. . . .
Monday, August 10, 2009
Michael (an analytical trader and blogger that I greatly respect) had been critical of that post, suggesting that asset correlation studies are inherently flawed and fraught with other difficulties. And I'm willing to admit that may be so . . . to a point.
First of all, trying to forecast anything the markets will do is an exercise that frequently requires the patience of Job, the wisodm of Solomon, and otherwise tests the limits of of one's sanity and capital account.
That being said, it's my intent to explore Michael's contentions in the near future.
Until then. . . .
The systems that I've posted on this blog are part of a research project begun back in May 2008. That project was intended to develop a basket of non-correlated systems that could be used to trade a small basket if stocks. I originally called the project KOP (Kit of Parts) and 2 deliverables from that research were the KOP10 for the IWM and the Qs KOP.
The idea was that by using non-correlated variables such as RSI2, the Detrend, standard deviations, MA crosses, linear regression, candlestick patterns, day of week, relative beta and a few others that a trading model could be produced that would work in both up and down trend markets and in both trading range and trending markets.
This was obviously an ambitious endeavor and I'm grateful for the input and suggestions that the on-going project refinement have elicited.
In the process of developing the KOP component systems I asked myself several questions:
What is the goal of the system?
It is designed as a stand alone system or part of a larger market timing package.
Does the system offer a unique timing perspective on the market?
Are the buy/sell signals in sync with other non-correlated timing models, or does the system offer opportunities to capture trades that might otherwise not be apparent.
Does the system offer unique risk exposure controls?
As I've mentioned many times on the blog, capital preservation is Job #1 in my book and the metrics that I consistently look for in evaluating system performance rank maximum drawdown and max. consecutive winners/losers more so that total net return.
Maybe not your ideal trading plan, but it works great for me. That's just my comfort level.
Does the system provide consistent trade confirmations of other non-correlated systems?
If it does deliver consistent confirmations then I feel a lot more comfortable risking my money using both it and the other related systems in the basket.
There are lots of other questions that I ask myself in the process of system develop and testing, but these are a few of the biggies.
Which brings us to today's post.
As with Friday's post, this is jumping off point, not a final destination, so bear with me over the next few days as we explore the potential profits and pitfalls of a simple system that sells a series of recent highs and buys a series of recent lows.
See the exact Tradestation code language by linking to the original Qs 7&7 system above.The optimized settings for the SPX over the specified time period are (3,3,9,10), meaning the system sells and buys 3 day highs and lows and holds 9-10 days. No stops and no other money management controls have been applied.
Were we to test other time frames, we might expect different optimized settings, both for entries and for exits and that one parameter we examine in future posts.
Connors original version of the system included a 200MA entry filter, which I have deleted in an attempt to increase trade frequency and total net return.
In future posts I'll look at several possible system refinements including:
Pyramiding the entries
Optimizing over a variety of time frames and what those differences might mean
The use of alternate exits (OR conditions)
and perhaps most importantly for me. . .
Ways to contract the drawdown
Friday, August 07, 2009
And now for something completely different. . . . . .
One thing that's caught my attention for a while is the daily % change in the VIX relative to the major indices, including the Dow, SPX and NDX.
I've noticed a tendency for exaggerated VIX moves to foreshadow subsequent short term trends in the indices and here's a simple little study to validate my intuition.
I choose to make the indices the target of the trade rather than the VIX since, by it's very nature, there's considerably more daily noise in the VIX than in the underlying indices.
That format almost makes it easy to test alternate indices performance relative to the VIX.
Simply put, the system sells the SPY if the daily % change in the SPY is greater than the daily % change in the VIX, and then exits the position after a fixed period. Vice versa for the sell side.
A 65% percent profitable return may not seem very impressive, but this is a very simple system and easily improved by adding entry filters such as a % change threshold, an MA trend, etc., as well as exits conditions, trailing stops, etc.
Those improvements could also improve the drawdown and consecutive winners/losers ratio. . . 2 of the most important metrics that I consider when evaluating a system.
And, as always, keep in mind that these studies are not meant to be canned solutions but rather jumping off points for your further exploration. That's one of the reasons they're free. The equity curve really begins to take off around trade # 20, which was initiated on 12/27/07 which is reflective of the beginning of the bear paradigm. Were we to rerun the data backtest with this as out start date, the results would be improved considerably without any further code modications.
Below is the TS 2000i code with the fixed bar exits optimized for the SPY. In future posts I'll look at some of the other indices, add a few filters and some other tweaks. So far this simple approach looks very encouraging for both capturing short term gains and creating a probability model for short term direction.
Thursday, August 06, 2009
TopTick had noted in the comments that it looked like both of the faster slopes had to cross the slower slope concurrently and questioned the validity of such an approach.
Above is the TS chart of the 6 trades showing exactly how the entries and exits transpired.
Even though the inputs for the long and short sides were somewhat different, the trades essentially resolved into a switch and reverse context.
Keep in mind that this study was only intended as a tangent to Michael's investigations and reflects just another attempt on my part to develop a basket of systems (begun back in May of 2008) to forecast market momentum and direction in an effort to protect my longer term capital positions.
While some may find fault with the execution on the system, the trades (and minimal drawdown) speak for themselves.
After Tuesday's post I reran the system with pyramiding turned on (shown below), which effectively increased the total number of trades to 18 and increased net return almost 300%.
I also made the system a pure switch and reverse and adjusted the inputs to 10,12 and 44. It turns out that the system performs almost as well with only the 10 and 44 inputs so I have to question the need for the confirming second fast slope crossover and will explore that further as time permits.
The big negative with the pyramid approach is that by scaling in to the trades you're ultimately committed to a 500 share position at various points, thereby increasing your overall risk exposure substantially. The plus side is that much of that extra capital is not exposed for the same period as in the original study.
This then becomes a matter of comfort level with the reliability of the system and the willingness to capitalize it. From my own risk management perspective I would focus on the original system and use others such as the 3 Finger Lead and Reverse systems for trend confirmation before adding capital to the original position.
The 3 Finger systems have proven robust in both up and down trending markets and those that lurk around the Market Rewind chat room where I hang out a few days a week, know that the intraday signals of the 3 Fingers have proven to be consistently reliable forecasters of the Qs short term direction.
Wednesday, August 05, 2009
The charts continue with the theme developed last week and offer few technical clues as to likely near term developments.
Overall, the VIX trend continues down while the Qs trend continues up. Today's open provided solid overhead resistance and for those that follow the new daily high counts and low counts, it was obvious that the field is getting thinned with a few notable exceptions such as CIT, AIG and C . . with a modest 1.5 billion shares traded.
The SPX (not shown) has shown a real affinity for the 1000 level for 3 days now and the break up or down is likely to be substantial.
With Thursday providing the greatest historical probability as a pivot high day we may yet see another push up before a fade of substance.
The markets continue to confound mean reversion and counter trend traders and I have no regrets about going to 95% cash last week for my longer term positions.
The one position I'm still holding, GE, continues to melt up although I'm prepared to exit as soon as the upslope trend looks at risk.
Tuesday, August 04, 2009
Being the inveterate tinkerer that I am, I utilized a similar concept using the cross and confirmation of linear regression slopes to see how well they tracked the Qs trend.
Turns out they track it pretty well.
I use somewhat different settings for the long and the short side because the short side tends to be more volatile and hence reflects a shorter time frame to complete.
There's not a lot of trades here. . . 6 in 4 years and the time in the trades is off the scale for my normal comfort level. . . 110 days for the longs, 200 days for the shorts but, what's truly amazing to me is the piddling drawdown that this system incurs considering the length of the exposure.
Equity curve shown above, reflecting the relative gain in each of the trades.
Not a loser in the bunch.
Not sure I'm ready to bet the farm on this simple timing model, but certainly worthy of further exploration for smaller time frames and the application of various filters to increase trade frequency and, hopefully, net return.
Below is the TradeStation 2000i code.
TS users with versions 8.3 and above know the BUY/SELL and EXIT command language requires minor modification to run properly.
Monday, August 03, 2009
And, while it would be easy to dismiss this statistical anomaly as the product of underlying put buying there's another approach to using the VIX that I've profiled many times before and it works even when the VIX is acting out like an unruly child who refuses to behave.
We're looking at a 2 minute chart of the Qs with the VXN overlaid as a line (white) study.
Below the Qs chart is a chart of the NYAD on a similar time frame with the VIX overlaid.
In the top chart I've used the red and green trend lines to reflect the VXN implied momentum of the Qs.
In the lower chart I've used the red and green trend lines to signify the trend of the NYAD.
Note in particular the VIX/VXN crosses generated at 13:00 in both the Qs and the NYAD.
These crosses will always keep you on the right side of the trend and help avoid those otherwise frustrating whipsaws that can easily drive a trader to distraction.
Just another setup to log into your technical trading toolbox.