by Andrew Coles, April 12, 2011, midasmarketanalysis.com.
One of the site’s regular readers emailed recently to ask if I could summarize the types of MIDAS curve now developed, as there was some confusion over references and terminology. The aim of this post is to cover this topic briefly. This is a MIDAS tutorial so it will also be stored in the folder ‘MIDAS tutorials’.
A final table at the end of this blog post is a concise summary of the various curves.
Two basic methods of evolving the curves
There are obviously two ways in which any formula or algorithmic procedure can be adapted. Very simply, the first would involve changing or extending the actual data input to the formula, while the second would involve either changing or adding to the basic maths. With these two ways in mind, what follows is a quick review of the recent development of MIDAS curves.
First Generation curves: Paul Levine’s original (minor) adaptation of the basic VWAP formula
First Generation curves are Levine’s modification of the basic VWAP formula upon which First Generation curves are based. He applied this basic modification (a simple subtraction procedure involving the launch bar) both to his standard MIDAS support/resistance curves and to his more complex Topfinder/Bottomfinder (TB-F) curves. This work was the subject of his original online lectures.
Second Generation curves: replacing market volume with constant volume to create “constant volume” or “nominal” curves
Second Generation (“nominal”) curves represent a vital move forward from First Generation curves in so far as they depend on an essential, more advanced, understanding of how to apply MIDAS curves. Since First Generation curves critically rely on the marked-derived cumulative volume input in the volume-weighted aspect of the VWAP formula upon which they’re based, their chart positioning is heavily influenced by this market-based cumulative volume input.
This volume influence in VWAP goes back to the original VWAP formula in so far as it was always understood that the more volume traded at a certain price level, the more impact it has on the VWAP. However, beyond this insight the actual role of volume was never investigated more deeply.
When analysed more thoroughly, it emerges that various (identifiable) relationships between price trends and volume trends have a crucial role in the plotting of MIDAS curves. For example, what happens when a MIDAS curve is plotted using data from a rising price trend plus rising volume, or a rising price trend plus declining volume? A lot actually, since curves will displace (move about) quite dramatically as a result of these different relationships between uptrending and downtrending price and volume data. Readers interested in this vital area should consult Coles’ Chapter 11 and Hawkins’ Chapter 6 in the book.
To provide one instance here, Chart 1 below is a monthly chart of the DJIA from 1981 to the present with a volume histogram and a 25 month moving average of volume beneath. A volume MIDAS support curve (solid) is launched from the bottom of the 1987 stock market crash and creates a powerful support level for the major 2003 bottom. The height of this curve is explained by Rule #1, that a rising price trend plus volume trend displaces a curve upwards. After the 2007 high, the subsequent 2008 subprime collapse broke this curve and thus provoked the launch from the same 1987 bottom of a nominal curve (dotted), the rationale being that the persistent uptrend in volume would create a significant downwards displacement of a nominal curve from the volume curve with the potential to influence the 2009 bottom. This is precisely what we see in relation to these two vital market bottoms in recent stock market history.
Nominal curves are described as constant volume curves because of the way in which the artificial volume that replaces actual cumulative volume from the market is inserted into the MIDAS algorithm. This is explained thoroughly in the book.
Readers looking at Chart 1 will see straightaway that Second Generation nominal curves don’t replace First Generation curves. Both types of curve are crucially needed. It all depends on the relationship between the price-volume trends a MIDAS analyst is confronted with. It also depends crucially on a MIDAS analyst’s familiarity with these relationships and his ability to apply the rules that are derived from them correctly.
Nominal curves can also be applied to contexts where there is no market volume. An example is the volumeless higher timeframe cash FX markets. See Coles’ Chapter 10.Second Generation curves: replacing the volume weighting in the MIDAS formula with open interest weighting (OIWAP instead of VWAP)
For longer-term futures traders who have a choice of working with volume trends or trends in open interest, it’s also possible to replace volume with open interest instead of constant volume. For example, in the case of Rule #1 above, if in a price uptrend with shallow pullbacks volume is declining but open interest is increasing, a MIDAS analyst would create more accurate MIDAS curves if he replaced volume with open interest. I’ve covered this in Chapter 12.Third Generation (Parallel) curves: Momentum, Volume, Volatility, Relative/Strength, Econcomic
Third Generation curves have appeared very recently and I’ve used the word “parallel” as a catch-all term to describe them.
I first created them in a very late chapter for the book (Chapter 16) where I began to apply MIDAS curves to datasets other than price with the same fractal trend characteristics. The idea was that price-based MIDAS curves often miss many important price inflection points despite their being launched correctly and despite the various innovations above.
In Chapter 16 I applied the MIDAS curves to Granville’s On Balance Volume, but since the completion of the book I’ve applied them to other datasets with considerable success. In particular, two new MIDAS trading setups have emerged I’ve called the Dipper Setup and the Inversion Setup. Anyone interested in further exploring Third Generation curves will find an article in the May 2011 and June 2011 issues of Active Trader magazine. Here I create Momentum Curves (MACD), Volatility Curves (VIX), Relative Strength Curves (R/S), and Economic Curves. For the latter I chose the Baltic Dry Index. Third Generation curves produce excellent trading opportunities, and the Momentum Curves (and OBV-style curves) also remove the age-old problem of timing divergences.
At the outset of this discussion, it was stated that there have been two ways in which the basic MIDAS formula has been manipulated. The first way has been in terms of changing the actual data input for the algorithm. As discussed above, First, Second, and Third Generation curves have all been created as a result of applying this methodology.
The second way in which the MIDAS formula has been manipulated is in terms of changing or adding to the basic maths behind the formula. In the remainder of this post I’ll discuss the innovations that have been made to Levine’s original curves with respect to this second approach.David Hawkins’ Calibrated Curves
It’s not entirely true to say that Calibrated Curves are based on changing or adding to the basic maths in the MIDAS formula, but it’s necessary to include them here because they rely on a different launch point to standard MIDAS curves. As a result, they do process information slightly differently. These atypical launch points are calibrated to subsequent important price turning points and thus capture additional, highly relevant, market moves that standardly-launched MIDAS curves miss. See Hawkins' Chapter 9 in the book.Bob English’s MIDAS Average and MIDAS Delta Curves
A presentation of Bob’s ideas can be found in the second half of Chapter 17, including these curves. MIDAS Average curves (MACs) are created by taking the average of a standard MIDAS support/resistance curve, while MIDAS Delta curves (MDCs) are plotted equidistant from the standard MIDAS S/R curve on its other side. Both types of curve are important when examining longer-term datasets. MACs were developed to help cope with much deeper pullbacks, which are a problem for standard MIDAS S/R curves in so far as deep pullbacks usually break straight through them. MDCs were developed to cope with sharply trending markets that will usually pull away from standard MIDAS S/R curves quickly. As such, MDCs plot in a similar way to the Topfinder/Bottomfinder curves, but without the parabolic component in the algorithm.Coles’ MIDAS Displacement Channel (NYSE:MDC)
Initially, the MDC was developed for sideways markets, where Levine’s original curves would simply move to the middle of the trading range and become ineffective. By displacing the original curve to create one or more upper and lower boundary channels, it was possible accurately to identify support and resistance areas in sideways conditions. However, it quickly became apparent that the MDC is also very effective when trends aren’t trending up and down too sharply. In such conditions, the MDC will also catch price highs in uptrends and price lows in downtrends. This indicator is discussed by Coles in Chapter 14.MIDAS Standard Deviation Bands
This indicator has evolved gradually. Bob English was the first to code the indicator in Tradestation while illustrating its potential on his website, wwwprecisioncapmgt.com. I was the first to code it in Metastock while replacing the VWAP formula with the MIDAS formula. Chapter 15 covers this indicator.Non-curve based MIDAS innovations
There are various non-curve based innovations, but since the results involve various types of oscillator or other related indicators, they won’t be covered here. Chapter 17 covers many of these ideas, especially those by Bob English in the second half.Summary of curves in table
As emphasized in the table, all of the innovations that involve changes or additions to the formula can be created as First Generation, Second Generation, or Third Generation curves.