by Rick Ratchford
The debate rages on. Since I started trading and teaching about dynamic cycles (the combination of fixed cycles resulting in market patterns), there has been many who would question the validity of cycle extraction. Many argue that the cycle turn dates produced by our cycle extraction methods are no better than just picking dates out of a hat at random. But is this true?
I've spent a great deal of energy trying to demonstrate the fact that random date picking is in no way equal to actual cycle extraction of future turn dates. While at the very least cycle extraction requires the use of real historical data to produce these future cycle turn dates, random date picking uses no information at all rather than picking a number out of thin air in order to blanket the market with enough dates to cover every possible trading day.
It has also been argued by some that we who use cycle turn dates are doing the same thing of blanketing the market with turn dates. Because of an allowance for most turns to occur within a day of the turn date, it is claimed that we are simply spacing our cycle turn dates in a way to cover almost every day in order to claim our high accuracy rate. The following information is to help you see the folly and nonsense these critics make about cycles and random dates.
A turn spread is the distance between one turn/swing/pivot and the next turn as viewed on a price chart. On average, a turn/swing/pivot is likely to occur within 3 days. Sometimes you get a 1-bar correction which is a turn, albeit the trend itself doesn't change but only corrects briefly. Sometimes turns occur 2 days apart. Sometimes 3, etc. Less often, turns occur 5, 6, 7 or more trading days away from the last turn.
Understanding what a turn/swing/pivot is and what is a spread is very important to come to understand the validity of cycle dates as opposed to the folly of random dates.
In simple Gann terms, a swing or pivot top is any price bar that makes a new high (in relation to the prior price bar high) and is then followed by a price bar that will make a lower low (in relation to the prior price bar low). Always keep in mind that you are comparing any price bar with the bar just prior.
Take a look at chart one above. The price bars that are considered swing/pivot tops are marked in red. The swing/pivot bottoms are marked in green. From here on out, I'm just going to refer to these turns as 'swings'.
Notice that one of those swing tops marked on Chart 1 is actually two price bars. Why? Because these both had the same high price and both were higher than the prior bar (which was a swing bottom, by the way). It happens.
The distance between any two swings is called the 'turn spread'. It's not a well known term because we just made it up in order to have a way of referencing the time element for this discussion. If you study Chart 1 carefully, you will quickly see that these swings do not occur at equal intervals of time. The turn spreads are all different. Some are small and some are longer.
Now let's consider the problem with random date picking. Since we know that all markets form swings at varying time spreads, how can someone determine these time spreads by picking dates randomly out of a hat or thin air? Logically it cannot be done. Can you guess that Coffee is going to turn tomorrow, then not again for 5 trading days, then quickly after that in 2 days, then...? Do you see the folly in that? In addition, how much confidence would you place in that forecast?
Proponents of the random argument suggest that all they have to do is space out their random dates every 3 days and that by doing so they will catch every turn in the future. In this they are correct because no day would be left unaccounted for.
But can you see the problem with this already? Let's use a real market situation to demonstrate how ridiculous it would be to do this.
Take a look at Chart 2 above. Now starting at any bar of your choosing, imagine marking off every 3rd day bar. Would you say that the result would accurately depict this chart pattern? Of course not.
Now let's take a look at the same chart again but this time I am going to place an arrow on those price bars that we calculated in advance and reported on our Ceppro Report for this market.
Chart 3 above has plotted the actual forecasted cycle turn dates that appeared as early as our 11/04/2005 Ceppro Report. Notice how these dates more accurately expose the market turn spread. Rather than arbitrarily assigning a date very 3 days, cycles give us REAL information that is calculated from REAL price data. Just note the turn spread of 11/30 to 12/9. This was a move of at least 7 trading days (10 calendar days) from one turn to when we are expecting the next. At the time of this writing, we have yet to see what will become of our 12/9 cycle date. But I believe there is enough information here to clearly demonstrate how cycle extraction provides information we can use.
Now you may be thinking that I picked an optimum chart to make this argument. Well, you would be correct. But I did this for logical reasons. For one, I needed an example that had an unusally large time spread to demonstrate the differences between actual cycle extraction and just picking random dates. Also, I would challenge anyone to pick a market at random and then forecast in advance another two turn dates that would have a spread of 7 or more trading days. The odds of doing so would be astronomical. Yet, we've done it often using our cycle extraction method.
Take a look at this next chart of Soybeans.
Chart 4 above is from a daily chart of Soybeans (12/9/2005). The swings tops and bottoms on this chart are identified by the green rectangle around them. Would plotting a random date every 3 bars accurately depict the pattern of this market? Again, this would be a no.
Yet on our Ceppro Report, provided to our members in advance, the following cycle dates were reported. 11/23 - 11/28 - 12/2 - 12/7. This more accurately describes the market swing pattern that has occurred in Soybeans. 11/23 is just one bar of 11/22. And 12/2 is one bar prior to the 12/5 top. In addition, our forecast of 11/28 - 12/2 is time spread beyond 3 days and we can see that the market did in fact produce a time spread beyond 3 days (11/28 - 12/5). Also note that another swing occurred on 12/7 that quickly followed the last. Our forecast called this correctly as well.
Now this discussion is in no way to suggest that we get these perfect all the time. We're not looking for perfection, although we continue to work toward that goal. These examples are to simply demonstrate that when it comes to turn dates, there is just no comparison to cycle extracted turn dates from real price data and just picking random dates out of thin air. And after you have actually seen many, many examples of future market turn patterns being forecasted in advance, your confidence in using them for timing your trades will increase. Naturally you could never develop that kind of confidence in random dates knowing they are, frankly, random.