Teaching You to Trade Stocks 77: Re-distinguishing Some Concepts

2007/9/5 23:24:01

Sorry, got home late. Only now have I finished writing up the lesson to post.

Dreamtalk can't go on endlessly — now back to the technical side. Let's re-distinguish some concepts, because certain concepts are so fundamental that if you don't get them straight, everything after will be one word: chaos. To avoid chaos, and hopefully for the last time, let me re-examine those most basic concepts from before.

Let me first use the example of gaps to illustrate the proper thinking methodology for correct prediction based on rigorous classification. All predictions must be based on rigorous classification — this is the most fundamental way of thinking. Otherwise, indulging all day in pure probability games is just treating boredom as entertainment.

Take gaps as an example — using upward gaps for illustration. First, you need to give gaps a clear definition — one that facilitates classification, because only a clear definition enables clear and complete classification. What is a gap? On a given K-line chart period, it's an interval between two adjacent K-lines where no transactions occurred. For example, on the Shanghai Index daily K-line chart, between July 29, 1994 and August 1, 1994, the interval [339, 377] had no transactions. So we say [339, 377] is a gap. And gap filling means that after the gap appears, the entire gap interval eventually sees transactions again. This process may occur on the very next K-line, or it may never occur again. For example, the [339, 377] gap — while we dare not say it will never be filled, before the stock market ceases to exist, there probably won't be much chance. Those of us like this ID who were fortunate enough to experience that day are truly blessed. This ID still remembers: the stocks this ID bought in the largest quantity on July 29, 1994 were — in Shenzhen, old Xingyuan; in Shanghai, big Flying Music.

Based on whether a gap is filled, we can construct a classification of trend strength. One: not filled — this is obviously strong. Two: filled but then continuing to new highs or new lows — this is neutral. Three: filled but unable to make new highs or new lows, thus causing a reversal of the original trend — this is weak.

Generally speaking, breakaway gaps are rarely filled. Continuation gaps — those during trend extensions — have roughly a 50/50 chance of being filled, but they always continue to new highs or new lows, meaning they are at least neutral. Once a gap is filled and no longer makes new highs or lows, it signifies that the original trend has reversed. This is the characteristic of an exhaustion gap. Once this situation occurs, there will definitely be at least a relatively large-level adjustment, of a level at least greater than the level of the trend that was underway when the gap appeared. That is, an exhaustion gap during a daily-level trend will produce at least a weekly-level adjustment. And an exhaustion gap during a 5-minute level trend will produce at least a 30-minute level adjustment.

Note: the levels here have nothing to do with the K-line chart period where the gap appears — they only relate to the levels of trend types in this ID's theory. Different period K-line charts and trend levels are like microscopes of different magnifications and the objects they observe. This metaphor has been repeated over and over — stop confusing them.

Obviously, a gap on the daily K-line chart will have corresponding gaps on any sub-daily period K-line charts, but filling the daily chart gap doesn't necessarily fill gaps on sub-daily period K-line charts. Additionally, gaps within consolidation trends differ in nature from gaps within trending movements — they are ordinary gaps. Such gaps are generally filled and don't have much analytical significance. The only significance is that within hub oscillation, there is a target: during the pullback process, it's almost certain that the price will at least pull back to fill the gap.

Done with gaps, now let's talk about the issues of fractals, strokes, and line segments.

Fractals need no further explanation. By definition, as long as you understand the inclusion relationship, even someone as clueless should be able to trace along. Without any inclusion relationships, 3 K-lines can determine one fractal. But note: any adjacent fractals must satisfy the associative law — that is, no K-line can belong to multiple different fractals simultaneously. This is not allowed.

Generally, for those unfamiliar with the method, you should first mark all fractals on the chart being analyzed according to the definition, after processing inclusion relationships and the associative law. Top fractals can be marked with downward arrows and bottom fractals with upward arrows — making everything clear at a glance.

With the above foundational work done, the chart can be viewed as containing only these fractals — the K-lines between fractals can be temporarily ignored. The next task is determining strokes. A stroke must go from one top to one bottom (or vice versa), and between the top and bottom, there must be at least one K-line that doesn't belong to either the top fractal or the bottom fractal. Of course, there's another obvious requirement: within the same stroke, the interval of the highest K-line in the top fractal must have at least some portion higher than the interval of the lowest K-line in the bottom fractal. If even this condition isn't met — meaning the top is entirely within the bottom's range or the top is lower than the bottom — this is obviously unacceptable.

Therefore, in the process of determining strokes, the above conditions must be satisfied. This allows unique determination of stroke division. The uniqueness of this division is easy to prove: suppose there are two divisions that both satisfy the conditions. For these two divisions to differ, there must be some point where the first N-1 strokes before it are identical for both, and from the Nth stroke onward, the first difference appears. N can equal 1, meaning they differ from the very beginning. Then the fractal at the end of the (N-1)th stroke obviously has the same nature for both divisions — both are tops or both are bottoms. For the case where it's a top, the Nth stroke's bottom must correspond to different bottom fractals in the two divisions; otherwise, the stroke would be the same for both, which is contradictory. Since fractal division is unique, these two different bottoms must have a sequential order and relative height difference, and between these two bottoms, there cannot be a top — otherwise, this wouldn't be a single stroke.

If the earlier bottom is higher than the later bottom, then the earlier division is obviously wrong, because under that division, the stroke hasn't been completed — a bottom followed by a lower bottom without an intervening top is the most classic case of an incomplete stroke.

If the earlier bottom is not lower than the later bottom, then if before the next top fractal appears, a bottom fractal lower than the earlier bottom emerges, both divisions are incorrect — the strokes they've divided are incomplete. If before the next top fractal appears, no bottom fractal lower than the earlier bottom emerges, then the next top fractal must be higher than the earlier bottom. Therefore, the earlier bottom and this top fractal form the new (N+1)th stroke. Thus, both the Nth and (N+1)th strokes have a unique division — contradicting the assumption that different divisions exist starting from the Nth stroke.

The case where the fractal at the end of the (N-1)th stroke is a bottom can be proven similarly.

In conclusion, clearly, stroke division is unique.

From the uniqueness proof above, you actually already know the procedure for dividing strokes:

One: Identify all fractals that meet the standard.

Two: If two consecutive fractals are of the same type — for tops, if the earlier one is lower than the later one, keep only the later one and cross out the earlier one; for bottoms, if the earlier one is higher than the later one, keep only the later one and cross out the earlier one. Cases not meeting the above, such as equal ones, can be kept for now.

Three: After Step Two's processing, among the remaining fractals, if adjacent ones are a top and a bottom, they can be connected as one stroke.

If adjacent remaining fractals are of the same type, there must be earlier tops no lower than later tops, and earlier bottoms no higher than later bottoms. After consecutive tops, a new bottom must eventually appear — connect the first top in the consecutive sequence with this new bottom as a new stroke, and cross out all the tops in between. After consecutive bottoms, a new top must eventually appear — connect the first bottom in the consecutive sequence with this new top as a new stroke, and cross out all the bottoms in between.

Obviously, through the above three steps, all strokes can be uniquely divided.

With strokes established, next comes line segments. The most basic principle of line segment division is that a line segment must consist of at least three strokes. This is perfectly obvious — otherwise, if a single stroke could constitute a line segment, what difference would there be between strokes and line segments? As for why two strokes can't constitute a line segment, the reason is even simpler: with two strokes, the fractals at the two ends of the line segment would necessarily be of the same type. Like strokes, a complete line segment's two endpoints can't have fractals of the same type. In other words, like strokes, a line segment cannot start from a top and end at a top, or start from a bottom and end at a bottom. From this, it follows that the number of strokes contained in a line segment is always odd.

Moreover, the first three strokes at the beginning of a line segment must have overlap. Starting three strokes without overlap cannot form a line segment.

Additionally, a line segment must be broken by another line segment for its completion to be confirmed. For the first type of line segment division: if after the first stroke produces a stroke-level break, the next stroke immediately makes a new high, and the stroke after that doesn't even touch the stroke that caused the break, then obviously this cannot constitute a line-segment-level break of the line segment, because these latter three strokes have no overlap and cannot possibly form a line segment.

Using the first type's diagnostic method, this becomes even clearer: the above situation cannot possibly form a fractal in the characteristic sequence, so it obviously can't signify completion of the line segment.

Furthermore, when a line segment is broken by another line segment, it must not be broken by a line segment of the same direction. That is, a line segment starting with an upward stroke cannot be broken by a line segment also starting with an upward stroke — it must be broken by a line segment starting with a downward stroke.

The second type of line segment situation actually includes this case. That is, under the first type's criteria, line segment A isn't broken by the following line segment B, but the subsequent line segment C breaks line segment B. Therefore, line segment B is complete, and naturally line segment A must also be complete. Note: line segments A, B, and C here are first divided using the associative law principle — with the contents within brackets satisfying the basic properties of line segments. Before this breaking relationship is confirmed, these are just provisional labels.

Everyone has surely noticed that in the second type's situation, special emphasis is placed on the fact that the second characteristic sequence — which actually corresponds to line segment C's breaking of line segment B — no longer distinguishes between first and second types. This is actually a simplification method. Why?

If we insist that the final break of a line segment must fill the characteristic sequence gap, then if line segment C's relationship to line segment B is also the second type, line segment C's interval must lie between line segment A's characteristic sequence gap and line segment B's characteristic sequence gap. Following this logic, eventually there must be some line segment X that fills the characteristic sequence gap relative to its predecessor; otherwise, these line segments' intervals would shrink infinitely, eventually converging to a point — which is obviously impossible. Anyone who's studied limits should understand this. So, in a string of line segments where each is the second type relative to its predecessor, there must ultimately be a first-type break, and working backwards, there must be continuous breaking among this string of provisional line segments.

Precisely because of this, in the second type's second characteristic sequence judgment, we no longer distinguish between first and second types. This avoids the hassle of working backwards through a string of converging line segments. Mathematically, this is of course absolutely rigorous, but operationally it's too cumbersome. Moreover, such special cases are rare, making it even more unnecessary.

So why distinguish the second type at all? Because we don't want situations at the line segment level where small levels transition to large levels in an uncertain manner. Using the second type resolves this problem.

There's a complex situation that appeared in today's 80-83 division. For 80-81, a first-stroke break appeared, followed by a movement A that meets line segment standards but didn't make a new low. Obviously, this can't be considered a continuation of the original line segment, but it can't be considered a line segment break either. Why? Because there aren't three qualifying strokes. Then, a rebound that also meets line segment requirements occurs, after which the price turns and continues to new lows. Here there's a subtle difference: if this rebound were only one stroke, it wouldn't break movement A, and the subsequent new low would mean movement A still extends — so movement A would be a continuation of the original 80-81.

But the actual problem is that this rebound broke movement A at the line segment level. Therefore, saying movement A still extends is obviously incorrect, so the subsequent movement has no relation to movement A. Therefore, the only reasonable division is to combine the first break stroke, movement A, and the rebound into one line segment — this fully satisfies the definition of a line segment. Hence, we get 81-82.

Line segment division is actually not difficult at all. The key is to work from the definitions. Moreover, using the two types of rules specified for line segment division, it's not hard to prove that line segment division is also unique.

If you have any questions, please continue asking. Getting things thoroughly clear is what matters most. These remedial lessons will be given from time to time, but they can't be every class. Next time, we'll be covering new content. If old questions accumulate to a certain degree, there will be another remedial session.