Price structure

Why the obvious level was only obvious afterwards

This is the most valuable lesson in the module and the least comfortable. Every chart you have ever learned from was finished when you saw it. The one in front of you tomorrow will not be, and almost everything that felt obvious came from the difference.

About 12 minutes to read. Free, no account.

Two charts, and they are not the same chart

When you study a chart you are looking at a resolved object. The swing points are all confirmed, the levels that mattered are visible because the market came back to them, and the ones that did not matter are invisible because nothing happened there. Your eye is drawn to the prices the future made important.

When you trade you are looking at an unresolved object. The right-hand edge is now. The most recent pivot is not confirmed yet. You cannot see which of the twenty candidate prices on the screen the next week will pick out, because it has not picked yet.

Those are not the same task, and skill at the first does not transfer to the second. Reading a resolved chart is comprehension. Reading an unresolved one is prediction. Almost all chart education is the first thing wearing the clothes of the second.

What bar 51 knew, and what you know

swing rule                       =  3 bars either side
the high printed on              =  bar 48
confirmed at the close of        =  bar 51
price then, below the high       =  4.63%

the chart you are studying ends at bar 200.
bars of information it has that bar 51 did not
                                 =  200 - 51  =  149

you are judging a decision made with 51 bars,
using 200.

A hundred and forty-nine bars is not a detail. It is the entire reason the level looks obvious, and there is no amount of experience that gives it to you at bar 51.

The selection problem, with numbers

The sharpest way to see the distortion is to count how many candidate levels a chart offers and how many survive into the picture you remember.

Candidates versus keepers

bars on the chart                =  200
a swing point roughly every      =  8 bars
pivots available                 =  200 / 8   =  25

levels drawn in the finished analysis   =  3
share of the candidates kept     =  3 / 25    =  12%

the other 22 pivots were discarded - using information
that did not exist when they formed.

Twenty-two pivots went in the bin because nothing came of them, and the only thing that marked them as unimportant was the future. In front of a live chart all twenty-five are equally available and equally plausible. The three-level chart is not a simpler version of the twenty-five-level chart. It is a different chart, assembled with the answer in hand.

The same arithmetic applies to any pattern. A chart with twenty-five pivots contains hundreds of pairs of pivots, so it is always possible to find a pair that forms a double top, a divergence, a trend line with three touches. Finding one is not evidence. Finding one before the chart resolved would be.

Why the lag makes this worse than it already sounds

Lesson 1 established that a pivot is confirmed only after the bars on its right-hand side have printed. Put that together with the selection problem and you get the real position you trade from:

The pivot you most want to use - the most recent one - is the one that is not yet confirmed. Structure is reliable in inverse proportion to how current it is.
By the time it is confirmed, price has moved away from it by a distance you can calculate, as the 4.63% above shows. The good entry was inside the window where you did not yet know.
If you act before confirmation, you are acting on a pivot that may not become one - which is allowed, and is a different and more honest decision than pretending the structure was there.

This is not an argument against using structure. It is an argument for knowing which of these three situations you are in, every time, because they carry completely different information and they feel exactly the same.

So how is structure still useful? By being written down first

Here is the part that keeps this lesson from being merely cynical. Structure remains genuinely useful, and the thing that makes it useful is a change of grammar: stop using levels to explain and start using them to commit.

An explanation is produced after the fact and can always be produced. A commitment is a price and an action named before the fact, and it can be wrong - which is the only property that makes it worth anything.

Write the rule before the chart. Swing rule, tolerance band, which edge, wick or close. Decided in advance, these turn a drawing into a procedure that produces the same levels for you as for anybody else.
Write the level and the time together. "Resistance 110.00, marked at 14:00 on the 3rd." A level without a timestamp cannot be checked, and a level you cannot check will be remembered as having worked.
Say what you will do if price arrives, before it arrives. That is the entire practical value of a level: it moves a decision from a moment of stress to a moment of calm.
Count everything the rule produced, not what you noticed. Including the dull instances. Especially the dull instances - they are the majority and they are what the average outcome is made of.

Done this way, structure is doing a job no indicator can do: it gives you specific prices, fixed in advance, at which you have already decided something. Whether the level "works" matters much less than whether you had a plan at it.

The covered-chart test, and the three questions

There is one test for any structural claim, it takes ten seconds, and it is the same test the RSI lesson applies to divergence: cover the right-hand side of the chart and ask whether you would have drawn this. Not whether you could have. Whether you would have, with nothing to the right of the line.

Most lines fail it. That is the useful result, not a discouraging one - a chart with three defensible levels on it is far more tradeable than one with twenty that were each obvious in retrospect.

And three questions to ask of any structural observation, your own or somebody else’s:

When was this knowable? Name the bar. If the answer involves bars to the right of the one you are pointing at, the observation is a description of the past.
What was the rule, and was it fixed beforehand? If the swing rule or the tolerance was adjusted until the picture looked right, the picture is an output of the adjusting.
How many times did this rule fire and nothing happen? If that number is unknown, the claim has no size. "It worked here" is one instance of an unknown total.

Honest about the honest version too

One last turn of the screw, because a lesson about self-deception that exempts itself is not worth much. Writing the rule down first removes the worst of the distortion. It does not remove all of it.

You chose the rule after years of looking at resolved charts, so your idea of a sensible swing rule is already shaped by outcomes you have seen.
If you test ten rules and keep the best, you have selected on the answer again, just one level up. Fewer rules, decided for a reason you can state, is the only defence.
A level that has worked for you recently will feel more credible than the identical level on a coin you do not follow. That feeling is not information.

None of this makes structure useless. It makes it modest, which is a much better place to trade from than certainty. The chart tells you where people previously changed their minds. What you do when price gets there is the only part you control, and the only part worth deciding in advance.

The terms, defined

Hindsight bias: Judging a past decision with information that arrived afterwards. On a chart it is structural, because the chart itself contains the information.
Selection effect: Keeping the instances that the outcome made visible. It is why three levels appear on a finished chart that offered twenty-five.
Resolved chart: One whose right-hand side you can see. Every chart you learn from is resolved; no chart you trade is.
The covered-chart test: Hiding everything to the right of a drawing and asking whether you would have drawn it. The cheapest honesty check there is.
Pre-commitment: Naming the price and the action before the price arrives. It converts a level from an explanation into a decision that can be wrong.
Complete record: Every instance a written-down rule produced, dull ones included. Without it, "it worked here" has no size.

What this means in practice

Before trusting any level, cover the right-hand side of the chart and ask whether you would have drawn it with nothing to the right.
Name the bar on which each observation became knowable. If that bar is to the right of the event, you are describing history.
Fix the swing rule and the tolerance before you look, and resist testing several and keeping the one that fits.
Timestamp every level you mark. An unrecorded level is always remembered as having worked.
Decide what you will do at a price before price gets there - that is what a level is actually for.
When you count a pattern, count the boring instances too. They are the majority, and they are most of the answer.
Ask the three questions of anybody showing you a chart, including us: when was it knowable, what was the rule, where is the complete record.

Nothing in this lesson is financial advice or an instruction to trade. See the risk disclaimer. We never recommend a leverage setting - that decision is yours.