Scoring, the State Machine, and the No-Trade Rules

Eleven lessons of separate pieces. This one puts them in order and gives them a shape you can run under time pressure: a score that records what you saw, a state machine that names what you are doing, and a list of conditions under which you do nothing at all. Start with the part most courses would put last, because it changes how you read everything after it.

What we measured about scoring, and what we did not find

The standard way to teach this is to build a quality score from the setup's components, show that high scores win more often, and hand you a threshold. We tried the first half and the second half did not arrive.

We reconstructed a compression-quality score over 490 SPY sessions and 489 first breakout attempts out of compressed ranges, 09:45–14:30 ET, and sorted outcomes by it. Sorting did not separate them: failure rates ran 71–77% across score buckets, against 75% for the sample as a whole. The broader compression finding has the same shape — 74% failure whether the range was compressed or not, outcome difference −0.137W with a 95% confidence interval of [−0.330, +0.056] straddling zero.

So be precise about what a score is for. It standardises a decision you would otherwise make from a feeling, and it forces the sequence into the right order. It is not a probability estimate, and on our data it did not raise the hit rate. A grade records what you observed. It does not forecast what happens next.

There is a related trap on the other side of this. If a scoring system appears to lift your win rate on a few hundred trades while you tune eight or nine weights, the likeliest explanation is that you fitted the noise. Nine free parameters against 300 observations will find structure in anything.

The eight components

Each is a lesson of this course compressed to one line. Score each before you need it, not while the bar is printing.

Six of the eight are readable before the boundary breaks, and that split survives the negative result above: whatever the score predicts, computing it early lets you stand down before the break rather than after, having spent attention and no money.

Two components behave differently. Clean traffic is a veto, not a score — with no room between entry and the next obstacle, a perfect setup still cannot reach a worthwhile target. And boundary pressure can run negative: a range with rising lows that breaks downward is breaking against its own internal pressure.

The state machine

The score records. The machine names what you are doing right now and what has to happen before you may do the next thing.

Every arrow carries a condition you wrote down in advance. The value is not the diagram. It is that at any moment there is exactly one answer to "what am I waiting for," and it is not a matter of opinion.

Two confusions that cause most of the losses

ARMED is not entry. Armed means everything computable has been computed: boundaries drawn, entry model chosen, stop priced at a named level, targets set, size derived from the stop distance, traffic measured. Nothing is on. A trader who begins the analysis at the trigger has already lost the trade, because there is no time to do it there; a trader who is armed has a two-second confirmation to make. Most of this state's value is in the setups it kills — you disarm at no cost and go find another.

IMPULSE is not validity. A break that will fail and one that will hold both produce a move. Across those 489 attempts, breaks that eventually failed still reached a median best price of 0.19W at five minutes; the ones that held reached 0.43W. The failures were worth less, not nothing. Being in an impulse tells you almost nothing about whether the break was real, which is why acceptance is a separate state with its own evidence, and why the 489-breakout study is the piece of this course to reread first.

The pre-trade checklist

Nine lines, assembled from the whole course. It runs in the bars between the structure arming and the boundary breaking.

CONTEXT     higher timeframe direction, VWAP side, session window
COMPRESSION range coiling, at least three clean touches, volume holding up
BOUNDARY    outer level and inner level drawn on the chart as lines
ARMED       every item below completed before the break, not after
ENTRY       model selected (trigger / retest / probe-and-add) and written down
TRAFFIC     nearest obstacle measured — runway versus risk        VETO
STOP        at a nameable level, and the position sized from that distance
TARGET      first target placed with the entry, in the same action
MANAGEMENT  stop moves to structure or does not move; never to your entry

Two lines carry most of the weight. A level you have drawn is one you compare price against; a level in your head is one you renegotiate. And the first target goes in with the entry, because it is the step that disappears exactly when the move looks strongest.

The no-trade conditions

Any one of these and the answer is no, regardless of how good the rest looks.

• The clouds are flat and tangled. No timeframe is in charge, so there is no direction to align with.
• Price is crossing its mean repeatedly. That is two-sided trade, the opposite of the condition a break needs.
• The structure is unreadable. If you cannot name the range's edges out loud, you do not have a boundary — you have a shape.
• The only entry is straight into an opposing level. A prior close, an untouched high or a round number sitting on your target.
• There is no runway. The obstacle is nearer than the risk. A veto on its own.
• You cannot watch the first five minutes. Median time to a confirmed failure is four minutes. If you will not be at the screen while the trade declares itself, you are not managing it. You are holding it.

The course in one page

Compression creates the condition. The boundary defines the battlefield. The break releases the impulse. Acceptance decides continuation. Traffic defines the opportunity. Risk defines the size.

And the thesis this course was built to deliver: you never had to know whether it was real. You cannot know at the trigger — the impulse appears on both kinds of break, and the scoring result above says setup quality did not tell you either. What you can do is enter where being wrong is cheap, take something out of the move while it exists, and let the market's behaviour after the break decide whether you are still in it thirty minutes later. That is not a weaker form of prediction. It is what remains once prediction is off the table.

What you should take from this lesson

1. A score standardises your decision; it does not raise your hit rate. On 489 breakouts, sorting by a reconstructed compression-quality score left failure rates at 71–77% across buckets. Use it to record and to enforce sequence, never as a probability.
2. Six of the eight components are readable before the break. That is what makes standing down cheap — you disarm before the trigger, not after the loss.
3. ARMED is not entry. Decisions are made in advance so the trigger is a confirmation, not an analysis.
4. IMPULSE is not validity. Failed breaks reached 0.19W at five minutes, held breaks 0.43W. Movement is not evidence; acceptance is a separate question.
5. The no-trade list only works if it is read before the trade. At the moment a rule would bind you are looking at a chart, and a chart is more persuasive than a memory.

Now go back to Lesson 1 and run the whole sequence once on a live chart, without taking a trade.