---

name: trendline-eval description: A skill that evaluates the validity of a descending/ascending line (trendline) in the 200EMA strategy. Always use this skill when the user uploads a 1m, 5m or 1h chart image (e.g. XAUUSD or USDCAD) and asks things like "evaluate this line", "does this meet the entry criteria", "judge this ascending/descending line", or "check this against the 200EMA strategy"; when the user names a Notion trade-log page and asks for an evaluation; or when the user asks for a pre-entry chart check, a trade review, or feedback on how a line was drawn. Also use it when a chart image has a trendline drawn on it and an evaluation or judgment is requested.

200EMA Strategy Descending/Ascending Line Evaluation Skill

Purpose of this skill

In STEP 3 of the 200EMA strategy (entry on a Dow-theory line break), structurally evaluate whether the drawn descending/ascending line is "a valid line worth entering on," using a fixed set of 5 observation elements and 8 evaluation elements (4 elements per axis × 2 axes), and output an overall verdict.

The core of the strategy that the evaluation assumes:

• A genuine corrective wave forms clear swing highs and lows as counter-trend traders enter and exit. A break of the line connecting those swings = the moment the counter-trend traders give up = the resumption of the impulse wave. That is exactly why the break means something.
• A slow grind down / slow grind up or a range is not a correction but accumulation (build-up), and its eventual resolution tends to come on the 200EMA-break side (= the SL side).
• A steep line captures only a small sub-wave at the tail end, not the whole correction. Its break is merely the end of one micro leg; the larger corrective structure tends to continue and move toward the SL.
• How "easy the line is to draw" is itself diagnostic. The break of a forced line is noise.

For the background of the strategy as a whole (the STEP 0 slope filter, etc.) refer to references/methodology.md, and for the characteristics of the model examples refer to references/ideal-examples.md, as needed.

About the execution timeframe

• The default execution timeframe is 5m. However, entries are also made on 1m and 1h.
• When the execution timeframe is 1h, apply this skill's criteria (5 observation elements, 8 evaluation elements, scoring tables) to 1h as is. No note about a timeframe mismatch is needed.
• The reference timeframes for STEP 0 (higher-timeframe filter) are 15m and 1h for 1m/5m entries, and 4h and daily for 1h entries. Use this mapping when mentioning them.

Conventions for reading the input chart

The chart images (TradingView) the user provides for evaluation use a different color scheme from the model images. Read them according to the following conventions:

• Green candles = bullish, black candles = bearish
• Dark blue = 200EMA, light blue = 20EMA
• Black diagonal straight line = the descending/ascending line (the subject of evaluation)
• Green/red box: the boundary is the entry price, the red side is the SL and the green side is the TP (same as the model images)
• Green vertical line = where a golden cross occurred; orange vertical line = where a dead cross occurred (the marker that fixes the directional bias in STEP 1)
• Green circle = the first 20EMA touch and the first 200EMA touch after a golden cross. Orange circle = the same first touches after a dead cross (the marker for STEP 2)
• Gray horizontal lines / wide boxes = higher-timeframe AOI. Irrelevant to the execution-timeframe line evaluation; ignore them
• The timeframe panel at the top right of the screen (arrows for 15m/1h/4h/D, etc.) shows the golden/dead cross state of each timeframe. Note that this is "directional bias" information and is different from the "200EMA slope" that STEP 0 requires. It may be mentioned as reference information but is not included in scoring

Evaluation workflow

Observation phase (fixed 5 elements)

Before scoring, always extract and put into words the following 5 elements. Both analysis and output are always done in terms of these 5 elements.

Element	What to extract
① Impulse wave	Direction / duration / price range / presence of an EMA cross (= the basis for the N calculation and scale comparison)
② Corrective wave	Origin (the end pivot of the impulse wave) / duration / retracement depth / whether it reached the 200EMA
③ Pivots	List of swing highs/lows recognized within the corrective wave (recognition criterion: accompanied by a pullback/bounce of at least 30% of the correction range)
④ Line	Location of the origin / number of pivots it passes through / number of price reactions (touch → bounce)
⑤ EMA relationship	Slope of the execution-timeframe 200EMA / location of the touch markers (circles) / squeeze between the line and the EMA (whether the confirmation requirements are met) / number of candles in the contact zone / quality of the 200EMA reaction (bounce, intermediate, clinging, undetermined)

Scoring phase

Each axis is judged on 4 fixed elements. Mark each element ○ (clearly met) / △ (partial / insufficiently confirmed) / × (not met), and determine the axis score with the conversion table.

Axis 1: Swing clarity (quality of the wave structure)

The axis that judges "is this pullback a genuine correction, or a build-up?"

Key	Element	○ criterion	Typical ×
S1	Wave rhythm	The corrective wave forms clear waves with pullbacks/bounces (not a grind or a range)	A slow grind down/up with candles continually overlapping, or a range
S2	Pivot reaction	The line connects 2 or more pivots that have reacted (flawless if there is a reaction at a 3rd point)	Price has never reacted; effectively 1 pivot
S3	Reversal structure	The pivots form a named reversal structure (double top/bottom, head and shoulders)	No structure can be recognized
S4	Uniqueness	Almost no other equally valid line can be drawn	Three equivalent lines can be drawn; tempted to draw through the middle of a cluster

The core elements are S1 and S2. If either is ×, Axis 1 is automatically 2 or lower.

Axis 2: Angle and scale fit

The axis that judges "does the scale of this line match the scale of the move we are aiming for?"

Key	Element	○ criterion	Typical ×
K1	Origin match (wholeness)	The line's origin is at the origin of the corrective wave (the end pivot of the impulse wave) and it covers the whole correction	Starts in the middle or at the tail of the correction and captures only the final accelerating sub-wave
K2	Relative slope	The line's slope is clearly gentler than the slope of the preceding impulse wave (corrections are slower than impulses)	As steep as or steeper than the impulse (scale mismatch, or suspicion of an impulse in the opposite direction)
K3	Sufficient scale	The correction has developed over enough time / candles to contain multiple swings, and the pullback has not gone beyond the origin of the impulse wave (not a trend negation)	A tiny line containing only a few candles. Or the pullback clearly exceeds the origin of the impulse wave (higher high/lower low) and the correction has turned into a trend negation
K4	Break location	Retracement ratio r ≤ 0.33 (break in the lower 1/3 of the correction range = near the EMA bounce)	r > 0.50 (break in the upper half of the correction = buying the pullback rather than the bounce; far from the EMA, SL far away) *Judge per "Objective judgment of K4" below

The core elements are K1 and K2. If either is ×, Axis 2 is automatically 2 or lower.

Important notes:

• It is normal for the corrective wave to take longer to form than the impulse wave; do not deduct for it. Impulses are fast and corrections are slow — if anything, a correction that takes time is a hallmark of a genuine correction. In the model examples too, the correction often takes as long as or longer than the impulse.
• Angle depends on the aspect ratio of the image, so do not judge by absolute angle (XX degrees). Always judge by relative comparison with the slope of the preceding impulse wave (K2).
• The essence of "steep" is not speed but scale mismatch: the line is based on a sub-wave that is too small rather than on the whole correction (K1 and K2 are two manifestations of the same pathology).

Do not confuse "retracement depth" with "trend negation" (★ preventing recurrence of misjudgment):

• Retracement depth is a descriptive metric measured as pullback size ÷ price range of the preceding impulse wave in %, and its reference is always "the preceding impulse wave." Do not write "deep" without stating what it is deep relative to. A 50–78% retracement is normal for a correction, and depth alone is not a deduction. Just write the % in Observation ②; do not make it × on K3.
• K3 × = trend negation is recognized only by the measurable event that the pullback clearly exceeded the origin of the impulse wave (higher high/lower low). "Close to the origin" or "deep" is not ×.
• Reaching the 200EMA is a premise of this strategy (the destination of the pullback) and is not in itself a deduction. "It reached the EMA = deep/dangerous" is a category error. Measure depth as a % of the impulse wave, separately from whether the EMA was reached. Danger at the EMA is evaluated by the "quality of the reaction (R)" and the "special rule," not by "depth."

Objective judgment of K4 (break location)

K4 judges "is the break near the 200EMA touch?" not by subjective "near/far" but by position normalized by the scale of the corrective wave. The aim is to catch the break = 200EMA bounce near the touch, so that the SL (the final high/low of the correction) is close and the RRR stays within 1:1–1:3. This quantification prevents mistaking "a break within the same corrective structure" for "near the touch" (preventing recurrence of past misjudgments).

The 4 points to measure (long; invert for short):

Key	Meaning
A	Origin of the correction (end pivot of the impulse wave = the high, for a long)
B	Extreme of the correction (the low of the 200EMA touch = SL anchor, for a long)
E	200EMA value at the time of the break
X	Break price (= entry price)
R	|A − B| = width of the correction range (the normalizing denominator)

Primary metric: retracement ratio r = what fraction of the correction range has already been retraced at the time of the break.

• Long: r = (X − B) / R / Short: r = (B − X) / R

r	K4	Meaning
r ≤ 0.33	○	Break in the lower 1/3 of the correction range. Near the EMA bounce, SL is close (measured on model examples: r≈0.29–0.33)
0.33 < r ≤ 0.50	△	Lower half but toward the middle. Somewhat late; RR degrades but is acceptable
r > 0.50	×	Break in the upper half of the correction. Buying the "pullback" rather than the bounce; far from the EMA, SL far away

Secondary metric: EMA deviation ratio g = \|X − E\| / R. g ≤ 0.20 reinforces ○; g > 0.40 reinforces ×.

RR cross-check: If the RRR with the SL slightly beyond B and the TP at the impulse wave's N-calculation value falls below 1:1, K4 is × regardless of r (the entry is too late and the edge is gone). Even if the RRR is within range, note in the supplement that the higher r is (= later), the more the RR is compressed toward 1:1.

Boundary handling: If r is within ±0.05 of a band boundary, decide using the secondary metric g and the RRR. Always judge by this normalized metric rather than the absolute angle (because it does not depend on the image's aspect ratio). A/B/E/X may be approximated from the chart image, but must not contradict the values read in Observations ⑤ and ④.

Axis score conversion table (common to both axes)

Axis score	State of elements
5	All 4 elements ○
4	Three ○ + one △ (no ×)
3	Two ○ + two △ (no ×), or exactly one × on a non-core element (S3/S4, K3/K4) (no other ×; stays at 3 regardless of △)
2	One × on a core element (S1/S2, K1/K2) (no other ×)
1	Two or more ×, or a forced line / a complete range

Judge from the top and take the first row that applies. Note that the handling of × branches on which element is ×:

• A single × on a core element (S1/S2, K1/K2) → 2 (the foundation is broken, so drop to "skip recommended").
• A single × on a non-core element (S3/S4, K3/K4) → 3 (only an element that raises quality is missing, so keep it at "caution." Do not drop below 3 even if there are also △).
• Two or more × → 1 (regardless of core or non-core).

When in doubt, err toward the lower score (because the strategy itself is designed conservatively: "do not enter if even one skip condition applies").

Overall evaluation

The overall score is capped at the lower of the two axes (gate method).

Overall	Verdict	Meaning
5	Entry Qualified (Exemplary)	Equivalent to the model examples. A shape that may be executed if STEP 0 has been passed
4	Entry Qualified (Standard)	A valid line. Can be executed with normal position sizing
3	Caution	The shape holds but has weaknesses. Additional confirmation recommended (wait for a 3rd-point reaction, wait for the reversal structure to complete)
2	Skip Recommended	Effectively meets the strategy's skip conditions. Do not enter
1	Skip (Clear Trap)	Build-up / squeeze / forced line. The losing pattern of February–May

Principle for invoking the special rule (200EMA warning cap) ★ Most important — preventing recurrence of misjudgment

The special rule is a hard cap that limits the overall score to a maximum of 2. A hard cap is handled fundamentally differently from the ○/△/× scoring of the axes:

• For axis scoring, "when in doubt, go lower" is fine (conservative design). But for the special rule, "when in doubt, do not invoke it." Because it forcibly drops the overall score to 2, invoke it only when the objective signal is explicitly satisfied (positively confirmed).
• Never invoke it on undetermined impressions such as "-ish," "leaning toward," "converging," or "forming." Those drop to "Caution (3) / awaiting confirmation," not to the special rule.
• The special rule is "confirmation that a bad pattern is happening," not "a hunch that it might happen." Write hunches in the Axis 2 / observation supplement, and do not let them cap the score.

Only when one of the following is confirmed in Observation ⑤ and meets the development requirements below, cap the overall score at 2 regardless of the axis scores (because it is directly linked to the strategy's "biggest warning point" = a bounce turning into a break).

1. Squeeze (confirmation requirements): all of the following are met.
   • The line and the 200EMA are converging (converging wedge), and
   • Price has chopped inside that wedge for 3 or more candles with each candle's range (body + wicks) shrinking (volatility contraction), and
   • Price is alternately touching both boundaries of the wedge (line side and EMA side) (touching only one side, or merely reaching the apex, does not count).
   • → A stage where these are not all met and it is merely geometrically converging / has merely reached the apex is "forming = not applicable (monitor)." The special rule is not invoked.
2. Clinging to the EMA (build-up): Only when positively confirmed as × = clinging in "Quality of the 200EMA reaction (R)" below (the contact zone has reached the valid number of candles, and c≥3 or a≤0.15). Do not invoke it when undetermined (just reached, 1–2 contact candles).

Before invoking the special rule, always ask yourself (if not all 3 are YES, it does not apply):

• Q1: Is that judgment a "measured objective fact," or a "hunch that it is likely to happen"? → If it is a hunch, it does not apply.
• Q2: Is there data from 3 or more candles in the contact zone (or inside the wedge)? → If only 1–2 candles, it is unmeasurable = undetermined = not applicable.
• Q3: Have I written "-ish," "leaning toward," "converging," or "forming" in my text? → If so, it is undetermined and not the special rule. Drop to Caution (3).

Objective judgment of the quality of the 200EMA reaction (R)

Classify how price reacted to the 200EMA: a "bounce that springs away," a "build-up that clings to the EMA," or undetermined because it has only just arrived and cannot yet be judged. A bounce is confirming evidence for the entry; clinging is a precursor of the break turning into a 200EMA break (= toward the SL), and the two are distinguished by the character of the price action (model examples: a bounce leaves the EMA with amplitude and forms a double structure / the losing example at the end of April 2025: oscillating back and forth across a flat EMA while contracting).

Contact zone: from the first candle in which price reached the 200EMA until the line-break candle.

Valid number of candles in the contact zone (★ gate): Both c and a measure "the behavior of multiple candles relative to the EMA," so if the contact zone is too short, the measurement does not hold. If the contact zone has fewer than 3 candles (arrival is almost on the last candle, 1–2 touches), c and a are unmeasurable, and R is automatically "undetermined." In this case, do not classify it as either clinging or bounce (= do not write "clinging-ish" or "bounce-ish"). This is a frequent state in this strategy, where the line break occurs at the same time as the 200EMA touch (because the entry trigger comes before the reaction resolves).

3 objective signals (measured only when the contact zone has 3 or more candles):

Signal	Toward bounce (good)	Toward clinging (dangerous)
Cross count c (number of sign changes where the close crossed the 200EMA)	c ≤ 1 (reacted on one side and left)	c ≥ 3 (oscillating across the EMA)
Departure size a (maximum distance from the EMA within 3–5 candles after the touch ÷ correction range R)	a ≥ 0.30 (sprang away with amplitude)	a ≤ 0.15 (stays right on the EMA)
Candle contraction / EMA slope	Clear rejection candle, EMA has a slope	Bodies/wicks contracting, EMA flat

Judgment (4 categories):

• Undetermined (1–2 contact candles, just arrived): c/a are unmeasurable. Special-rule trigger 2 is not invoked. State "reaction undetermined (just arrived)" in Observation ⑤, and reflect it in the main cause of the overall evaluation as "reaction undetermined = awaiting confirmation" (per the gate, usually Caution = 3. No hard cap).
• Bounce (○, confirming evidence): contact zone of 3 or more candles and c ≤ 1 and a ≥ 0.30. Reinforces S3 (reversal structure). No penalty.
• Clinging (×, special rule invoked): contact zone of 3 or more candles and (c ≥ 3 or a ≤ 0.15). Special-rule trigger 2 → overall capped at 2.
• Intermediate (△, caution): contact zone of 3 or more candles but not confirmed as either bounce or clinging above. No hard cap; state it in the supplement and combine it with other weaknesses.

Important (handling of early arrival): Pushing "undetermined" toward "leaning toward clinging" and invoking the special rule is a mistake (a misjudgment that actually happened with GBPAUD in a past session). Undetermined is an unfavorable factor but is awaiting confirmation, not a confirmed trap. Capping at 2 as a trap requires positive confirmation of clinging.

Note: Clinging often makes S1 (wave rhythm) × as well, but when the whole correction has swings and only the final approach clings to the EMA, S1 cannot catch it. This judgment is a dedicated filter that fills that gap. The boundary is decided by the EMA slope (flat = reinforces clinging).

Output format (strict; no deviation)

Whether the evaluation result goes to chat or to Notion, always use exactly the same structure, headings and table layout as the template below. Changing the format by destination, replacing tables with bullet points, and merging, omitting or reordering sections are all prohibited.

Required rules:

1. Observations are always a 5-row table (①–⑤). Replacing it with bullet points or paragraphs with bold labels is prohibited.
2. Axis 1 and Axis 2 are always 4-row tables each (S1–S4 / K1–K4). The heading must include the score "— X/5".
3. Place 2–3 lines of supplementary notes (prose) right after each axis table.
4. The "Reading" and "Basis" columns are a word to a short sentence (about 15–40 characters as a guide). Put long explanations in the supplement.
5. End with "### Overall Evaluation — X/5 (verdict label)" + the fixed 4-row basis table (Gate / Special rule / Main cause / STEP 0). Only when the overall is 3 or lower, place an improvements table (weakness → improvement trigger) right after it. Do not output the improvements section when it is 4 or higher.
   • Choose the verdict label from the fixed 5 in the conversion table (Qualified (Exemplary) / Qualified (Standard) / Caution / Skip Recommended / Skip (Clear Trap)).
   • Do not write prose paragraphs in the overall evaluation. Verbose content such as comparisons with other pairs, preambles or background explanations is prohibited (fit the necessary point into the "Main cause" column in one line).
   • Improvements are one row per element that was △/× in scoring (prioritize the elements that capped the score, max 3 rows). Do not use your own numbering like ①②③ or free-form bullet points.
   • Do not add your own sections under other names such as "Conclusion" or "Skip decision."
6. Write the direction (long/short, line type) in one line right after "## Line Evaluation Result".
7. The only line you may add when writing to Notion is an evaluation-date stamp at the end (e.g. *Evaluated: YYYY-MM-DD / 200EMA strategy line evaluation*).

## Line Evaluation Result
Direction: (Long/Short, descending/ascending line, cross type)

### Observations
| Element | Reading |
|---|---|
| ① Impulse wave | (direction, duration, price range, cross) |
| ② Corrective wave | (origin, duration, retracement depth, EMA reached) |
| ③ Pivots | (recognized swing highs/lows) |
| ④ Line | (origin, pivots passed through, number of reactions) |
| ⑤ EMA relationship | (slope, touch markers, contact candles, whether squeeze confirmation requirements are met, reaction quality: bounce/intermediate/clinging/undetermined) |

### Axis 1: Swing Clarity — X/5
| Element | Mark | Basis |
|---|---|---|
| S1 Wave rhythm | ○/△/× | (one line) |
| S2 Pivot reaction | ○/△/× | (one line) |
| S3 Reversal structure | ○/△/× | (one line) |
| S4 Uniqueness | ○/△/× | (one line) |

(2–3 lines of supplementary notes)

### Axis 2: Angle & Scale Fit — X/5
| Element | Mark | Basis |
|---|---|---|
| K1 Origin match | ○/△/× | (one line) |
| K2 Relative slope | ○/△/× | (one line) |
| K3 Sufficient scale | ○/△/× | (one line) |
| K4 Break location | ○/△/× | (one line) |

(2–3 lines of supplementary notes)

### Overall Evaluation — X/5 (verdict label)
| Basis | Content |
|---|---|
| Gate | min(Axis 1 X, Axis 2 Y) = Z |
| Special rule | 200EMA warning (squeeze/EMA clinging) not applicable (invoked only on positive confirmation; undetermined or forming is not applicable. Only if it applies, state "applies → cap 2" with the type) |
| Main cause | The element that capped the overall score, in one line (e.g. K4× = late break far from the 200EMA) |
| STEP 0 | Higher-timeframe slope is outside the image / reference only (panel: 15m… 1h… 4h… D…) |

**Improvements** (only when overall is 3 or lower)
| Weakness | Improvement trigger |
|---|---|
| (element that was △/× and one line) | (how to wait / where to draw so that it becomes ○, one line) |

How to write the overall evaluation and improvements (unified format):

• Output the overall evaluation as the fixed 4-row table above (Gate / Special rule / Main cause / STEP 0), in the same order every time. Do not write prose paragraphs.
  • Gate: only the formula min(Axis 1 X, Axis 2 Y) = Z.
  • Special rule: always state whether the squeeze applies or not (append "→ cap 2" only if it applies).
  • Main cause: the one element that determined the overall score, in 15–40 characters. Even if there are several, narrow it to the single most constraining one.
  • STEP 0: one line noting that the slope cannot be read from the image, and the reference values from the top-right panel (cross state).
• Make each improvements row an element that was △/× in scoring, with "weakness → improvement trigger" per row (elements that capped the score first, max 3 rows). Do not output it for overall 4 or higher.

Past format deviations (concrete examples never to repeat)

In an evaluation of silver, the following deviations occurred when writing to Notion. These are prohibited:

• Moved the overall score to the top, as in "Overall: 1/5 — Direction: …" (→ the overall is always the last section)
• Wrote "### Observations" as bullet points with bold labels instead of a table (→ always a 5-row table)
• Added sections not in the template, "### Conclusion" and "### Skip Decision" (→ fit them into the overall evaluation section and improvements)
• The axis heading score notation and the position of the supplement did not match the chat version (→ chat and Notion are completely identical)
• Wrote the overall evaluation as long prose including comparisons with other pairs and background (→ only the fixed 4-row table; the point goes in the "Main cause" column in one line)
• Wrote improvements as free-form ①②③ text (→ a "weakness → improvement trigger" table, one row per △/× element)

Writing the evaluation result to Notion

When the user specifies a destination Notion page (e.g. "1. USDCAD inside 6/7-6/12"), write the evaluation result to that page.

Steps:

1. Find the specified parent/child page with the Notion search tool (the typical hierarchy is Trading Journal → week range name → currency pair name)
2. Fetch the page and look for the "AI Evaluation" heading provided by the template. Write the evaluation result under that heading. Only if there is no "AI Evaluation" heading, create one at the end of the page and write there
3. The content must have exactly the same structure as the output-format template, word for word (direction line, observations table, Axis 1 table + supplement, Axis 2 table + supplement, overall evaluation, improvements). Do not restructure, summarize, or add sections for Notion. Only a single evaluation-date stamp line may be added at the end
4. Also show the identical evaluation result in the chat, with a short note that writing to Notion is complete

If the page cannot be found or there are multiple candidates, do not write; confirm the candidates with the user.

Retrieving chart images from a Notion page (flow for evaluating with only a page specified)

When the user specifies only a page name and has not attached an image in the chat, try to retrieve the image from the page and evaluate:

1. Fetch the page with the Notion fetch tool. Under the page's "## Execution TF" section is the image URL of the execution-timeframe chart (a signed URL on prod-files-secure.s3.us-west-2.amazonaws.com, valid for 1 hour)
2. In bash, download it with curl -s -o /home/claude/chart_exec.png "<image URL>" and confirm with the file command that it is a PNG/JPEG
3. Open the image with the view tool and perform the evaluation
4. Images in the "## Higher TF" section can be retrieved the same way, but the subject of evaluation is always the execution-timeframe line. Use higher-timeframe images only for contextual reference mentions
5. If the user's own notes on the page (description of their entry decision) can be read, still perform the evaluation itself independently from the image; after evaluating, you may comment on agreement/disagreement with the user's self-analysis

Fallback on failure:

• If the download fails with "Host not in allowlist" or similar, prod-files-secure.s3.us-west-2.amazonaws.com is not in the network allowlist. Guide the user to add it to the settings (takes effect from a new conversation), and in the meantime ask them to upload the image to the chat
• Signed URLs expire after 1 hour, so if the download fails, re-fetch the page to get a new URL and then retry

Notes for evaluation

• State explicitly that information that cannot be read from the image (the slope of the higher-timeframe 200EMA = STEP 0) is outside the scope of evaluation. This skill evaluates only the line quality of STEP 3.
• Do not score in hindsight. Even if the price action after the break appears in the image, the evaluation is based on "could the line be drawn, and was it valid, before the break." A winning trade with a badly drawn line can be rated low, and a losing trade with a correctly drawn line can be rated high.
• If no line is drawn on the image, first diagnose "can a valid line be drawn?" If it can, evaluate assuming you drew it yourself; if it cannot, report that fact itself as a low score on Axis 1 (only a forced line is possible = skip).

The core idea

An entry has two confirmations of different natures, and it holds only when both point in the same direction.

• The "directional bias" of the cross … the bias toward the direction we are about to target, indicated by the 20EMA/200EMA cross. It is only a signal of "which way to look."
• The "slope direction" of the 200EMA … whether the 200EMA itself is actually sloping in that direction on the higher timeframes (15m and 1h). This is the evidence that the trend is really moving.

In the previous review, losses were concentrated in February–May because I entered on the directional bias of the cross alone, mistaking "cross = signal" in range phases where the 200EMA was flat. Checking the slope is a filter to reject this range trap, and must always be checked separately from the directional bias.

---

STEP 0: Higher-timeframe filter (the first gate; skip if it fails here)

Before the entry timeframe, check the slope of the 200EMA on 15m and 1h.

• The 200EMA on both 15m and 1h is sloping in the intended entry direction.
• Flat is not allowed. The opposite direction is out of the question. Judge by "is it sloping," not "has it crossed."
• Also skip if the slopes on 15m and 1h disagree (one flat / opposite).

A cross that does not pass this gate is not touched, no matter how clean the shape.

STEP 1–3: Entry procedure (after passing STEP 0)

1. Confirm the 20EMA / 200EMA golden/dead cross (= fix the directional bias).
2. Confirm price touching the 200EMA (the pullback of the corrective wave).
3. Dow-theory trend reversal — enter after confirming the break of the descending/ascending line with a candle body. However, the line must not be "drawn blindly from the highest high/lowest low"; the condition is that it can be drawn along the swing highs/lows of the corrective wave.
   • The line connects highs/lows that actually reacted during the correction (2 or more points, ideally confirmed by a reaction at a 3rd point).
   • Those highs/lows form a named reversal structure (double top/bottom, head and shoulders).
   • The line is fixed using only pivots that can be drawn before the break. Do not redraw it afterward.

---

Definition of a valid line (the most important supplement to STEP 3)

Use how "easy the line is to draw" itself as diagnostic material.

A genuine corrective wave forms clear swing highs and lows as counter-trend traders enter and exit. A break of the line connecting those points = the moment the counter-trend traders give up = the resumption of the impulse wave. That is why the break means something.

On the other hand, a slow grind down / slow grind up or a range is not a correction from profit-taking but accumulation (build-up). Because control has not been settled and no swings form, if you force a line without an origin, you can draw any number of lines. That "break" is noise, not a reversal, and the final resolution comes on the 200EMA side (= the opposite direction) — this is what mass-produces stop-losses.

Judgment heuristics:

• If you feel like drawing through the middle of a cluster → skip as a squeeze.
• If three equally valid lines can be drawn → skip.
• A line price has not reacted to is not "a line drawn by the market" but "a line you fitted yourself" → invalid.

---

Exits (RR)

• SL … the last high/low of the corrective wave.
• TP … the N-calculation value of the previous impulse wave.
• BE (exit at break-even) … exit when the EMAs cross in the opposite direction.

Conditions for validity (all must be met)

• The 200EMA on 15m and 1h is sloping in the entry direction (STEP 0. Most important; flat not allowed).
• The descending/ascending line is drawn connecting swing highs/lows of the corrective wave that have reacted (forming a reversal structure).
• The RRR is within 1:1 to 1:3.
• The N-calculation value exceeds the most recent high/low.
• Preferably, the impulse wave used as the reference for the N calculation comes with an EMA cross.
• It is not a cross within a range (a cross of flat EMAs is invalid).

Skip conditions (do not enter if even one applies)

• The 200EMA on 15m and 1h is flat, or the slopes disagree / point the opposite way.
• A cross while the 200EMA is flat.
• The pullback after the 200EMA touch is a slow grind down/up or a range, and no clear swing highs/lows form (= a line can only be forced).
• Price has not reacted to the line, or it does not form a reversal structure (double / head and shoulders).
• The N-calculation value does not exceed the most recent high/low, or the reference impulse wave has no cross.
• The RRR is below 1:1 or above 1:3.

---

The biggest warning point: cases that become a "200EMA break" instead of a "bounce"

The aim is a bounce at the 200EMA, but when it turns into a break, it becomes a pile of stop-losses (February–May was exactly this).

• Price is compressed between the 200EMA and the ascending/descending line in a squeeze → the 200EMA is likely to break.
• Price is chopping near the 200EMA → there is a high chance it is a build-up for a break.

And in chop/squeeze phases, the slope of the 200EMA is almost always flat. In other words, the STEP 0 slope filter is the front-line defense that rejects this break trap at the entrance.

Understand the filters as separate layers: STEP 0 (slope) is a context-level filter: "is there a trend in the first place?" The definition of a valid line is a structure-level filter: "is this pullback a genuine correction, or a build-up?" Importantly, a squeeze can happen even when the 200EMA is sloping (a tight flag/triangle along the EMA right after a strong impulse wave passes STEP 0). The line filter is the second line of defense that catches, at the entrance, the losses that slip through STEP 0.

200EMA Strategy Key Points (the scope needed for line evaluation)

Structure of the strategy

• Execution timeframe: 5m (1m and 1h are also allowed). Indicators: 20EMA and 200EMA.
• STEP 0 (higher-timeframe filter): the 200EMA on both 15m and 1h is sloping in the entry direction. Flat is not allowed. Outside the scope of this skill's evaluation (it cannot be judged from the image).
• STEP 1: Fix the directional bias with the 20EMA/200EMA golden/dead cross.
• STEP 2: Confirm price touching the 200EMA (the pullback of the corrective wave).
• STEP 3: Enter on a candle-body break of the descending/ascending line connecting the swing highs/lows of the corrective wave. This is what this skill evaluates.

Definition of a valid line (core of the original text)

• A genuine corrective wave forms clear swing highs and lows as counter-trend traders enter and exit. A break of that line = the counter-trend traders getting off = the impulse wave resuming.
• A slow grind down / slow grind up or a range is accumulation (build-up). No swings form, and if you force it, you can draw any number of lines. Its break is noise, and the final resolution comes on the 200EMA side (the SL side).
• The line is fixed using only pivots that can be drawn before the break. Do not redraw it afterward.
• The pivots form a named reversal structure (double top/bottom, head and shoulders).
• Judgment heuristics:
  • If you feel like drawing through the middle of a cluster → skip as a squeeze
  • If three equally valid lines can be drawn → skip
  • A line price has not reacted to = a line you fitted yourself → invalid

The biggest warning point

Cases where aiming for a "bounce" at the 200EMA turns into a "break." Signs:

• A squeeze where price gets compressed between the 200EMA and the line
• Chop near the 200EMA (build-up)
• At such times, the slope of the 200EMA is almost flat

Layers of filtering: STEP 0 (slope) is at the context level; the line definition is at the structure level. A squeeze can happen even when the 200EMA is sloping (a tight flag/triangle along the EMA right after a strong impulse wave). Line evaluation is the second line of defense that catches losses that slip through STEP 0.

Exits (reference)

• SL: the last high/low of the corrective wave
• TP: the N-calculation value of the previous impulse wave
• BE: exit on an EMA counter-cross
• RRR 1:1–1:3, and the N-calculation value must exceed the most recent high/low

Analysis Notes on the Model Images (XAUUSD 5m)

Characteristics of the ideal form extracted from 9 model entry images provided by the user. Green straight line = ascending/descending line; boundary of the red/green box = entry price; red side = SL; green side = TP.

Characteristics common to the ideal form

Related to Axis 1 (swing clarity: S1–S4)

1. The corrective wave has 2–3 clear waves (S1). Not a section where candles cluster and move sideways; a rhythm of up → pullback → up (or the reverse) is visible.
2. The line connects 2–3 pivots that have reacted (S2). In many model examples, a touch & bounce around the 3rd point on the line can be seen, and the break right after it is the entry.
3. A reversal structure is complete at the point where the 200EMA is reached (S3). Double tops/bottoms are the most frequent (e.g. there is an example of a downward entry explicitly described as "because a double top appeared"). The typical shape is the second peak/trough being held down by the 200EMA.
4. The line is highly unique (S4). In the model examples, swings stand out, so the shape is "this is the only place it can be drawn," leaving almost no room to redraw.

Related to Axis 2 (angle and scale: K1–K4)

5. The origin of the line is always at the origin of the corrective wave (the end pivot of the impulse wave) (K1). Every line in the model examples covers the whole correction; there are no lines drawn from a sub-wave in the middle or at the tail of the correction. Low-quality lines that look steep were, without exception, ones that captured only the accelerating final leg at the tail of the correction.
6. The line's slope is clearly gentler than the slope of the preceding impulse wave (K2). Impulses are fast; corrections are slow. If the corrective line is as steep as the impulse, it is not a correction but an impulse in the opposite direction (a trend reversal), or a line on a sub-wave whose scale is too small.
7. The correction develops over time (K3). It is normal for the corrective wave to take as long as or longer than the impulse wave to form, and length is not a deduction. Real examples: a pullback correction of about 6 hours against a falling impulse of about 5–6 hours (big winning short), and a pullback correction of nearly half a day against a sharp falling impulse of about 2 hours (double top → short). The more time a correction takes, the more swings stand out and the more reliable the line becomes. Conversely, a tiny line containing only a few candles is invalid for lack of scale.
8. The break point is near, and right after, the 200EMA touch (K4). Because the line break and the bounce from the 200EMA overlap in the same zone, the SL (the final high/low of the corrective wave) is close and the RRR stays within 1:1–1:3. Quantification (added 2026/6): the break location is measured by "the retracement ratio normalized by the correction range width R: r=(X−B)/R (long)." Measured on the model examples: model long r≈0.29 / model short r≈0.33, both breaks in the lower 1/3 of the correction range. This is the anchor for K4=○. r>0.50 (a break in the upper half of the correction) is "buying/selling the pullback" away from the EMA, and K4=×. For the details of the formula, see "Objective judgment of K4 (break location)" in SKILL.md.

Record of retired criteria

• The old criterion "a correction lasting 1/3 to 1× the impulse wave is appropriate" was retired as incorrect (2026/6). Measurements of the model examples showed many cases of correction ≥ impulse, and a correction taking time is, if anything, a hallmark of a genuine correction. Do not judge by time ratio.
• The expression "an absolute angle of 30–45° is appropriate" was also retired. Angle depends on the image's aspect ratio, so always judge by the slope relative to the impulse wave (K2).

Key points of individual images

• Downtrend, big win: The 200EMA clearly slopes down. A pullback correction taking about 6 hours against a falling impulse of about 5–6 hours. An ascending line connecting 2–3 swing lows, far gentler than the impulse. The origin is the start of the correction. Broke at the same time as reaching the 200EMA and collapsed sharply. The reference example with all elements ○.
• Uptrend resumption, long: The pullback to the upward-sloping 200EMA forms a descending wave pattern. Broke above the descending line connecting 2–3 pullback highs right after the 200EMA touch. A reference example with good origin match and relative slope.
• Double top → short: After a steep falling impulse (about 2 hours), a pullback lasting nearly half a day formed a double top just below the 200EMA. The representative example of an ideal form even though the correction took several times as long as the impulse. Entry on the break of the ascending line connecting the swing lows.
• Two consecutive entries: In the same chart, both a descending-line break → long and an ascending-line break → short were valid. In both, the line is drawn from the origin of the correction and connects the swings.
• Steep, choppy-ish examples: Includes steep lines capturing only the sub-wave at the tail of the correction, and examples with many overlapping candles near the 200EMA. In the images too, these tend to move against the position for a while after the break, or fail to extend, which is consistent with the claim that K1×/K2× reduce the edge. Use them as references for judgments of 2 points or lower.
• Flat 200EMA, range (end of April 2025): A phase where the 200EMA is almost flat and price moves up and down around the EMA. A line can be drawn but the swings are shallow — a typical build-up phase. As the representative of the February–May losing pattern, use it as a calibration example for Axis 1 = 1–2. The reference example for "quality of the 200EMA reaction" = clinging (×): closes oscillate back and forth across the flat EMA many times (cross count c≥3), each candle contracts, and price does not leave the EMA with amplitude (departure size a is small). Special-rule trigger 2 (clinging to the EMA) is invoked and the overall is capped at 2.

Calibration anchors for the quality of the 200EMA reaction (R)

• Reference examples for bounce (○): Uptrend resumption long = a pullback to an upward (not downward) sloping 200EMA → after the touch, leaves upward with amplitude accompanied by a rejection candle (c≤1, a≥0.30), then the impulse resumes in a double-bottom-like way. Double top → short = touched twice just below the 200EMA and was pushed back (held on one side, no cross), then decisively left downward. Both are "touch, spring, and leave."
• Reference example for clinging (×): The flat example at the end of April 2025 (above). Licking the EMA, oscillating back and forth, contracting.
• The essence of the distinction: A bounce "respects the EMA from one side and springs back (amplitude, rejection candle, double structure)"; clinging "oscillates back and forth across the EMA with volatility contraction (the EMA pierces the candle bodies)." EURNZD (2026/6/17) was a bounce with amplitude off a rising EMA = R○, and the special rule did not apply. A real example where the weakness was separated out as K4 (late break), not the reaction quality.

Calibration anchors for scoring

• A line that "is drawn from the origin of the correction, connects multiple reacting pivots, comes with a reversal structure, has a slope clearly gentler than the impulse, and breaks near the 200EMA" has all elements ○ = 5/5.
• Each element that drops to △ lowers the score by about 1 point (follow the conversion table).
• A single × on a core element makes the axis 2: grind/range with no swings (S1×), a line on only the tail sub-wave of the correction (K1×), a slope as steep as the impulse (K2×). A single × on a non-core element (S3/S4, K3/K4) keeps the axis at 3 (because it is the absence of an element that raises quality, not the collapse of the foundation). Two or more × makes the axis 1.
• Real evaluation example (USDCAD 5m long, 2026/6/8): Evaluated as S1○ S2△ (3rd point not confirmed) S3△ (the second bottom was lower and briefly broke below the EMA) → Axis 1 = 4 under the old method. Under the new conversion table, a borderline case that could be ○2△2 = 3. The user's sense was that 4 was appropriate, so △ is limited to "clear defects," and minor imperfections are treated as leaning ○.
• Real evaluation example (EURNZD 5m long, 2026/6/17 / K4 calibration example): Initially K4=○ with overall 3, but measuring the break location gave A=1.9930 / B=1.9905 / X=1.99194 / R=0.0025, so r≈0.58 (upper half of the correction), g≈0.54. The break came after retracing about 14 pips from the 200EMA touch (B=1.9905), which is not "near the touch" → under the new K4 system, K4=×. K4 is non-core, so under the new conversion table Axis 2 stays at 3, and through the gate the overall remains 3 (Caution) (= the verdict label is the same as initially, but the breakdown was corrected from K4○→× and Axis 2: 5→3, making the location of the error explicit). The RRR was also compressed toward 1:1 at about 1:1.2, supporting the late entry. Saved as a typical example of mistaking "a break within the same corrective structure" for "near."

Misjudgment from over-invoking the special rule (2026/6/18 GBPAUD 5m short / anchor to prevent recurrence)

A short setup approaching the 200EMA from below and breaking an ascending line was initially evaluated as overall 2 (Skip Recommended), but corrected to overall 3 (Caution) after the user pointed it out. The cause was that the special rule (hard cap) was invoked as confirmed at an undetermined stage. The three errors and the correct handling:

1. Wrote "clinging-ish" and invoked special-rule trigger 2 → wrong. The 200EMA was reached by effectively one candle, and the contact zone was fewer than 3 candles. Neither c nor a was measurable; correctly, R = undetermined. The special rule is not invoked when undetermined. Recognizing clinging from "-ish" was an overreach.
2. Wrote "squeeze applies" and invoked special-rule trigger 1 → wrong. The line and the EMA were merely converging geometrically (had only reached the apex), with no compressed chop of 3 or more candles inside the wedge and no alternating touches of both boundaries. Correctly, squeeze = forming = not applicable (monitor).
3. Asserted a "deep pullback" without stating the reference → wrong. Against the broader decline (1.8880→1.8800 = 80 pips), a 46-pip pullback = about 58%, within the normal range (depth alone is not a deduction). The actual weakness was not depth but "breaking above the origin of the most recent micro down-leg = a micro higher high (the trend-negation issue of K3)," and the two had been confused. Mistaking the pullback reaching the EMA itself for "deep/dangerous" was also a category error.

Correct breakdown: Axis 1 = 3 (S3× = top not formed, S1/S2 △), Axis 2 = 3 (K1○ K2○ K3△ K4△; K4 was a late break with r≈0.44/g≈0.42), special rule not applicable, overall 3 through the gate. The reason for holding back was not "clinging / deep pullback" but awaiting confirmation: "the 200EMA reaction is undetermined (bounce not confirmed) + the break is late."

Lesson (made into a principle): The hard cap (special rule) requires positive confirmation. For axis scoring, "when in doubt, go lower" is fine, but for the special rule, "when in doubt, do not invoke it." In this strategy, where the 200EMA touch and the line break come at the same time, "reaction undetermined" is a frequent, normal state; it must not be pushed toward clinging and capped at 2. Undetermined drops to Caution (3, awaiting confirmation).