Level provenance · reaction study
6C Support and Resistance: Define and Test Levels
Give two traders the same 6C chart and one may draw yesterday's high, a weekly swing and an overnight range while the other marks round numbers and volume nodes. Both can later point to a reaction. Until the source and reaction rule are fixed in advance, neither drawing is evidence.
Objective inputs
Choose a Small, Declared Level Vocabulary
Support and resistance are roles assigned to price areas by a method. A prior high does not become “resistance” by law, and an old low does not promise support. Use neutral labels until a test records what happened.
| Candidate source | Point-in-time rule | Main contamination risk |
|---|---|---|
| Prior session high / low | Freeze from a declared session clock after that session closes | Changing session boundaries or daylight-saving conversion |
| Opening range | Use a fixed clock window and do not redraw after it completes | Selecting the window after seeing the day's move |
| Swing pivot | Require a fixed number of completed bars on both sides | Using future bars before the pivot could have been known |
| Volume-derived node | Name the exact contract, data feed, sample and binning rule | Combining incompatible expiries or using incomplete trade data |
| Settlement / event reference | Use the authoritative timestamp and exact dated-contract price | Confusing spot USD/CAD, a fixing or another 6C month |
| Round-price grid | Declare the fixed spacing before the sample | Adding only round prices that visibly reacted |
Do not begin with all six. Choose the smallest set tied to the decision being tested. Every added source increases the number of zones on a chart and the chance that price appears to react to something by coincidence.
Tolerance
A Level Needs a Reproducible Width
Markets trade on a discrete grid, and fills occur across spreads and queues. Treating a historical price as an infinitely thin line creates false precision. Define a zone around the anchor with a rule that uses only information available then.
Fixed tick band
Example: anchor plus or minus two standard 6C outright ticks. This is simple and stable, but its economic width does not adapt as volatility changes. Current standard mechanics are verified on the 6C contract specification page.
Volatility-scaled band
Example: a predeclared fraction of a trailing true-range estimate calculated only from completed bars. This adapts, but adds estimator and lookback choices that must be frozen and validated.
A miss is part of the data. If the method repeatedly fails by similar distances, propose a revised width, freeze it and test it on new data. Editing historical zones to capture each reaction destroys the denominator.
When nearby anchors overlap, apply a declared merge rule. For example, merge zones whose edges overlap and use their outer boundaries, while retaining all source labels. Do not count the merged area as several independent confirmations.
Eligibility gate
Score Conditions Before Price Enters the Band
The same zone can have a different test history depending on age, prior contacts, approach and market state. Record those fields before the touch so they can later be analyzed rather than used as post-hoc explanations.
- Verify the contract.The displayed month and execution month match, or the basis between them is explicitly handled. Continuous back-adjustment is documented.
- Count prior contacts.Define when two prints form one contact and when separation creates a new one. Do not call a level fresh after repeated chopping through it.
- Measure approach.Record distance, elapsed time, realized range and whether price approached directly or after compression. These are descriptive inputs, not assumed predictors.
- Mark event state.Record scheduled Bank of Canada, Statistics Canada and material U.S. releases, plus holiday and roll conditions.
- Measure execution state.Capture spread, usable depth and recent trade intensity in the exact contract. A visible line cannot overcome an untradeable book.
- Apply the exclusion rule.Overnight gaps, missing data, crossed markets or a predeclared blackout make the observation ineligible rather than a discretionary skip.
Reaction taxonomy
Use Labels That Can Be Scored Without a Story
A touch can produce noise, rejection, temporary hold, acceptance through the zone or no resolution within the horizon. Directional trading rules come later.
No touch
Price never enters the tolerance band before the level expires. Keep it in the opportunity record so untouched levels do not disappear.
Touch, unresolved
Price enters but neither the declared rejection nor acceptance threshold occurs before the evaluation horizon ends.
Rejection
After entry into the band, price moves a predeclared distance away on the approach side before closing beyond the far edge.
Acceptance
Price closes beyond the far edge and satisfies the declared time or displacement condition. This is a label, not proof of continued trend.
Whipsaw
Both rejection and acceptance thresholds occur within the horizon. Score the declared first-event rule or preserve it as a separate state.
Invalid observation
Bad data, wrong contract, undeclared roll adjustment or a prohibited event window prevents a defensible label.
Worked annotation
One Hypothetical Level, Scored Without Hindsight
Suppose the prior declared session high in a standard 6C contract is 0.72600. The protocol creates a two-tick band from 0.72590 to 0.72610 because the current standard outright tick is 0.00005. The evaluation horizon is 20 minutes after first touch.
| Timestamped field | Recorded value | Interpretation limit |
|---|---|---|
| Source | Prior declared-session high at 0.72600 | Known before current session contact |
| Band | 0.72590–0.72610 | Illustrative width, not an optimized rule |
| First contact | Price trades 0.72590 | Starts the evaluation clock; not an entry signal |
| Maximum advance | 0.72615 | One tick beyond far edge |
| Subsequent retreat | 0.72555 within 12 minutes | Nine ticks below anchor; seven below the lower band edge |
| Label | Depends on frozen thresholds | Cannot be called rejection until the declared rule is applied |
If the rule defined acceptance as one close beyond 0.72610, the bar-close record decides. If it required three closes, a single print at 0.72615 is insufficient. If rejection required a six-tick move below the anchor before acceptance, 0.72555 may qualify. The point is not which label sounds best; it is that the same rule must score every case.
Research design
A Level Is Not “Reliable” Until It Beats a Fair Baseline
No original performance study is reported on this page. A proper test should ask whether declared levels change reaction distributions compared with matched prices that were not selected as levels.
Freeze the algorithm
Source, session clock, pivot delay, width, merge, expiry and label rules are versioned before evaluation.
Build matched controls
Compare with random or grid prices matched on contract, session, volatility, event state and distance from current price.
Prevent overlap bias
Define event independence when zones cluster or repeated contacts occur. Use grouped uncertainty by day and level.
Separate description from trade
First test reaction labels. Then, separately, simulate entry, fills, invalidation, costs and position sizing.
Report counts, conditional probabilities with uncertainty, excursion distributions, time to outcome and stability across roll, volatility and event states. Testing many sources and widths creates multiple-comparison risk; preserve all attempted variants and use untouched later data.
Common invalidators
Reject the Level When Its Provenance Breaks
Research failures
- Drawing after seeing the reaction
- Using future bars to define a swing
- Changing zone width by outcome
- Deleting untouched or failed levels
- Counting overlapping anchors as independent
- Testing many definitions and reporting only one
Execution failures
- Trading a continuous-chart price in a different month
- Ignoring the daily break or holiday clock
- Entering inside a scheduled blackout
- Using a spread wider than the planned reaction
- Placing invalidation inside the same tolerance band
- Treating a stop as a guaranteed fill
Level-quality scorecard
A Gate, Not a Prediction
Score one point only when the evidence is recorded before contact. A high score means the observation is cleaner for the declared method; it does not mean price must reverse.
| Gate | Pass condition | Fail action |
|---|---|---|
| Provenance | Source and creation time are auditable | Discard |
| Point-in-time | No future data helped define it | Discard |
| Contract | Exact expiry and adjustments are known | Wait or repair data |
| Zone | Width and merge rule were frozen | Discard |
| Pre-touch state | Contacts, event and market quality are recorded | Do not score |
| Risk room | Invalidation and opposing structure leave room after costs | Reject trade use |
An observation is eligible for the research sample only when all five research-validity gates—provenance, point-in-time construction, exact contract, frozen zone definition and recorded pre-touch state—pass under the predeclared policy. Any listed discard or do-not-score action overrides a numerical total. Risk room is a separate trading-use gate; even a research-valid observation can still be rejected as a trade. For a breach, continue with the breakout decision tree.
Sources and methods
- CME Group FX Product Guide 2026 for the standard 6C unit, quotation and outright tick used in the illustrative band arithmetic.
- CME Group holiday and trading hours for current session and holiday checks.
- CME Group FX calendar-spread FAQ for dated-contract and roll context.
- CFTC Futures Market Basics for general futures-risk context.
Sources and methods were reviewed August 13, 2026. This article provides a measurement design and hypothetical annotation, not an original finding that any named level source predicts 6C reactions or produces profitable trades.