Max Agent
Articles

Build a crypto scenario with four evidence cards

A practical research workflow for connecting price, liquidity, positioning and on-chain observations, with explicit timing rules, competing explanations and a notification testing plan.

MaxData EditorialPublished Updated

Write the research question before opening the dashboard

Our proposed workflow begins with a short research brief: what must happen, which observations will qualify, what will contradict the hypothesis, and when the review ends. Treat this as an editorial convention for organizing an investigation. Do not assign a trading instruction to an isolated reading.

For a first exercise, ask: “After price crosses my reference level, what combination of observations will I classify as sustained acceptance above that level?” Then write a competing question: “What would make me classify the move as a return into the previous range?” Leave room for an unresolved result.

Use four evidence cards: price, liquidity, positioning and on-chain activity. Give each card a distinct question rather than asking all four to produce a bullish or bearish vote. In our proposed worksheet, conflicting observations require an explanation or an unresolved classification. Do not remove a series simply because its reading disagrees with the opening hypothesis.

This article is an educational methodology exercise. References describe what the supplied documentation says; they do not certify which features are available in a particular account. The account invitation at the end is commercial.

Build a data contract for each card

The Pro Alerts documentation names BTC price, on-chain metrics and exchange metrics among its categories. Its examples include Long Term Holders SOPR, Bitcoin Reserve on Exchanges and Funding Rates. Pro Alerts documentation. Use those names as starting points for documentation searches, without assigning them an interpretation from their labels alone.

For the price card, our checklist asks for the exact instrument, venue, market type, quote currency and chosen observation, such as a completed candle close. For the liquidity card, specify whether your question concerns hypothetical order execution or a particular reserve series. Require a separate source justification before treating either as an answer to the other question.

For the positioning card, choose one precisely identified series and obtain its provider definition. For the on-chain card, name the network, measurement and coverage you intend to examine. These are proposed research roles, not assertions about how the four categories cause price movements.

Every card should contain a definition, unit, provider, aggregation method, sampling interval and timestamp convention. Add the documented publication delay and any documented revision policy. If either is unavailable, write “not established” and keep it visible. Do not fill gaps with an assumed universal delay.

Give each card an analysis horizon and an explicit limitation. Add an alternative interpretation as a question to investigate, not an asserted explanation. For example: “Would I classify this observation differently using another documented coverage universe?” Require a source before writing the answer as a factual conclusion.

Our proposed readiness rule is simple: a card without a definition or identifiable source remains outside automated evaluation. You can still list it in the research brief as unfinished work. Do not quietly replace the missing definition with the meaning you hoped the metric would have.

Our proposed checklist

Make an observation eligible for review

  1. 1

    Assign a question

    State what the card is intended to investigate.

  2. 2

    Document the series

    Record the provider definition, unit, coverage and aggregation.

  3. 3

    Check timing

    Record sampling, availability and documented delay.

  4. 4

    Set boundaries

    Write limitations, an alternative question and an invalidation condition.

An editorial preparation workflow, with unknown fields left explicit. A four-step checklist covering the research role, provider definition, timing and interpretation boundaries.

Set the clock and define missing information

Adopt one review clock for the exercise. We propose UTC, completed hourly price candles and a six-hour review horizon. These are illustrative design choices, not recommended trading parameters. Select the remaining series intervals only after checking their documentation.

At each review, record both the observation time and the time the value became available to your process. Our proposed convention admits only values recorded as available by that review deadline. Preserve later values for the next review rather than inserting them into the earlier snapshot.

Set a separate maximum age for each series, with a written rationale linked to its documentation and the research question. If a publication delay remains unknown, label timing compatibility unresolved. Do not treat an unresolved timestamp as evidence that a condition passed.

Use three worksheet states: pass, fail and unassessable. Store the raw reading beside the state. Also record the rule version and any manual intervention. If you change a threshold, start a new version and retain the previous record. Our suggested audit convention is to keep observations made under old rules visibly separate from observations made under revised rules.

Work through a fictional range break

Consider a fictional instrument called NORTH with a reference level of 250 units. Assume hourly price observations and a review that expires after six hours. All values and conditions in this example are invented for teaching. They describe no observed market and imply no expected return.

Before observation begins, specify the thresholds for the liquidity, positioning and on-chain cards. Do not choose them retrospectively to match the price path. Our proposed exercise uses three competing classifications.

Scenario one: acceptance. Require two completed hourly closes above 250, valid readings from the other three cards, and satisfaction of their previously written conditions. Record the exact readings rather than reducing the result to “four confirmations.” The scenario name describes our classification rule.

Scenario two: return. Require a completed close below 250 after a completed close above it. Preserve the other three card readings for comparison with the acceptance hypothesis. Do not append an unsupported causal story about who bought, sold or moved assets.

Scenario three: unresolved. Use this when required data are missing, timing compatibility is unknown, or neither price classification applies. End the exercise at the six-hour deadline even if it remains unresolved. In our workflow, the deadline is part of the original specification.

Add two diagnostic comparisons to the review: classify the same exercise using price alone, and repeat it using an alternative documented source where a suitable one has been identified. Treat both as proposed tests. Do not assume that the four-card version will outperform either comparison.

Invented NORTH exercise

Classify the observation three ways

Acceptance
Price requirement
Two completed hourly closes above 250
Other evidence
All three cards valid and their written conditions met
Journal entry
Record the complete condition set
Return
Price requirement
A close below 250 after a close above
Other evidence
Preserve readings for comparison with acceptance
Journal entry
Record the return and accompanying readings
Unresolved
Price requirement
Neither classification established
Other evidence
Missing readings or unknown timing compatibility
Journal entry
Record the reason assessment stopped
Proposed exercise rules using an imaginary reference level of 250; these are not forecasts. Three alternative classifications: acceptance above the reference, return below it, and unresolved evidence.

Specify notifications before connecting execution

The documentation distinguishes a single-condition Alert from a Pro Alert combining several conditions. It also describes a time window for meeting those conditions and a sleep period after activation. Alert and Pro Alert definitions.

Our proposed configuration sheet should name every input, comparison operator, threshold, value source and evaluation time. State whether a completed interval is required, how missing readings are handled, and when another notification may occur. Write the intended behavior in plain language before mapping it to interface settings.

Check which parts the interface can express and retain unsupported checks for manual review. For this educational workflow, begin with notifications and a journal. Treat order execution as a separate implementation decision requiring its own specification.

Require each test notification to contain the scenario name, rule version, data timestamps and conditions that passed. Label it as a research observation. During review, compare the recorded payload with the original specification; do not accept an unexplained label such as “strong opportunity” as a substitute for the underlying conditions.

Review historical observations without projecting a return

The Backtests documentation describes checking historical alert trigger counts and displaying average and median price changes before and after an alert. It also describes testing either a single condition or several conditions together. Backtests documentation.

Our proposed report includes the date range, event count, missing observations, overlapping review windows and outcome at the predefined horizon. Compare price alone with the combined specification using the same reporting rules. Keep historical observations in a section separate from hypotheses for the next evaluation.

If you add a hypothetical transaction simulation, explicitly document fees, slippage assumptions, order size and the liquidity criterion used to admit an observation. Mark unknown execution assumptions as unresolved. Do not present an incomplete simulation as a realized result.

We also propose splitting threshold development from evaluation on a separate period. Keep a ledger of every configuration tried and explain why you selected the final version. Use that ledger as your editorial check for fitting choices to historical results. Do not turn a selected historical average into an expected future return.

Finish with a decision about the research process: retain the specification, revise a named assumption, or suspend evaluation until documentation is complete. Record the reason and the next review date. No transaction is required to complete this exercise.

Our proposed evaluation sequence

Separate design, evaluation and observation

  1. Freeze the specification

    Save thresholds, timing rules, missing-data handling and review horizon.

  2. Inspect historical events

    Count observations and compare price alone with the combined conditions.

  3. Evaluate separately

    Apply the selected rules to a separate period without retuning them.

  4. Observe notifications

    Keep timestamped payloads and record reasons for revisions.

A staged research convention without a claim of predictive performance. Four research stages: freeze the specification, inspect history, evaluate a separate period and journal notifications.

Documentation questions and commercial invitation

The Help Hub presents MaxData Info Bot as an assistant for questions about MaxData. MaxData Help Hub. Our suggested question is: “Where is the provider definition, sampling interval and publication delay for this exact series?” Check the relevant provider documentation before completing the data card.

Commercial invitation: create a MaxData account if you want to explore the product environment. Verify the data and settings available in your account before attempting to reproduce the exercise.

Sources

  1. MaxData Help Hub – Pomoc i przewodniki Mój workspace — agent.maxdata.app; accessed October 30, 2026
  2. Pro Alerts — agent.maxdata.app; accessed October 30, 2026
  3. Backtests — agent.maxdata.app; accessed October 30, 2026

Start working with your own agent

The agent knows this article context and still follows the assigned department checklist.

Create a MaxData account
Max Agent