Build a Crypto Trading Strategy From a Testable Rule
Specify venue and product identity, causal messages, fill assumptions, funding and operating ownership before evaluating a crypto rule.
Name the product and the observable hypothesis
A crypto strategy needs more identity than an asset name. Specify venue, spot or derivative product, base and quote units, contract convention and account scope. A stablecoin-quoted spot pair and an inverse perpetual can use different cash and quantity rules. A common symbol does not prove identical history, fees or permissions across venues. Save the source's product identifier and effective specification.
Write a testable entry objective, eligible population and future label before collecting attractive charts. A trend, reference recovery and order-book hypothesis are different experiments. Record delisted or renamed products and unavailable observations rather than study only today's active survivors. CFTC's advisory is a consumer-risk reference, not a conclusion that a named strategy is manipulative. This guide provides no current token recommendation, profitable rule or assumed all-market access.
Build an availability-aware data contract
Identify exchange event time, vendor processing, local receipt and later corrections. A candle's closing timestamp does not automatically establish when its final value arrived. Quote snapshots, trade prints and incremental book messages have different coverage. Define sequence-gap recovery and when a derived feature becomes eligible again; quietly carrying stale depth can manufacture a predictive series.
Save product changes, outages, maintenance and missing intervals. A continuous-looking chart should not create observations during a documented data gap. API documentation explains a particular interface, not the completeness or licence of the historical dataset in a study. If the feed lacks the precision needed for the proposed fill or event-order claim, narrow the scope. A bar-level reference calculation can remain useful without pretending to reconstruct queue priority or hidden liquidity.
Specify orders and terminal states before simulating results
Write activation, requested quantity, minimum increment, order type, validity and amendment ownership. Coinbase's order reference is product-specific and should not be generalised to every exchange or derivative. A triggered condition is not already an account position, and a limit touch is not proof that the account traded the full requested amount. Keep accepted requests and executed quantities separately.
In a fictional one-unit request, 0.40 units fill before cancellation of the remaining 0.60. Exposure is 0.40 until a closing execution changes it. A cancellation message cannot erase that exposure, and an API timeout cannot safely be interpreted as definitely no order. Retain client and exchange identifiers so retries and later reports can be reconciled without duplicating the instruction. These fixtures test state reasoning, not actual exchange reliability.
Compare gross movement and product-specific costs
Use an invented fixed notional of 10,000, one-percent favourable move, fourteen basis points of combined execution friction and six funding events costing two basis points each. Gross scenario outcome is 100, execution cost fourteen, funding cost twelve and net seventy-four. With the same costs and an adverse one-percent move, net is -126. This is the worksheet's fixed-notional arithmetic, not a market backtest.
If the signed funding assumption changes to a two-basis-point credit per event, funding cost becomes -12 and the favourable net becomes 98. Do not present that scenario as a forecast of real funding. Actual rates, event times, quantities and eligibility require the exact derivative contract. Spot, linear and inverse products need their own accounting. Borrowing, collateral changes and liquidation are outside this simple fixed-notional model.
Freeze selection, native exits and operating constraints
Keep entry ranking on its declared causal objective and freeze identity before applying one predeclared exit for economics and exposure. Preserve native whole-position, minute-exit and maximum-hold rules; partial exchange fills are execution states, not a new partial-exit strategy. Later policy changes or discretionary rescues need effective versions and an appropriate evaluation boundary.
Check account permissions, funding, shared exposure and missing-state response independently. A conviction grade cannot choose personal leverage or calibrate a winning probability. Production operating claims require managed-cloud ownership, durable state and failure evidence so normal work continues with personal computers off. A local research notebook can test accounting but cannot become a hidden production publisher or verifier. This guide introduces no new service, connected account or automated trading job.
Deliver a complete experiment rather than a flattering curve
The package should contain data identity, availability map, product units, candidate history, frozen entry, exit contract and every eligible instruction state. Save requested, filled and remaining quantity, costs and unresolved exposure. Check label spans and fit scope at reserved boundaries; a chronological starting date is not enough to exclude future information.
The blank construction record organises these decisions before a real experiment. Reproduce the state and fixed-notional examples outside empirical results. Software correctness does not independently validate an edge, and a research live label does not establish a subscriber route. A valid outcome can be no acceptable candidate or insufficient execution evidence. This page describes how to make a crypto research claim inspectable without inventing provider returns, strategy significance or an assurance of future profitability.
Review checkpoints
| Object | Preserve | Do not infer |
|---|---|---|
| Identity | Venue, product and unit contract | One asset ticker implies identical instruments |
| State | Accepted, filled and remaining quantity | A timeout proves no exposure exists |
| Economics | Signed funding and fixed-notional scope | A worksheet scenario is a validated return |
Download blank research record (CSV)
Headers only. No trades, account outcomes or operating receipts are supplied.
Primary Sources
- Coinbase Advanced: product order documentation
- Coinbase Advanced Trade API: documented interface scope
- CFTC: virtual-currency pump-and-dump advisory
Crypto Friction Budget Worksheet
Continue the Research
A complete amount packet is only one component
The filled workbook can complete an equation without completing instrument research. Real entry ranking retains the raw causal objective, frozen identity, one predeclared exit and independently verified master. The teaching engine has no historical price route or parameter selector. Its manually entered label does not implement a missing asset-class adapter, and its decimal checks are not independent financial samples. Preserve the original rule and causal-data checklist when using the amount examples. Actual product identity, full eligible history, receiving quantities and cost vintages remain separate evidence requirements.