A crypto screener automates finding markets that match conditions. A trading bot can automate submitting and managing orders under assigned rules. The important difference is where authority to act begins. A filtered result, an alert and an exchange fill are different records.
These functions can live in separate tools or inside one platform. A bot may consume a screener’s output, use strategy signals or run its own checks. A product name alone does not tell you which steps are automatic. Map the workflow before choosing how much responsibility to delegate.
Identify the four stages
Our crypto screener introduction explains filters and results. Here, the question is who takes the next action. Use this four-stage worksheet to compare a proposed setup:
| Stage and output | Question to ask |
|---|---|
| Discovery: a matching market or shortlist. | Which conditions and data produced the match, and how recent are they? |
| Notification: a message about an event. | Who receives it, and what happens if it arrives late or twice? |
| Decision: permission to act or a recorded rejection. | Does a person approve each trade, or do preapproved rules decide? Who sets size, expiry and exposure limits? |
| Execution: an order and its actual status. | Who submits, checks, cancels and manages exits? Where can you verify filled quantity? |
The decision stage might be entirely manual, partly automated or delegated to a strategy within limits. Execution can be automated even when a person supplies the decision. Conversely, automatically finding a setup does not mean an account is authorized to trade it.
A notification needs an explicit next step
An alert can prompt human review or feed another system. TradingView’s webhook documentation describes sending an alert message to an external application and notes that delivery can fail. The message is an input to that application; it is not evidence of an exchange fill.
For a manual workflow, define who reviews the notification and when it becomes stale. For an automated workflow, define the accepted event, account, applicable rules and rejection reasons. A repeated delivery should not silently become another intended entry. If a response is missing, check the actual order state before deciding whether another submission is appropriate.
Keep the event time, receipt time, decision time and exchange response separately. Our chart-confirmation lesson explains another timing boundary: when a condition first becomes knowable.
Follow one hypothetical handoff
Consider a fictional AAA/USD spot market on Venue V. A saved screening rule matches when a completed five-minute candle closes above 100 USD per token. At 10:00:00 UTC, the completed 09:55–10:00 UTC candle arrives with a close of 100.20. That creates a candidate, not an order.
Assume a separate teaching plan has already specified a possible purchase of 2 tokens, a buy limit of 100.30 USD and an entry expiry at 10:01:00 UTC. Assume its separate balance, size, exposure and exit-plan checks pass. These arbitrary settings are not a recommended strategy or a description of any product.
Compare two possible handling paths for that same candidate:
- Human review: at 10:00:08 UTC, the reviewer sees a best ask of 100.80. Under this example’s rule, a quote above 100.30 means skip the entry. The alert remains recorded; no order is submitted.
- Delegated execution: at 10:00:04 UTC, a hypothetical controller sees a best ask of 100.25 and submits the authorized two-token limit order at 100.30. Suppose the exchange reports only 0.5 token filled at 100.25. The remaining 1.5 tokens are still open. At expiry, the controller requests cancellation and verifies a final cancellation response with no further fills.
The second path ends with 0.5 token held, not 2. Entry expiry did not sell that holding; the separate exit policy still applies. These invented timings do not establish that bots are faster, safer or more profitable. Different arrival times or market conditions could reverse the comparison.
Order submission is not completion
Kraken’s market-and-limit-order guide explains that a limit order restricts the execution price but may fill partially or not at all. The example’s limit therefore does not reserve two tokens at the observed quote. Fees are separate from the quoted execution price.
Compare requested, filled, remaining and canceled quantities. A cancellation request alone is not a cancellation confirmation; inspect the final exchange state, including fills that occurred while the request was in transit. A paused entry process also need not close a position or cancel existing orders. Ask precisely what each control does.
A trade signal’s reference entry belongs to the proposal. An account’s fills belong to execution. Evaluating a bot using only the proposal can miss skipped entries, rejected orders, partial fills, fees and exit handling.
Choose the responsibility you want to delegate
If the bottleneck is finding markets worth reviewing, screening and a daily watchlist may address it. If the plan is already precise and the task is executing it repeatedly, investigate an execution service’s actual controls and records. Automating an undefined plan merely leaves important decisions unspecified.
Before delegating, write down which account can be traded, what permissions are required, who sets risk limits, how stale or repeated events are handled, and who responds when the service is unavailable. Specify what happens to resting orders and open positions when new entries stop. Use the risk-management guide for the separate sizing discussion.
Test the handoffs with small, explicit scenarios: no match, rejected candidate, late message, repeated message, rejected order, partial fill and interrupted connection. A simulation can test behavior without establishing future live fills or returns. Keep the expected response beside the observed result.
Judge discovery and execution separately
For discovery, inspect data freshness, useful candidates and exclusion reasons. For execution, inspect account-level orders, fills, cancellations, costs and unresolved positions. For strategy quality, compare appropriate historical evidence and live observations using the backtest guide. None of these records substitutes for the others.
The useful question is not which label sounds more advanced. It is which step you want automated, under whose rules, with what evidence that the intended action occurred.
Educational information, not financial advice. All markets, prices, times and fills in the example are hypothetical. Automation does not establish a profitable strategy or remove trading risk.
