Skip to main content

From a setup to a strategy

Who it’s for
Users who found something on a discovery screen and want to act on it
Assumes
You have read Discovery & research

Every discovery screen ends at the same place: a Build strategy button that hands off to the AI engine. It does not hand off to an order, and this page explains why that is the design rather than an omission.


What the handoff does

It gives the builder a starting context, the instrument, the timeframe the detector evaluated, and the shape of the condition that was found. You then build a strategy the way you build any strategy: state your rules, set your risk, review the assembled object, backtest it.

What the handoff deliberately does not do

Not carried acrossWhy
A position sizeSizing is a decision about your capital and risk budget. A scanner has no basis for it
A stop-lossThe invalidation level says where the pattern voids, which is a different question from where you should exit
A take-profitNothing on the screen implies a target
An orderThere is no path from a research screen to a live order, by design
A recommendationThe screen reported a condition; it did not suggest trading it
Invalidation is not your stop-loss

This is the single most important distinction on this page. The invalidation level is the price at which the technical pattern the detector identified no longer exists. It is a statement about the pattern.

Your stop-loss is a statement about your risk budget: how much of your capital you are willing to lose on this trade. Those two numbers coincide only by accident.

Using the invalidation level as a stop means your position size is being set by a scanner's geometry rather than by your risk tolerance. Size against your risk budget, then check whether the resulting stop makes sense relative to the pattern. Not the other way round. See Position sizing.


The honest workflow

StepDo thisAsk yourself
1A setup catches your eye
2Read the evidence, not the scoreDo the "why now" statements describe something you believe in? On the Trading Map, open the verification pack
3Check the contextEvent risk? Freshness? Coverage in this slice? Is this a broad move or one instrument?
4Form a rule, yours rather than the screen'sFor example: buy RELIANCE when close breaks the 20-session high on 1.5 times average volume
5Build it in the AI engineHave I stated my own stop, target and per-trade risk?
6Backtest itDoes the rule have any history behind it, or only today's instance?
7Paper it
8Decide

Steps 2, 4 and 6 are the ones that turn a scanner hit into something you can defend.

Step 2 deserves the emphasis

The score is a summary. The evidence is the content. "Close 2,847.30 is above 20-session high 2,841.00 (+0.22%)" is a claim you can evaluate, a 0.22% break is a marginal one, and knowing that is worth more than knowing the score was 82.

Step 6 deserves it too

A discovery screen tells you a condition is present now. A backtest tells you what that condition has been worth historically, with costs applied. Those are different pieces of information and the second one is the one that bears on whether the rule is tradeable.

A setup with a STRONG grade whose rule backtests to grade D is telling you something clear: the pattern is present and has not been profitable. See Backtesting.


A worked example

From the Radar: RELIANCE, breakout, daily, Setup Quality 82, trigger 2,841.00, invalidation 2,769.50, event risk quarterly results in 3 sessions.

What a careless workflow does: clicks Build strategy, accepts defaults, deploys with the invalidation as the stop.

What a defensible workflow does:

  1. Reads the evidence. The break is +0.22% over the 20-session high (marginal. Volume at 1.50× average is real confirmation.
  2. Notes the earnings print in three sessions. Decides either to size smaller or to wait.
  3. Forms the rule generally, not for this instance: "long RELIANCE when close breaks the 20-session high with volume above 1.5× the 20-day average; exit when close falls back below the 20-session high; 1.5% stop; risk 1% per trade."
  4. Backtests it over the available window. Reads the grade, the drawdown, the trade count and the cost drag.
  5. If it holds up, papers it for a fortnight.
  6. Then decides) with the earnings date in mind.

The difference between those two is not effort. It is whether the decision was yours or the scanner's.


Things worth knowing

A rule is not one instance. Building a strategy from a setup means encoding the condition, so it can be tested across history and will fire again. If your strategy only ever fires on today's instance, you have not built a strategy.

Detector conditions are expressible in the builder. The three detectors are defined technical conditions, an N-session high with volume confirmation, multi-session return against a moving average, range against ATR. All of that is ordinary builder vocabulary. See Indicators.

The timeframe matters. A daily-timeframe detector hit does not imply a 5-minute strategy. Build on the timeframe the condition was measured on unless you have a reason not to.

Snapshot age. If the candidate is DELAYED or STALE, the condition may already have changed.


Next