Concepts

Reading a setup

A real setup walked through field by field: lines, trigger, plan, track record.

A setup is one current pattern on one ticker: two lines, the rule that turns it into a trade, the plan if that rule fires, and the record of how the same pattern family resolved before. This page walks through a real one field by field. The full field table is on the Setup page.

A forming falling wedge on NVDA
{
  "setup_id": "NVDA:D:falling_wedge:1",
  "lineage_id": "lineage-1",
  "symbol": "NVDA",
  "interval": "D",
  "family": "falling_wedge",
  "family_label": "Falling wedge",
  "pattern_type": "falling_wedge",
  "direction": "bullish",
  "status": "forming",
  "pattern_start": "2026-08-08",
  "pattern_end": "2026-09-07",
  "as_of_session": "2026-09-25",
  "last_close": 101.5,
  "price_scale": "log",
  "lines": {
    "upper": { "start": { "date": "2026-08-08", "price": 110.0 }, "end": { "date": "2026-09-07", "price": 104.0 }, "touch_count": 3, "touches": [{ "date": "2026-08-08", "price": 111.2 }] },
    "lower": { "start": { "date": "2026-08-08", "price": 90.0 }, "end": { "date": "2026-09-07", "price": 98.0 }, "touch_count": 2, "touches": [] }
  },
  "trigger_rule": "A daily close above the upper line; the trigger price moves with the line.",
  "plans": [{ "direction": "bullish", "trigger_price": 105.25, "stop_price": 96.0, "target_price": 120.0 }],
  "track_record": { "family": { "…": "see Track records" }, "ticker": null }
}

Identity

setup_id names this occurrence and is what chart.png takes. It is stable while the pattern stands; if the nightly build redraws the pattern, the id changes and lineage_id is what still ties the two together. symbol and interval (always D, daily) say what it is drawn on.

Family and direction

family is one of five: rising_wedge, falling_wedge, bullish_pennant, bearish_pennant, triangle. family_label is the human name. pattern_type is the detector's more specific label when it has one. direction is bullish or bearish, and it is null for a symmetrical triangle until price breaks one way, which is why a triangle can carry two plans.

Status

forming means the pattern stands and the trigger rule has not fired. Once it fires, status changes and an entered object appears with the entry date and price and the stop and target that were locked at entry. Occurrences that have entered also appear in the history route, first with outcome open, then win or loss.

The lines

Each of lines.upper and lines.lower is two dated endpoints in dollars. That is the whole definition: there is no slope field on purpose, because the lines are straight in log price and a slope in dollars per day would be wrong at the far end. To draw one, put the chart's price axis on a log scale and connect start to end; extend it to the right if you want to see where the trigger would be tomorrow. touch_count and touches are the bars whose high or low touched the line, useful for showing why the line is where it is. A line can be null for pattern shapes that only have one.

The trigger and the plan

trigger_rule is a sentence you can show: the condition that turns the pattern into a trade. plans is what the trade would be if it fires, one per direction: trigger_price (as of the last session; it moves with the line), stop_price and target_price (the measured-move target). last_close beside trigger_price tells you how close the setup is to firing. A symmetrical triangle carries two plans; everything else carries one.

The Edge Filter tier

Most nights the scan finds hundreds of patterns. The question a trader asks first is simple: which ones are the best tonight? The Edge Filter answers it with a letter grade, the same one shown on each setup card on the site.

Think of it like a school report. S is the top grade, then A, B and C. F means the Edge Filter would skip the trade, and so would Autopilot. Inside the passing grades, the setups are also put in order, best first. So "S3" means an S-grade setup that came third out of everything the Edge Filter kept tonight. An F has no place in that order, because it was not kept.

Each setup carries its grade in edge_filter. tier is the letter (S, A, B, C or F) and rank is the place in tonight's order (1 is the best; empty on an F). On an F, refused is true, and reason says why when there is one. For example, every short trade is F, because the Edge Filter only grades buys.

A pattern can have two plans, one for each direction. The grade on the setup is the one its card shows, which is its best plan's. Each plan also carries its own grade, so you can see which side earned it.

To ask for tonight's best, give the scan a grade and the Edge Filter's order: GET /scan?tier=S&sort=edge lists tonight's S grades, S1 first (in MCP, get_scan with tier: ["S"] and sort: "edge"). edge_filter is empty when a setup was not graded that night. The keyless demo shows the field with locked: true and no letter, the same as the free site.

The estimates

Before a setup triggers, the obvious question is how this particular trade is likely to go. The TradingPal model answers with three estimates for each plan, the same three the setup card on the site shows: how often a trade like this one wins, what it makes on average once losers are counted, and how far a typical winner runs.

They sit on each plan as estimates: win_chance_pct (Win chance), avg_return_pct (Est. avg return, the net percent per trade, winners and losers together) and est_win_pct (Est. win, the middle winning move from the trigger; target_price is that move as a price). A summary sentence carries all three.

These are the model's estimates for this order, not a record of past trades, so present them as estimates and keep their names. estimates is null when the night's grading gave the order none. The keyless demo shows Est. win only, as the free site does.

Similar trades

The estimates are a model's answer. Similar trades is the plainer one: find the earlier trades whose pattern looked most like this one, and see what happened to them.

Each plan carries their record as similar_trades: up to 50 earlier trades on any stock, on the same side, each one entered and closed before this setup existed. It gives count, wins, losses, win_rate_pct, avg_return_pct and avg_r, with a summary to quote whole. avg_r is the average result measured against the risk: a trade that made twice what it stood to lose at its stop is +2R, one that hit its stop is -1R. It is null when the evening search found no answer for the order.

To show the trades themselves, closest match first, with each one's entry, stop, target, exit and result, call GET /tickers/{symbol}/similar or the get_similar_trades tool.

The track record

track_record.family is the stat block for the whole family; track_record.ticker is the same measurement for this ticker alone when there are enough occurrences, else null. Both carry a summary sentence written to be quoted whole.

Time

as_of_session is the last trading session the nightly build covered, and pattern_start to pattern_end is the span of bars the pattern is drawn across. Present as_of_session beside anything you show; setups change nightly and should not be cached across sessions.

Stock facts

The answer for a ticker also carries stock_facts, once beside setups: the ticker's facts as the TradingPal screener's quick view shows them, in the same groups (Price, Performance, Momentum, Volume and volatility, Company), with the screener's own labels and ready-to-show values such as "EMA8 vs EMA21": "+2.5%" or "RSI (14)": "59". Show them as given. The chart card draws them as a table under the chart.

What to do with one

For a person, the useful sentence is: "NVDA has a bullish falling wedge forming as of 2026-09-25. It triggers on a daily close above 105.25 (last close 101.50), with a stop at 96.00 and a target at 120.00. Falling wedges have resolved 1,287 times since 2006: 54% reached the target, average win +8.1%, average loss -4.2%. This is educational chart analysis, not advice." Every number in it is a field above, and the last sentence is the disclaimer from the envelope.