Skip to content
Browse documentation

Breakout Radar

Potential breakouts validated against the chart, with posts translated to English and a mechanical risk plan.

GET/api/radar
Radar & ScannersIncluded in every plan ยท one key.

Overview

**Historical endpoint: `/api/influencers`.** The board now lives at `/api/radar`, the path that names what it returns; the old one answers identically while clients migrate and carries `Deprecation`, `Sunset` (2027-03-19) and a `Link` to its successor. `/api/influencers` keeps `view=accounts` and `view=traders` for good โ€” those really are about influencers. Hybrid pre-breakout radar whose social evidence comes from the influencers you choose. `leaderboard` and `candidates` hold the same variable core set (0โ€“30), never padded. Every row must clear all filters: favourable regime, verified market cap of at least $300M, earnings clear, 0โ€“8% from the pivot and a mature pattern measured on closed daily candles. Posts keep the original wording in `text` and add English in `textEn`; `translation` reports whether that coverage is complete or partial. Cached for about 3 hours, rebuilt on demand.

Use it toSee only pre-breakout setups that clear every filter, and choose which influencers provide the social evidence.

**Stocks only.** The radar returns the stocks and their lines: no stop, target, trailing stop or position size โ€” how to trade each one is up to you (since 2026-11-11; `risk_budget_usd` and `max_position_usd` are still accepted and ignored). **How many setups are there?** `candidateCount` setups are near breakout and `watchlistCount` more are on the watchlist (both set by the radar's model); together they are today's setups, listed in `candidates` and `watchlist`, best first. **`likely`:** the setups ranked by the same model (`likelyProb`, 0โ€“1: the model's probability), the top 3 (`likelyCount`), each with `ticker`, `price`, `pivot`, `distToPivot`, `likelyProb`, `sector`, `industry`, `onRadar`, `earningsSoon`/`earningsDate`, `setupSummary` and its `base` block, re-checked on live prices when read. Kept for integrations; since 2026-11-04 the site shows one radar. **Live, `GET /api/radar/live`:** with your key, today's list against quotes at most a minute old: each setup's `state` is `broke` (above its line now), `touched` (reached it today and back under) or `below`, `crossedAt` is the New York time it first reached the line, plus `vsLinePct`, `changePct` and `volumeVsAvg`. A setup that reached its line is ready to buy by the radar's rule (a buy stop at the line); the free plan gets its top setups, paid plans all of them. **Confirmed breakouts, `confirmedBreakouts` (from November 2026):** after each close, every stock that closed the night before within 5% under its 50-session high and in this session broke that high and closed above it is scored by a model; the best ones (the top 1% of a year of scores) are signals to buy at the next open. `confirmedBreakouts` carries `session`, `generatedAt`, `today` (the session's signals with `close`, `line`, `prob`, `industry`), `open` (bought and still inside their 4 sessions), `finished` and a `record` (`signals`, `finished`, `wins`, `avgRet`); each bought signal carries `best`, the highest price since the entry open against it; a signal counts as a win when it reaches 4% over the entry open within 4 sessions before a close 3% under the line. No stop or target is given. Today's and the open signals are for paid plans; other keys get `locked: true` with `todayCount` and `openCount`. Null until the after-close job has run. **Good breakouts, `goodBreakouts` (from 2026-11-02):** after each close every breakout of the session โ€” a base's first close above its pivot, in the whole universe โ€” is scored by a model retrained on past years (gradient-boosted trees on what the daily chart, the stock's industry, its SEC-reported sales and earnings, the market and recent breakouts show at that close). A breakout scoring at or above the model's `threshold` (the best 5% of its training) while SPY is above its 50- and 200-day averages is a good one: bought at the next open (the same as at that close in the test: the next open gapped +0.0% on average) and sold 20 sessions later. `goodBreakouts` carries `session`, `generatedAt`, `today` (that session's good breakouts with `entry`, `pivot`, `score`, `industry`), `open` (earlier ones still inside their 20 sessions, with `last`, `lastDate`, `ret` so far and `spyRet` over the same days) and `closed` with a `record` (`trades`, `avgRet`, `avgVsSpy`, `winners`). Tested year by year from 2019 to October 2026, each year with a model trained only on the years before: +3.85% per trade on average, but most of it from a few very large winners (median trade +0.3%; without the 10 best of 537 trades the average is +1.4%), about half of the trades losing. The live record (`record`, `closed`) is what counts. Today's and the open signals are for paid plans; other keys get `locked: true` with `todayCount` and `openCount`. It is null until the after-close job has run. `stockCount` is not a setup count: it is every stock the build looked at, rejected ones included. Codes and field names stay stable so integrations do not break; `labels` holds their human-readable English labels. That now includes the section codes: `labels.section` reads `setup-explicado`, `mencionado` and `descartado`, and `labels.sections` reads the keys of the `sections`/`returnedSections` counters, which are plurals of those codes rather than the codes themselves. The three sections are a two-tier board: `setup-explicado` is a social mention that passes the full structural contract and carries an explained setup; `mencionado` keeps a social mention visible when it only fails quality gates (the chart is not verified, the mention is); `descartado` is reserved for hard reasons โ€” already broke its pivot, outside the forming window, unknown or imminent earnings, or market cap below the floor. A client can therefore tell a verified setup from a raw mention without losing either. The counters keep their keys on purpose โ€” publishing English duplicates beside them would double any client that sums the object, so they are translated rather than aliased. `priceAsOf` and `technicalAgeSeconds` say how fresh the price and the chart lines are; the chart cache is rebuilt after about two hours. If the board is stale and its rebuild fails, `sourceStatus` gives the cause, `sourceNeedsAttention` is true when retrying cannot help (for example an inactive X subscription), `ageDays` gives the age in whole days and `sourceDetail` explains it; the rebuild stops at the first batch on an inactive subscription. With personal accounts, `personalization.sourceStatus` appears when their refresh failed for an account-wide X cause; X is then not queried for 10 minutes and the previous overlay is served if it exists. **What the pattern quality grade means.** `patternQualityScore` starts at 42 and adds six measured components, each capped so no single one can carry a chart: proximity to the pivot (up to 16, `16 โˆ’ distToPivot ร— 1.5`), base maturity (up to 12, `(baseAgeSessions โˆ’ 10) ร— 0.6`), volume dryness (up to 12, `12 ร— (1.15 โˆ’ volContraction)`), base compactness (up to 12, `12 โˆ’ range20Pct ร— 0.45`), how many times the pivot was tested (up to 12, `pivotTouches ร— 3`) and a structure quorum (2 points for each of six signals: range contracting, rising lows, dry volume, a strong pivot cluster, price in the upper half of the base and a compact recent range). The total is clamped to 0โ€“100. `grade` reads that score: **A at 80 or above** (textbook), **B from 65 to 79** (solid), **C below 65**. The score only ever RANKS a valid setup: `patternQualityPass` is a separate hard contract, and a chart that trips any of its checks โ€” already broke, immature base, illiquid, extended, choppy, untested resistance, a recent spike or gap โ€” is rejected whatever it scored, with the failed checks listed in `patternQualityReasons`. A high score cannot rescue a chart the contract rejects, and every rejection says which check failed. `screenFunnel` reports how the technical lane reached its candidate count โ€” charts scanned, with data, with a setup, selected โ€” and names the gates that rejected the rest, so a zero can be read as a weak tape or as a broken gate. **Chart contract: the trendline base.** A chart qualifies only when it has a base drawn with lines โ€” a resistance touched at least three times that is flat or leans down gently, and a flat or rising support under the price. The resistance line is the `pivot`; `broke` means the price is already above it. **The setup rule (since November 2026).** A base in play โ€” within 8% under its line, at least $5 and $10M a day, no close above the line in the 20 sessions before โ€” is a setup (`baseTier` `ready`) when the price is above a rising 200-day average (`base.aboveMa200`, `base.ma200Rising`), the stock rose at least 15% before the base (`base.metrics.priorTrendPct`), its 14-day ATR is at least 4% of the price (`base.metrics.atrPct`) and it trades within 15% of its 52-week high (`base.metrics.fromHighPct`). Every setup is listed, best first by `base.breakoutProb` (0โ€“1, the model's probability that it breaks out), and some nights there is none. There is no `watch` tier any more; `eligibility.excluded.baseWatchOnly` stays at 0. The radar finds breakouts; it is not a trading plan. The six conditions that set the tiers until 2026-11-01, taken from the user's own chart ratings, are still measured and published as information: `flatTop` (resistance between โˆ’0.3 and +0.1 ATR per 10 sessions), `tight` (last 10 sessions within 3.0 ATR, or 3.5 when the other five conditions all hold), `aboveEma21` (at least 55% of the base closes on or above the EMA21), `accumulation` (up-day volume at least 0.82ร— down-day volume inside the base), `priorTrend` (at least +15% over the ~6 months before the base) and `volumeDryUp` (average volume inside the base at most 1.12ร— the 50 sessions before it). `base.metrics.supportSpan` (share of the base the support holds up) is published too. Each row's `base` block publishes the tier, `stage` (`continuation` or `stage1`), `sessions`, the `resistance` and `support` values at the last closed candle, `resistanceTouches`, `supportTouches`, `resistanceSlope`/`supportSlope` (ATR per 10 sessions), `convergence` (width now รท width at the start), the six `conditions`, their `metrics` (`atrPct` โ€” the 14-day ATR as a % of the last close โ€” `fromHighPct` โ€” the last close against the 52-week high, in % โ€” `toLineAtr`, `last10RangeAtr`, `aboveEma21Pct`, `upDownVolume`, `priorTrendPct`, `baseVolumeRatio`, `recentUpVolumeDays` โ€” up days in the last 10 sessions on volume above the previous 50-session average โ€” `recentMaxUpVolume`, the strongest of them as a multiple of that average, published as information and never as a gate, `dollarVolume50`) and the `failed` ones. `patternQualityPass` is true for a `ready` or `watch` base, `patternQualityScore` is the base score and `patternQualityReasons` names the failed six conditions (information) in the same terms the funnel uses (`base_top_not_flat`, `base_not_tight`, `base_below_ema21`, `base_no_accumulation`, `base_no_prior_advance`, `base_volume_expanding`, `base_not_near_line_or_converging`, `no_trendline_base`); `screenFunnel.method` reads `trendline-base-conditions`. Context published with each base, as information and never as a gate: `relativeStrength` compares the stock with SPY on the same closed sessions โ€” `rsVsSpy3m` and `rsVsSpy6m` are the % change of the RS line (price รท SPY) over ~3 and ~6 months, `rsLineNearHigh` is true when that line is within 2% of its high over the `sessions` measured (up to 126), the classic sign of a leader waiting under its resistance; `sector` and `industry` are Nasdaq's classification, and `industryBases` counts the industry's `members` in the universe and how many have a `ready` or forming (`watch`) base in the same build, published only for industries with at least 8 members. `base.aboveMa200` and `base.ma200Rising` state the 200-day average facts behind `stage`, and `base.metrics.priorTrendBars` says over how many sessions the prior advance was measured (120, or fewer when the history is shorter). `baseCount` says which base this is since the advance began, counted the O'Neil way on ~2 years of daily bars: `baseNumber` 1 is the first base (the user's "Stage 1 base"), 3 or more is late; `priorBases` lists the earlier bases cleared (start, end, `depthPct`, `sessions` โ€” a base is a pause of at least 5 weeks and 8% that was later cleared on a close), `advanceStart` where the advance began, and `truncated` is true when the advance may have begun before the bars available. The count starts again when the price undercuts the low of the last counted base or falls more than 50% from a high. `screenFunnel.nearMisses` lists up to 25 charts that failed exactly one check โ€” the ticker, that check, the pattern score, the distance from the 52-week high, `adrPct` and whether the base shows strict contraction (last 10 sessions' range under 0.8ร— the 10 before, volume not expanding) โ€” and `nearMissReasons` counts them by check, because the rejection counts are not exclusive and cannot show where good charts are lost.

Your influencer list is private and editable

Add, rename, replace or hide X accounts from your dashboard. A change affects only your own responses, never the shared list or another customer. The same list is used in two places: the Breakout Radar reads it automatically, and Stock Sentiment reads it when you ask with include=mine, which adds a separate myInfluencers block and leaves the shared numbers untouched. Posts are delivered in English through textEn and always keep the original text in text.

Coverage

Universe
US equities from chart scans and tracked traders
History
Current board
Updates
On demand ยท ~3h board / ~2h technical cache
Source
X timelines + Yahoo technicals + AI classification

Call it

Send your ra_live_โ€ฆ key in the x-api-key header. Pick your language:

curl "https://raspberrytrades.com/api/radar?setup=breakout" \
  -H "x-api-key: ra_live_your_key_here"

Parameters

setup
string
Filter by pattern. Friendly aliases: breakout (includes base_breakout), base, flag; exact values also work: vcp, staircase, pullback, bull_flag, base_breakout, cup_and_handle, ascending_triangle, high_tight_flag, double_bottom, bull_retest, support_bounce (optional).
ticker
string
Check a single symbol, e.g. NVDA (optional)
view
string
tickers: only the stocks โ€” `candidates` and `watchlist` as two lists of tickers, nothing else; evaluated: audit-only list with per-ticker exclusion reasons; accounts: trusted accounts; traders: trader statistics. ticker/setup and sizing parameters apply only to board and evaluated views. Invalid combinations return 400 before source requests (optional).
x-api-key
header
Your API key (required)
Example response

Example, not a live result. Field shapes are exact; the values, prices and timestamps are illustrative. Call the endpoint for current data.

200 OK ยท application/json
{
  "generatedAt": "2026-09-18T14:00:00Z",
  "locale": "en",
  "translation": { "targetLanguage": "en", "originalField": "text", "translatedField": "textEn", "posts": 18, "completed": 18, "translated": 18, "status": "complete", "preservesOriginal": true },
  "regime": "ON",
  "cached": true,
  "stale": false,
  "signalFresh": true,
  "refreshDeferred": false,
  "count": 13,
  "strictCount": 13,
  "candidateCount": 13,
  "watchlistCount": 9,
  "returnedSections": { "explicados": 13, "mencionados": 0, "descartados": 0 },
  "returnedSources": { "trader": 2, "technical": 9, "hybrid": 2 },
  "candidateFilters": {
    "explainedSetupRequired": true,
    "conservativeTechnicalValidationRequired": true,
    "patternQualityRequired": true,
    "marketCapRequired": 300000000,
    "readyMaxDistToPivotPct": 8,
    "selection": {
      "available": 13, "returned": 13,
      "limit": 30, "truncated": false
    }
  },
  "filters": {
    "explainedSetupRequired": true,
    "broke": false,
    "earningsSoon": false,
    "maxDistToPivotPct": 8
  },
  "leaderboard": [
    {
      "ticker": "MU", "breakoutScore": 82,
      "source": "hybrid", "section": "setup-explicado",
      "candidateState": "ready", "strictEligible": true, "actionable": true,
      "pattern": "vcp", "patternLabel": "VCP base โ€” volatility contracting above the EMA21", "grade": "A",
      "setupTraders": 2, "traderCount": 3, "postAgeHours": 11.2,
      "price": 198.2, "distFromHigh": -4.1,
      "pivot": 205, "distToPivot": 3.4,
      "broke": false, "earningsSoon": false,
      "patternQualityPass": true, "patternQualityScore": 91,
      "patternQualityReasons": [],
      "patternMetrics": {
        "structuralAsOf": "2026-09-17",
        "baseAgeSessions": 24,
        "pivotTouches": 3,
        "range20Pct": 14.2,
        "volContraction": 0.72
      },
      "setupSummary": "@srxtrades (bull flag): $MU coils under 205; watching for the breakout",
      "setupReasons": [
        {
          "evidenceSource": "trader",
          "trader": "srxtrades", "setupType": "bull_flag", "setupTypeLabel": "bull flag",
          "text": "$MU se comprime bajo 205; vigilando la ruptura",
          "textEn": "$MU coils under 205; watching for the breakout",
          "images": ["https://pbs.twimg.com/media/...daily.jpg"],
          "createdAt": "Wed Jul 01 06:09:25 +0000 2026",
          "url": "https://x.com/srxtrades/status/..."
        }
      ],
      "why": "the same array as setupReasons โ€” kept because clients read this name"
    }
  ],
  "candidates": [
    {
      "ticker": "MU", "source": "hybrid",
      "section": "setup-explicado", "candidateState": "ready",
      "strictEligible": true, "actionable": true, "distToPivot": 3.4,
      "technicalQualified": true, "patternQualityPass": true,
      "marketCapQualified": true,
      "technicalSetupReason": {
        "evidenceSource": "technical",
        "author": "Raspberry Radar",
        "setupType": "vcp",
        "text": "MU presenta un vcp grado A: cotiza a 3.4% del pivote..."
      }
    }
  ],
  "watchlist": []
}
Radar & Scanners โ€” related datasets
Get your API key

Create a free account, copy your ra_live_โ€ฆ key and make your first call to https://raspberrytrades.com/api/radar.

Keep your key on your server โ€” never ship it in front-end code.