---
title: "Api V2 Forced-Flow Backtest Get"
method: GET
path: "/api/v2/forced-flow/backtest"
tags: ["Forced Flow (v2)"]
---

# Api V2 Forced-Flow Backtest Get

`GET /api/v2/forced-flow/backtest`

## Request

- Base URL: `https://api.zerogex.io` (the document declares no server; this is the origin it was published from)
- URL: `https://api.zerogex.io/api/v2/forced-flow/backtest`
- Auth: HTTP bearer

Charm-into-Close track record: when the morning charm flow is stronger
(or weaker) than its own recent normal, does the noon->close return lean the
same way?

Scores the charm read RELATIVE to its trailing baseline rather than by raw
sign -- for an index book the raw sign is structurally almost constant ("buy"
nearly every day), so it degenerates to the naive baseline and tests nothing;
the deviation from recent normal is the part that varies. Reads the persisted
``forced_flow_profile`` + ``underlying_quotes`` history and reports the hit
rate against a naive directional baseline -- honestly, including when the
sample is thin or the edge is nil. Scores TWO forecast definitions over
identical sessions (``full`` = 0DTE-inclusive close flow; ``smooth`` =
first-order charm only) so the A/B is visible. Returns 200 with zeroed counts
(not 404) before enough sessions have accrued, so the page can show an honest
"collecting" state rather than an error.

## Query parameters

- `symbol` string
- `lookback_days` integer: Calendar days of history to score

## Response `200`

Successful Response

- EnvelopeBacktestResponse
  - `data` BacktestResponse, required
    - `symbol` string, required
    - `lookback_days` integer, required
    - `full` BacktestVariant, required
      - `total_sessions` integer, required
      - `evaluated_sessions` integer, required
      - `warmup_sessions` integer
      - `hits` integer, required
      - `hit_rate` number, nullable
      - `hit_rate_ci_low` number, nullable
      - `hit_rate_ci_high` number, nullable
      - `baseline_rate` number, nullable
      - `edge` number, nullable
      - `edge_p_value` number, nullable
      - `significant` boolean
      - `signal_mean_return` number, nullable
      - `signal_t_stat` number, nullable
      - `records` BacktestRecord[], required
        - `date` string, required
        - `charm_flow` number, required
        - `charm_baseline` number, required
        - `charm_deviation` number, required
        - `return_pct` number, required
        - `predicted_dir` integer, required
        - `realized_dir` integer, required
        - `hit` boolean, required
    - `smooth` BacktestVariant, required
      - `total_sessions` integer, required
      - `evaluated_sessions` integer, required
      - `warmup_sessions` integer
      - `hits` integer, required
      - `hit_rate` number, nullable
      - `hit_rate_ci_low` number, nullable
      - `hit_rate_ci_high` number, nullable
      - `baseline_rate` number, nullable
      - `edge` number, nullable
      - `edge_p_value` number, nullable
      - `significant` boolean
      - `signal_mean_return` number, nullable
      - `signal_t_stat` number, nullable
      - `records` BacktestRecord[], required
        - `date` string, required
        - `charm_flow` number, required
        - `charm_baseline` number, required
        - `charm_deviation` number, required
        - `return_pct` number, required
        - `predicted_dir` integer, required
        - `realized_dir` integer, required
        - `hit` boolean, required
  - `freshness` Freshness, required: The v2 freshness envelope. Present on every ``/api/v2/*`` response.
    - `evaluated_at` string, date-time, required: When the API evaluated this response (server wall clock, UTC). Endpoint-health signal: advances on every request even when the underlying data is unchanged.
    - `generated_at` string, date-time, required: When the served data was computed. Equals the snapshot stamp for cycle-backed endpoints, or evaluated_at for endpoints computed on demand for this request.
    - `source_timestamp` string, date-time, nullable: Timestamp of the upstream market observation the payload derives from. This is what freshness_status is measured against. Null when the endpoint has no upstream feed or returned no rows.
    - `latest_event_at` string, date-time, nullable: Newest event timestamp present in the payload (last bar, last trade, last scored row). Best-effort on very large responses, which are sampled rather than fully walked.
    - `age_seconds` number, nullable: evaluated_at - source_timestamp, in seconds. Null when source_timestamp is null.
    - `market_session_status` string, required: pre-market | regular | after-hours | closed (US/Eastern).
    - `expected_update_cadence` string, nullable: ISO-8601 duration for how often this endpoint's data is expected to change right now, e.g. PT60S. Null means no updates are expected (weekend, holiday, or a feed that only runs during the session).
    - `expected_update_cadence_seconds` number, nullable: expected_update_cadence as a number of seconds.
    - `cadence_profile` string, required: Identifier of the cadence class this endpoint belongs to, e.g. analytics_cycle. Stable across releases; the seconds behind it may be retuned.
    - `stale_after` string, date-time, nullable: Instant after which this payload should be treated as stale. Null whenever expected_update_cadence is null.
    - `freshness_status` 'fresh' | 'aging' | 'stale' | 'session_closed' | 'static' | 'unknown', required: The rolled-up freshness verdict for one response. ``fresh`` ``source_timestamp`` is within one expected update cadence. ``aging`` Past one cadence but not yet past ``stale_after`` — the grace band that absorbs normal cycle jitter. Not an error; a client that polls faster than the cadence sees this routinely. ``stale`` Past ``stale_after`` while updates *were* expected. This is the actionable state: the feed behind this endpoint is late. ``session_closed`` A feed-backed endpoint whose feed is not due to produce anything right now (weekend, holiday, or a feed that only runs during the session). The payload is the last good value — deliberately NOT reported as ``stale``, because an overnight consumer polling a closed market is not observing a fault and should not page on it. ``static`` The endpoint is not feed-backed at all: completed history, or a result computed from the caller's own inputs. Age here carries no health meaning and never becomes a fault. ``unknown`` No source timestamp could be resolved from the payload (e.g. a pure calculator, an admin endpoint, or an empty result set). A consumer should fall back to ``evaluated_at`` for endpoint health and make no claim about data age.

## Other responses

- `422`: Validation Error

---

[API](https://skmtc.dev/zerogex/apis/zerogex-api.md) · [All operations](https://skmtc.dev/zerogex/apis/zerogex-api/llms.txt) · [OpenAPI document](https://skmtc.dev/zerogex/apis/zerogex-api/revisions/753b26bf0592?raw)