Testing analog spread over one and ten observed bars

Editorial guide prepared September 8, 2026. The study's registration and recorded decision date are September 2, 2026. This page presents existing research.

An extension on the same records

Study 90 tested whether a wider distribution of historical analog returns was associated with a larger later absolute excess move at five observed bars. Study 91 examined one and ten observed bars using the same recorded anchors and analog sets.

The original protocol states that the additional horizon statistics had not been read before this extension was registered. Their inputs had already been recorded. This is a reuse of the same sample, not an independent replication or a prospective test on new dates.

The shared starting sample contained 388 anchors. Study 90 reported five errors and 383 successful records. The checked Study 91 result table does not list the eligible denominator separately for each horizon. Those counts should not be inferred from the shared starting sample.

Two separate primary results

Each horizon had its own positive-correlation primary, with a one-sided permutation threshold of p <= 0.05. The minimum was 250 eligible anchors with at least 30 measurable members at that horizon. Both primaries were recorded as passing.

The predictor was the interquartile range of the analogs' excess returns. The target was the magnitude of the anchor's excess return, regardless of sign. These correlations are not percentages of correct forecasts.

The additional value of situation matching remained uneven

The shape-only comparison was a secondary analysis. Its reported interval included zero at one and five bars and excluded zero at ten bars. This supports a qualified historical finding at ten bars; it does not establish an advantage at every horizon.

The intervals above are the original separately reported 90% intervals. They are not simultaneous confidence bounds across all horizons and comparisons. The protocol reports the two primaries separately, and the implementation does not provide a joint multiplicity correction for the broader set of claims.

Raw coverage and abstention are sample-specific

The report gives raw analog-range coverage of 83.6% at one bar and 80.4% at ten bars, versus 81.2% and 78.9% for the respective fixed historical ranges. These historical percentages do not certify current API coverage.

The frozen abstention candidate marked 44.9% of eligible sets at one bar and 18.0% at ten bars as uninformative. Those are secondary results for the original rule, not current user-level availability rates or evidence that the thresholds are optimal.

What the review could and could not verify

The publication review checked the original protocols, shared result report and existing analysis code. It did not recompute statistics or open new market outcomes. The original registration clocks differ: the result says 13:28 UTC, while the specification says approximately 13:40 UTC. Both give September 2, which is the date displayed here; this review does not resolve the exact minute.

As in Study 90, the inference resamples individual anchors rather than whole date or issuer groups. Shared market dates, repeated symbols and overlapping outcomes can affect its uncertainty. The original pass labels remain historical verdicts, not an independent confirmation of the method.

The native return horizons count subsequent observed bars. Missing bars can extend the elapsed exchange-session horizon. A member-origin cutoff alone does not authenticate historical data vintage or prove that every member's later outcome was already available at the anchor. The underlying historical sample also selected for available five-bar returns and used stored price and liquidity fields. These limitations carry through to the reused sample.

The finding concerns the size of excess moves. It does not establish a direction, a trading strategy, a sizing rule or better future forecasts than both shape-only and volatility-based alternatives.

Read the record

The shared original result document contains both Study 90 and the Study 91 extension. The Study 91 specification records the separate horizon questions and decision criteria. The original wording and recorded verdicts are preserved.

Pre-registration: research/specs/happening_memory_dispersion_test_2026_09_02.md (study_ledger 90, registered 00:58Z; no dispersion statistic computed before it). Successor to study 89 (FAIL on direction) on the same 388 anchors and the same three sets. Ships nothing, trades nothing. Chart Library never picks a side.

The question

Does the happening-then-shape frozen set know how much the anchor will move, even though study 89 showed it does not know which way?

Receipt

388 anchors ran with HAPPENING_IDENT_INDEX=1 on both the production retrieve and the identity-only control. 5 anchors errored (the random control's 120 s timeout on very large ETF-identity pools: JPST, IJR, SPYV, QLD, AXS). 0 lookahead violations. 383 anchors returned status ok with full 300-member sets. Base rate IQR_base = 4.0569 pp (liquid excess_5d, 2019–2023, n 1,208,699), frozen before any statistic.

Primary: PASS

Secondaries

A. Beyond trailing volatility. Trailing 20-session realised vol alone has IC 0.417 with realised |excess|, and correlates 0.62 with the set's IQR. The rank-partial of IQR given trailing vol is +0.231, 90 % CI [0.141, 0.314], P(> 0) = 1.000. The set's width carries information the anchor's own recent volatility does not.

B. Controls. Identity-only random IC 0.311; shape-only IC 0.340; both far outside the null. Happening-then-shape minus shape-only on 372 common anchors: +0.067, 90 % CI [−0.006, +0.142], P(> 0) = 0.935. Situation-first is better than shape alone on width, just short of the 5 % one-sided line.

C. Coverage of the raw set's [p10, p90], no calibration layer applied.

Split by trailing-vol halves: low-vol half, happening-then-shape 86.5 % vs fixed 87.5 %; high-vol half, happening-then-shape 73.8 % vs fixed 64.9 %. The adaptive band degrades less than a fixed band when volatility is high, and still under-covers there. That residual is exactly what the vol-regime × tightness multiplier in the calibration layer corrects, and it is why the layer exists.

D. Abstention rule (thresholds frozen at registration). r = IQR_set / IQR_base; uninformative when 0.80 ≤ r ≤ 1.25 and |median| < 0.25 pp. The rule abstains on 108 / 383 (28.2 %). IC among retained anchors 0.498; among abstained 0.094. Ratio quantiles q10 / q50 / q90 = 0.56 / 1.13 / 1.68. The rule separates informative sets from base-rate lookalikes without having been tuned on this sample.

What this means, with study 89

Same anchors, same retrieve, same universe:

This is the empirical form of the product law. The memory's information is the second moment: analog sets, calibrated widths, and abstention. It licenses surfacing the band, abstaining when the set is a base-rate lookalike, and sizing from width. It licenses no side.

Study 91 — the same test at 1 d and 10 d (ledger row 91, registered 13:28Z): PASS at both

Same records, no new retrieval. Base rates frozen before any statistic: 1 d IQR 1.7543 pp, 10 d IQR 5.8856 pp (liquid excess 2019–2023).

Width information holds at every horizon the product serves. The situation screen adds over shape alone at 10 d with the interval clear of zero, not at 1 d; at 1 d the frozen rule abstains on nearly half of sets, which is the rule doing its job (one-day width is mostly base rate). The receipt may be shown at all three horizons.

Shipped from this result (PR #100)

• services/informative.py: the frozen rule as a receipt (verdict in informative / uninformative / insufficient / unavailable, ratio, median, p10/p90, rule constants). Computed after the freeze from member 5d returns; never touches ranking or membership; never a side. Exposed as informative on /api/v1/cohort, /api/v1/cohort_analyze (always-on core key), v2 data.informative plus a warning when uninformative. Flag INFORMATIVE_RECEIPT (default on).
• liquid_date_medians (date → med1/med5/med10/n; dates with < 20 liquid names carry NULL medians) built on the box 2026-09-02 (2,044 dates) and refreshed nightly at the end of scripts/precompute_returns.py.
• Whether /app blanks the band on uninformative is left as a product decision; the members stay real analogs either way.

Next

A successor may register: the same test on 1 d and 10 d; the interaction with the calibration multiplier (does the layer's multiplier shrink when the set is informative?); and whether /app should blank the band on uninformative.

Registered 2026-09-02 ~13:40 UTC, before any 1 d or 10 d dispersion statistic was read. Tier T2, two primaries (one per horizon, reported separately; neither rescues the other). Successor to study 90 (5 d width PASS) on the SAME records: the study 90 runner already recorded per-set dispersion of 1 d and 10 d excess and the anchor's realised 1 d / 10 d excess for all 388 anchors. No new retrieval; nothing is re-run or re-cut.

The question

Does the frozen set's width information hold at the horizons the product also serves (1 d and 10 d), or is it a 5 d artefact?

Data (frozen; identical records to study 90)

• arm_c.jsonl: 388 anchors, three sets (happening-then-shape, identity-only random, shape-only), per-set ex1 / ex10 dispersion (IQR, std, MAD, p10, p90, median) and the anchor's abs_excess_1d / abs_excess_10d; trailing 20-session vol per anchor.
• Base rates from the study 89 extract, liquid excess 2019-01-01..2023-12-31, computed 2026-09-02 13:35Z before any statistic:
  • 1 d: IQR_base 1.7543 pp (p10 −2.1592, p90 +2.1424; n 1,208,699)
  • 10 d: IQR_base 5.8856 pp (p10 −6.9947, p90 +7.0654; n 1,208,696)
• Members on dates without a 10 d median (the newest sessions) are simply not measurable at 10 d; the informative floor below handles thin sets.

PRIMARY 91-A (1 d) and 91-B (10 d)

Spearman IC across anchors between the happening-then-shape set's IQR of excess at the horizon and the anchor's realised |excess| at the same horizon. Null: 2,000 permutations. PASS if one-sided p ≤ 0.05 (positive). Floor: ≥ 250 anchors with status ok and ≥ 30 measurable members at that horizon; below it UNDERPOWERED, not FAIL.

Secondaries (reported; cannot rescue)

Same as study 90 at each horizon: A partial on trailing vol; B controls and paired difference vs shape-only; C coverage of the set's [p10, p90] vs the fixed base band, split by trailing-vol halves; D the abstention rule with the SAME thresholds as study 90 (ratio 0.80..1.25 and |median| < 0.25 pp), against the horizon's IQR_base.

Kill / void

Any lookahead violation in the underlying records voids both primaries (already 0 in study 90). No re-cuts.

What a pass licenses

The same as study 90, at that horizon: surface the band, abstain on lookalikes, size from width. No side. A fail at a horizon means the receipt should not be shown for that horizon.