kamino_xstocks_weekend_20260417.md

Kamino xStocks — weekend comparator: 2026-04-17 → 2026-04-20

Generated 2026-04-27T12:36:08.874362+00:00; reserve config snapshot from 2026-04-26. Forward-running comparator that scores Soothsayer bands and free-data baselines against the realized Monday open under Kamino's actually-deployed xStock reserve parameters. All 8 xStocks live in lending market 5wJeMrUYECGq41fxRESKALVcHnNX26TAWy4W98yULsua and consume Scope as the primary oracle. The numbers below are reproducible from the on-chain klend program (KLend2g3cP87fffoy8q1mQqGKjrxjC8boSyAYavgmjD) plus the historical reserve snapshot artifact consumed by scripts/score_weekend_comparison.py; the original scripts/snapshot_kamino_xstocks.py producer was retired in the April 2026 scryer cutover.

Section 1 — What was observed

Per-symbol Friday close (Yahoo-underlier reference via Scryer), Monday open (Yahoo-underlier reference via Scryer), Kamino reserve config (LTV at origination / liquidation threshold / heuristic guard rail), and weekend tape coverage. Kamino's PriceHeuristic is a validity guard rail — Scope reads outside the range are rejected — not a coverage band, so it is reported here as a recorded incumbent parameter rather than scored against the realized move.

Symbol	Reserve PDA	LTV / liq	Gap (pp)	Heuristic guard rail	Fri close	Mon open	Realized (bps)	V5 tape obs	Scope tape obs
SPYx	UvXjBuC7…	73 / 75	2	[$515, $858]	$710.14	$708.78	-19.2	0	0
QQQx	2jerdAXR…	70 / 72	2	[$400, $700]	$648.85	$648.04	-12.5	0	0
TSLAx	5iTiczqg…	55 / 65	10	[$300, $520]	$400.62	$402.58	+48.9	0	0
GOOGLx	4wg6rEkG…	60 / 70	10	[$224, $416]	$341.68	$340.76	-26.9	0	0
AAPLx	CKJbqakb…	40 / 50	10	[$190, $300]	$270.23	$270.33	+3.7	0	0
NVDAx	7B66Az3t…	55 / 65	10	[$100, $250]	$201.68	$199.98	-84.3	0	0
MSTRx	Cwy2WJos…	30 / 40	10	[$65, $250]	$166.52	$162.30	-253.4	0	0
HOODx	4UBJu5Xp…	30 / 40	10	[$54, $100]	$90.75	$89.70	-115.7	0	0

Section 2 — Coverage context for the same weekend

Coverage and half-width comparison across the three methods that publish a coverage band: Soothsayer at τ=0.85 (deployment default), Soothsayer at τ=0.95 (stricter oracle-validation comparator), and a simple free-data heuristic (Friday close ± max(|Chainlink tokenizedPrice − Friday close|) over the V5 tape). This section is descriptive rather than load-bearing: the more important question for Kamino-shaped protocols is how the same weekend interacts with the real reserve buffer and near-threshold borrower states. Excess width is half-width minus the realized absolute move; positive values indicate over-protection, negative values indicate the band missed the move.

Symbol	Realized (bps)	Soothsayer τ=0.85 cov / hw / excess	Soothsayer τ=0.95 cov / hw / excess	Pyth (regular) cov / hw / excess	Simple heuristic cov / hw / excess
SPYx	-19.2	✓ in / 114.7 / +95.5	✓ in / 181.2 / +162.0	✗ out / 0.0 / -19.2	✗ out / 0.0 / -19.2
QQQx	-12.5	✓ in / 120.7 / +108.3	✓ in / 244.2 / +231.7	✗ out / 0.0 / -12.5	✗ out / 0.0 / -12.5
TSLAx	+48.9	✓ in / 374.7 / +325.7	✓ in / 786.4 / +737.5	✗ out / 0.0 / -48.9	✗ out / 0.0 / -48.9
GOOGLx	-26.9	✓ in / 202.7 / +175.7	✓ in / 296.4 / +269.4	✗ out / 0.0 / -26.9	✗ out / 0.0 / -26.9
AAPLx	+3.7	✓ in / 193.5 / +189.8	✓ in / 403.5 / +399.8	✗ out / 0.0 / -3.7	✗ out / 0.0 / -3.7
NVDAx	-84.3	✓ in / 309.3 / +225.1	✓ in / 754.7 / +670.4	✗ out / 0.0 / -84.3	✗ out / 0.0 / -84.3
MSTRx	-253.4	✓ in / 510.0 / +256.5	✓ in / 1244.9 / +991.5	✗ out / 0.0 / -253.4	✗ out / 0.0 / -253.4
HOODx	-115.7	✓ in / 704.2 / +588.5	✓ in / 956.2 / +840.5	✗ out / 0.0 / -115.7	✗ out / 0.0 / -115.7

Aggregate coverage on this weekend (method ✓ / total scorable rows): Soothsayer τ=0.85 = 8/8; Soothsayer τ=0.95 = 8/8; Pyth regular = 0/8; simple heuristic = 0/8.

Pyth regular is the publisher-dispersion confidence band from Pyth's regular-session feed at Friday close — the closest existing on-Solana analog to a published "band". It widens during off-hours via Pyth's aggregation rule (publishers thinning out → wider $\sigma$). The comparator's intellectual question is publisher-dispersion-based confidence vs calibration-transparent coverage; Pyth regular is the publisher-dispersion side of that comparison. One weekend is a tiny sample by design — this report is a forward-running tape; the cross-week aggregate becomes the meaningful comparison after several weekends.

Section 3 — Primary comparison: lending consequence under real reserve params

For each method's lower bound, what does the same reserve config classify a near-origination borrower (debt sized to LTV = max-LTV − 0.5pp) and a near-liquidation borrower (debt sized to LTV = liquidation threshold − 0.05pp) as on Monday? Kamino-incumbent uses the Scope-served Monday-open price (or Friday close as fallback when Scope tape is absent for past weekends) — this is the actually-deployed incumbent decision rule, not a reconstruction. Soothsayer rows use the band's lower bound as the conservative collateral price. This is the section to read first if the question is whether the closed-market signal changes reserve-buffer-relevant decisions.

Symbol	LTV / liq / gap	Method	Near-origination	Near-liquidation	Effective LTV (orig / liq)
SPYx	73 / 75 / 2pp	kamino_incumbent	Safe	Caution	0.725 / 0.750
SPYx	73 / 75 / 2pp	soothsayer_t085	Caution	Liquidate	0.742 / 0.767
SPYx	73 / 75 / 2pp	soothsayer_t095	Caution	Liquidate	0.749 / 0.774
SPYx	73 / 75 / 2pp	simple_heuristic	Safe	Caution	0.725 / 0.750
SPYx	73 / 75 / 2pp	pyth_regular	Safe	Caution	0.725 / 0.750
QQQx	70 / 72 / 2pp	kamino_incumbent	Safe	Caution	0.695 / 0.720
QQQx	70 / 72 / 2pp	soothsayer_t085	Caution	Liquidate	0.711 / 0.736
QQQx	70 / 72 / 2pp	soothsayer_t095	Liquidate	Liquidate	0.725 / 0.751
QQQx	70 / 72 / 2pp	simple_heuristic	Safe	Caution	0.695 / 0.720
QQQx	70 / 72 / 2pp	pyth_regular	Safe	Caution	0.695 / 0.720
TSLAx	55 / 65 / 10pp	kamino_incumbent	Safe	Caution	0.545 / 0.649
TSLAx	55 / 65 / 10pp	soothsayer_t085	Caution	Liquidate	0.567 / 0.676
TSLAx	55 / 65 / 10pp	soothsayer_t095	Caution	Liquidate	0.587 / 0.699
TSLAx	55 / 65 / 10pp	simple_heuristic	Safe	Caution	0.545 / 0.649
TSLAx	55 / 65 / 10pp	pyth_regular	Safe	Caution	0.545 / 0.649
GOOGLx	60 / 70 / 10pp	kamino_incumbent	Safe	Caution	0.595 / 0.700
GOOGLx	60 / 70 / 10pp	soothsayer_t085	Caution	Liquidate	0.613 / 0.721
GOOGLx	60 / 70 / 10pp	soothsayer_t095	Caution	Liquidate	0.622 / 0.731
GOOGLx	60 / 70 / 10pp	simple_heuristic	Safe	Caution	0.595 / 0.700
GOOGLx	60 / 70 / 10pp	pyth_regular	Safe	Caution	0.595 / 0.700
AAPLx	40 / 50 / 10pp	kamino_incumbent	Safe	Caution	0.395 / 0.499
AAPLx	40 / 50 / 10pp	soothsayer_t085	Caution	Liquidate	0.407 / 0.515
AAPLx	40 / 50 / 10pp	soothsayer_t095	Caution	Liquidate	0.416 / 0.526
AAPLx	40 / 50 / 10pp	simple_heuristic	Safe	Caution	0.395 / 0.499
AAPLx	40 / 50 / 10pp	pyth_regular	Safe	Caution	0.395 / 0.499
NVDAx	55 / 65 / 10pp	kamino_incumbent	Safe	Caution	0.545 / 0.649
NVDAx	55 / 65 / 10pp	soothsayer_t085	Caution	Liquidate	0.569 / 0.678
NVDAx	55 / 65 / 10pp	soothsayer_t095	Caution	Liquidate	0.615 / 0.733
NVDAx	55 / 65 / 10pp	simple_heuristic	Safe	Caution	0.545 / 0.649
NVDAx	55 / 65 / 10pp	pyth_regular	Safe	Caution	0.545 / 0.649
MSTRx	30 / 40 / 10pp	kamino_incumbent	Safe	Caution	0.295 / 0.400
MSTRx	30 / 40 / 10pp	soothsayer_t085	Caution	Liquidate	0.324 / 0.438
MSTRx	30 / 40 / 10pp	soothsayer_t095	Caution	Liquidate	0.366 / 0.495
MSTRx	30 / 40 / 10pp	simple_heuristic	Safe	Caution	0.295 / 0.400
MSTRx	30 / 40 / 10pp	pyth_regular	Safe	Caution	0.295 / 0.400
HOODx	30 / 40 / 10pp	kamino_incumbent	Safe	Caution	0.295 / 0.400
HOODx	30 / 40 / 10pp	soothsayer_t085	Caution	Liquidate	0.316 / 0.428
HOODx	30 / 40 / 10pp	soothsayer_t095	Caution	Liquidate	0.323 / 0.437
HOODx	30 / 40 / 10pp	simple_heuristic	Safe	Caution	0.295 / 0.400
HOODx	30 / 40 / 10pp	pyth_regular	Safe	Caution	0.295 / 0.400

Decision divergence summary

Cases where Soothsayer (τ=0.85) and Kamino-incumbent reach a different classification for the near-liquidation borrower. Treat this as a one-week illustration of the decision surface, not a welfare conclusion by itself:

Symbol	Realized (bps)	Kamino near-liq	Soothsayer τ=0.85 near-liq	Direction
SPYx	-19.2	Caution	Liquidate	Soothsayer more conservative
QQQx	-12.5	Caution	Liquidate	Soothsayer more conservative
TSLAx	+48.9	Caution	Liquidate	Soothsayer more conservative
GOOGLx	-26.9	Caution	Liquidate	Soothsayer more conservative
AAPLx	+3.7	Caution	Liquidate	Soothsayer more conservative
NVDAx	-84.3	Caution	Liquidate	Soothsayer more conservative
MSTRx	-253.4	Caution	Liquidate	Soothsayer more conservative
HOODx	-115.7	Caution	Liquidate	Soothsayer more conservative

Single-week decision divergence: 8 of 8 symbols on the near-liquidation borrower. The meaningful aggregate is the cross-week false-liquidation and missed-risk rate under the same reserve parameters.

Section 4 — LTV-gap breach analysis (the actually-Kamino-shaped metric)

The metrics above (coverage, decision-classification) ask Soothsayer-shaped questions. Kamino consumes a Scope-served point and applies an (LTV-at-origination, liquidation-threshold) gap, not a coverage band. The question its risk team would actually ask is: did the realized Monday move cross the threshold below which a borrower originated at max-LTV gets liquidated, and did each method's lower bound warn about that breach beforehand?

For a borrower originated at the LTV ceiling on Friday, the trigger drop is (max_ltv / liq_threshold − 1) × Friday close. Below that price the new LTV exceeds the liquidation threshold. Per-method classification is a 2×2: a method flags the breach if its lower bound is at or below the trigger price; the realized move breaches if Monday open is at or below the trigger price.

• matched — flagged AND realized → correct warning
• preemptive — flagged but not realized → safe-side false positive
• missed — not flagged but realized → dangerous false negative
• silent_safe — not flagged AND not realized → correct silence

Symbol	Trigger drop (bps)	Realized (bps)	Realized breach?	Kamino-incumbent	Soothsayer τ=0.85	Soothsayer τ=0.95	Pyth regular	Simple heuristic
SPYx	-266.7	-19.2	—	✓ silent_safe (+0bps)	✓ silent_safe (-230bps)	⚠️ preemptive (-323bps)	✓ silent_safe (+0bps)	✓ silent_safe (+0bps)
QQQx	-277.8	-12.5	—	✓ silent_safe (+0bps)	✓ silent_safe (-222bps)	⚠️ preemptive (-418bps)	✓ silent_safe (+0bps)	✓ silent_safe (+0bps)
TSLAx	-1538.5	+48.9	—	✓ silent_safe (+0bps)	✓ silent_safe (-387bps)	✓ silent_safe (-710bps)	✓ silent_safe (+0bps)	✓ silent_safe (+0bps)
GOOGLx	-1428.6	-26.9	—	✓ silent_safe (+0bps)	✓ silent_safe (-297bps)	✓ silent_safe (-432bps)	✓ silent_safe (+0bps)	✓ silent_safe (+0bps)
AAPLx	-2000.0	+3.7	—	✓ silent_safe (+0bps)	✓ silent_safe (-294bps)	✓ silent_safe (-511bps)	✓ silent_safe (+0bps)	✓ silent_safe (+0bps)
NVDAx	-1538.5	-84.3	—	✓ silent_safe (+0bps)	✓ silent_safe (-415bps)	✓ silent_safe (-1143bps)	✓ silent_safe (+0bps)	✓ silent_safe (+0bps)
MSTRx	-2500.0	-253.4	—	✓ silent_safe (+0bps)	✓ silent_safe (-886bps)	✓ silent_safe (-1933bps)	✓ silent_safe (+0bps)	✓ silent_safe (+0bps)
HOODx	-2500.0	-115.7	—	✓ silent_safe (+0bps)	✓ silent_safe (-664bps)	✓ silent_safe (-851bps)	✓ silent_safe (+0bps)	✓ silent_safe (+0bps)

Aggregate this weekend: 0 of 8 symbols realized a breach (Monday open below the LTV-gap trigger). Per-method tallies:

Method	matched	preemptive	missed	silent_safe
kamino_incumbent	0	0	0	8
soothsayer_t085	0	0	0	8
soothsayer_t095	0	2	0	6
pyth_regular	0	0	0	8
simple_heuristic	0	0	0	8

Welfare-relevant ratios over time: the cross-week ratio of matched / (matched + missed) is the recall on actual breaches, and matched / (matched + preemptive) is the precision on flagged breaches. A single weekend with no breaches doesn't distinguish the methods on this axis; the comparison emerges over multiple weekends.

Honest framing

• The Kamino-incumbent row uses the actually-deployed Scope oracle path; it is not a reconstruction or proxy.
• The Soothsayer rows are the deployed methodology serving at consumer τ; they are not tuned to this weekend.
• The simple heuristic exists to prevent the comparison from looking like 'Soothsayer vs one arbitrary number'. It is intentionally weak and is expected to be the worst of the three coverage methods most weeks.
• Coverage on a single weekend is too small a sample to support a welfare claim. The quantity that matters across weeks is the false-liquidation rate and missed-risk rate under the same reserve parameters — that aggregation is in scope for Paper 3 (reports/paper3_liquidation_policy/plan.md) once enough forward weekends have accumulated.