Date: 2026-09-08 JST. Research only; no new trade instructions and no exchange order writes. Do not change the separate live trading handoff or Strategy ledger from these results.

User supplied RSI Trigger + D/W/M/Y Pivot v2.16 and requested an NVDA-first backtest of pivot reliability on tradeXYZ, reusable for other symbols. The supplied code is an indicator, not a complete strategy. This research tests pivot-only experimental entries/exits, NOT the full RSI/YENZOU/divergence/star state machine. Original attachment SHA256: 695571c2f7a4107419b7e8dd82c95d437d52d1ba6f2f633ee3135f0ead109fcd; 106709 bytes. The original source was included in the user's local download package, not published in this val.

Val: kamibaba/hyperliquid-market-bridge; storage type: val (project-scoped). Blob prefix: research:pivot216:NVDA:20260908

• :manifest = acquisition record, asOf 2026-09-08T00:32:12.227Z.
• :raw: = source payload, request, fetch timestamp and SHA256.
• :bars: = normalized closed OHLC bars.
• :summary:5m / :summary:15m / :summary:1h = compact completed scenario results.
• :result: = full individual simulated trades, statistics, chronological split, cost sensitivities, D/W/M, side and session attributions, 20 shifted-grid controls. Dataset names: xyz_NVDA_5m (5038), xyz_NVDA_15m (5012), xyz_NVDA_1h (5002), xyz_NVDA_1d (300), cash_NVDA_5m (4681 raw normalized, 4680 clean), cash_NVDA_1d (502). Scenario IDs: XYZ_{5m,15m,1h}{CASH_RTH,XYZ_UTC}{TOUCH,REJECT}; plus CASH_5m_CASH_RTH_{TOUCH,REJECT}. Total14 primary scenarios, 280 controls. Files: PIVOT216_RESEARCH_PROTOCOL.md (frozen before outcome inspection), pivot216_data_collect.ts, pivot216_research_engine.ts, pivot216_research_run.ts. Runner's current saved config is 1h. Do NOT rerun the collector into the same prefix: it uses rolling dates and can overwrite archived inputs. Preserve this baseline and use a new prefix for new dates/symbols.

All14 scenarios had negative average R and negative average basis-point returns after ASSUMED 6bps (0.06%) round-trip aggregate trading friction. Funding EXCLUDED. This is not a measured all-in actual trading Net and not proof that pivots or the full original indicator cannot work. TOUCH primary results, mean cost-adjusted R:

• Cash-RTH-derived levels on XYZ5m: n193, win37.8238%, meanR -0.689562.
• Cash-RTH-derived levels on XYZ15m: n353, win47.3088%, meanR -0.324435.
• Cash-RTH-derived levels on XYZ1h: n714, win47.3389%, meanR -0.154107.
• XYZ-UTC-derived levels on XYZ5m: n191, win35.6021%, meanR -0.797978.
• XYZ-UTC-derived levels on XYZ15m: n389, win43.7018%, meanR -0.428866.
• XYZ-UTC-derived levels on XYZ1h: n747, win46.0509%, meanR -0.194299. Cash-based REJECT on XYZ5m/15m/1h: n125/271/584, meanR -0.635427/-0.309774/-0.177379. Full14 rows in summary blobs. The14 early70% and late30% segments were also negative at6bps. This is a retrospective chronological check, not genuinely prospective holdout evidence. Different timeframes use different dates AND holding horizons; do not rank them as equivalent period tests. XYZ15m CASH_RTH TOUCH by selected level: D n253 win44.2688% meanR-0.381314; W n58 win53.4483% meanR-0.185102; M n42 win57.1429% meanR-0.174219. M equal-notional meanbps is +0.520010, unlike equal-risk R; do not hide the weighting distinction. Small M sample, not validated edge. Same scenario by signal-time session: RTH n120 meanR-0.138854; outside_RTH n180 meanR-0.385059; cash_closed n53 meanR-0.538728. Closed days had TP33/53=62.3% but cost-positive23/53=43.4%, illustrating tiny target profits versus assumed friction. Cash15m TOUCH costs0/6/12/20bps produce meanR +0.009076/-0.324435/-0.657946/-1.102628. 'Cash15m' here means cash-derived pivot levels executed on XYZ15m, NOT cash-price fills. 20 shifted-grid comparisons: actual XYZ15m CASH_RTH TOUCH at85th percentile; XYZ1h CASH_RTH TOUCH at50th. Descriptive, not a valid formal p-value.

Successful checks in completed runs: Traditional formulas, Wilder ATR, year-crossing Monday week keys, no same-period OHLC selection; execution candle OHLC/order/interval validation had bad/duplicate/gap=0 (normal cash overnight gaps excluded). pivot216_funding_collect.ts creation was blocked: no funding data acquired. pivot216_research_diagnostics.ts was created, but its run was blocked: do NOT claim its extra alignment tests or execution fixture completed, and do not expect :diagnostics or :quality_checks blobs.

No TradingView per-bar parity certificate or exact user's source/settings. Source period mapping UTC assumption; instrument-specific session/time_tradingday semantics need confirmation. Yahoo cash feed is not TradingView feed identity. Visible D/W/M tested, not full hidden R/S depths or yearly levels. R/S role reversals untested. Monthly nearest selection uses prior close to avoid hindsight selection. Study's 1ATR SL/TP and next-open entry are not original indicator rules. No historical funding, Mark triggers, L2 queues, partial fills, measured slippage, latency, large-size capacity or tick rounding. Stop-first within-bar ambiguity is conservative, not proof of real ordering. Weekly-bootstrap intervals are exploratory; multiple comparisons and prior NVDA selection remain.

First reconcile actual TradingView pivot exports with this study's point-in-time levels. Then separately define/freeze original star/RSI confirmation, trend regime and session filters and evaluate on new untouched periods. Preserve all tested variants, including failures. Do not repeatedly optimize SL/TP on this same sample and call the best variant validated. For other symbols reuse the mathematical engine, but parameterize exact execution symbol, pivot-source instrument/price multiplier, source and execution sessions/calendar/holidays/DST, tick and costs. Current NY session diagnostic and the collector are NVDA-specific; do not claim all XYZ assets are plug-and-play.