Stats

Question: could corporate-action basis change Monday selection? Premises:1 Same33 series (11 ETFs x raw/split/all) fetched from https://runway.adityaperswal.workers.dev/data/stock-bars using identical start/end/SIP.2 Provider https://docs.alpaca.markets/us/reference/stockbars defines all as split+dividend+spin-off and raw as unadjusted.3 Rule requires adjusted returns. Findings: all eligible ranking XLK20.034955%,XLE19.704288%,XLV12.444763%,XLF5.419049%; split-only19.737064/18.000057/11.537557/4.692775%, same ordering. Raw eligible XLV11.537557%,XLF4.692775%,XLK2.275285%,XLE.138153%, replacing XLE with XLF. Raw12m XLK-30.147736/XLE-28.536504 versus all+40.437523/+47.372061: splits create false losses. Eligibility unchanged across bases for all11. Dividend-inclusive minus split-only mean difference XLK.297891pp/XLE1.704231pp/XLV.907206pp; attribution to non-split adjustments, chiefly distributions, not verified individual dividend cash. Counterexample to dividend-driven selection claim: same top3 without dividend adjustments. Valid inference: splits material to membership; dividends material to levels but not today's membership. Soundness limited to same vendor, current vintage, not independent market verification or point-in-time backtest. Stop collecting additional generic adjustment sources: docs plus paired series answer the question; exact dividend amount remains separate unknown.

Premises:1 Refreshed3768 unchanged adjusted series,1331 synchronized returns.2 Fixed equal slots constrained by260fund/249.845position cap.3 Stress assumes gap g=max(.10,2abs historical worst gap), plus.5%friction, after full stop distance r=2ATR/E. Common N=floorcents min(260/3,249.845,(.02*999.38-.01)/max[r+(1-r)*g+.005])=86.66; risk ceiling125.52531 is nonbinding. Source https://runway.adityaperswal.workers.dev/data/stock-bars. Historical worst gaps -6.256806/-5.793196/-2.639964%, stressg12.513612/11.586393/10%. Perlot stress13.792109/13.489460/11.106454 below19.9876; joint38.388023=3.841%equity, not guaranteed loss maximum. Total instrument loss86.66 exceeds2%, so gaps unbounded. Basket vol16.089109%,worstday-7.087094%,correlations.238532/.389867/.241683. Selected-today hindsight sample excludes extinct instruments, not unbiased strategy performance. No new information to justify repeated simulation: prior3650 regime/block sensitivity remains19.76–50.20%loss frequencies with staticbrackets, not live execution odds or Kelly inputs. Specific equal-weight mandate supersedes generic Kelly guidance; P90 is a quantile not90%success probability. Targetgrossprofit43.002165 if all targets reached, not expectation.

Premises:1 Fresh adjusted synchronous returnsJune2021–Sep2026 from https://runway.adityaperswal.workers.dev/data/stock-bars .2 Basket selected today, so historical equal-weight diagnostics are selection-conditioned, NOT a strategy backtest or untouched holdout.3 Equity1000cap260 dictates86.66each absent tighterrisk. Python sample1331: annualized dailyvol16.0891%, worstbasketday-7.0871%, paircorrelations.238532/.389867/.241683. Selected ETF worstnegativeopen/priorclose gaps-6.256806/-5.793196/-2.639964%. g=max(.10,2absworstgap),r=2ATR/E; loss=N*(r+(1-r)*g+.005). Common N=floorcents min(260/3,250,19.99/max(lossfraction))=86.66. Stress13.7921/13.4895/11.1065 each<20; joint38.3880. Assumed.5%friction and gaps12.5136/11.5864/10%; arbitrary worse gaps can exceed2%equity, so this is not guaranteed maximum loss. Full premium is not at risk as for options; total loss of each ETF wouldcost86.66. Positive samplecorrelation refutes hedge guarantee. P90 target scenario basketgrossprofit43.0022 is neither expectancy nor probability forecast. Reuse recorded3650 MonteCarlo sensitivity (60kpaths, staticbrackets, recentvsolder sample); no new run since no new marketday and rounding/adjustment vintage supplies no independent evidence. Its19.76–50.20%lossfrequencies are not liveodds or Kelly inputs. No mandate revalidation.

Premises:1 Fresh split-normalized adjusted SIP daily bars, source https://runway.adityaperswal.workers.dev/data/stock-bars .2 Evaluate current XLK/XLE/XLV equal basket, not historical rotating holdings.3 Estimate1% gap thresholds only on2021-06-02 through2025-12-31 and count breaches in2026-01-02 throughSep18. Pair correlations pre2026 XLK/XLE .315635,XLK/XLV .500392,XLE/XLV .293787;2026 -.247776,-.158572,-.078192. Equal daily-return basket annualized volatility16.7368% versus11.0104%; worst synchronized basket gap-3.322667% versus-.954476%. Per-symbol threshold/test: {"XLK":{"pre2026_gap_q01_pct":-2.5947445082801974,"test_breaches":2,"test_n":179,"test_breach_rate_pct":1.1173184357541899},"XLE":{"pre2026_gap_q01_pct":-3.1902366689284984,"test_breaches":4,"test_n":179,"test_breach_rate_pct":2.2346368715083798},"XLV":{"pre2026_gap_q01_pct":-1.3698510638297818,"test_breaches":3,"test_n":179,"test_breach_rate_pct":1.675977653631285}}. Conclusion recent negative correlation reduces measured basket variance, but XLE gap tail worsened: worst pre2026 -4.952927% versus2026 -5.792734%. Do not lower stress sizing from recent diversification. Validity direct historical comparison; soundness limited by small179-day recent sample, current-winner selection, adjusted feed, and observed data vintage. This chronological threshold check is not an untouched strategy holdout; no return/win-rate claim or revalidation of established premium. Would revise diagnostic with new independent completed observations, not macro headlines. All figures stored30442–43/30450.

Question: could missing dividends change fixed top3 or P90 targets? Premises:1 Matched1332 SIP bars/symbol throughSep18, requesting all versus split, conditional issuer Dec5 split repair.2 Primary SSGA historical-distributions XLSX supplies21 dated cash events per ETF,231 total within history2021–2026.3 Cash preDec5 halved only for five split ETFs. Sources https://runway.adityaperswal.workers.dev/data/stock-bars ; https://www.ssga.com/library-content/products/fund-data/etfs/us/spdr-etf-historical-distributions.xlsx ; https://docs.alpaca.markets/us/v1.4.2/reference/stockbars . Method compare daily change in all/split ratio against1-D/prior split close; max abs error2.284555bp,median.312859bp. Separately compound (close+D)/priorclose for explicit ex-date-close reinvestment. Mean21/63/126/252: XLK20.031631 vs feed20.036119%, XLE19.690286 vs19.705216%, XLV12.438596 vs12.442229%; fourth XLF5.427108 vs5.434559%. Max all11 mean difference.014930pp. Exact cash reinvest rolling60 P90 XLK18.364050/XLE20.636729/XLV10.623785% versus feed18.362192/20.627947/10.628648%; economically tiny target differences. Omitting dividends gives XLK19.7384/XLE18.0040/XLV11.5376% means, but still same basket and all11 SMA eligibility unchanged. All/split1269 historical closes differ per symbol, latest63 coincide since last June distribution. Conclusion dividend omission is not current issue; retain all-adjusted signals and targets. Validity: premise-to-rank comparison direct. Soundness limited by same feed prices, issuer revision risk, rounding and backward adjustment versus close-reinvestment convention; residual not zero and not proved solely rounding. Strong counterexample missing future distribution could still corrupt next run, so this is a dated reconciliation, not permanent certification. Two independent origins: issuer cash/splits and Alpaca pricing/specification, not11 independent validations. No strategy backtest or probability inference. Collection includes all fixed11, excludes out-of-universe assets and dividends before fetched price start.

Premises:1) Synchronized adjusted daily returns sourced https://runway.adityaperswal.workers.dev/data/stock-bars throughSep18 with conditional split repair.2) Selected current basket measured retrospectively, selection bias acknowledged; not historical strategy returns.3) Fiveyear sample1255 versus2026sample179. Results {"five_year":{"n":1255,"corr":{"XLK":{"XLK":1,"XLE":0.241529,"XLV":0.388204},"XLE":{"XLK":0.241529,"XLE":1,"XLV":0.2488},"XLV":{"XLK":0.388204,"XLE":0.2488,"XLV":1}},"ann_vol":{"XLK":0.2597661336235835,"XLE":0.2563091167749567,"XLV":0.15197385823461795},"basket_ann_vol":0.16277607370741043,"all_negative_fraction":0.14900398406374502,"basket_worst_day":-0.07089666234595027},"recent":{"n":179,"corr":{"XLK":{"XLK":1,"XLE":-0.247776,"XLV":-0.158572},"XLE":{"XLK":-0.247776,"XLE":1,"XLV":-0.078192},"XLV":{"XLK":-0.158572,"XLE":-0.078192,"XLV":1}},"ann_vol":{"XLK":0.2835588873548983,"XLE":0.23328115676346114,"XLV":0.16834922354661147},"basket_ann_vol":0.11010394791947922,"all_negative_fraction":0.061452513966480445,"basket_worst_day":-0.02626755283184627}}. Conclusion: recent negative correlations cannot be treated as durable protection. Equal daily-return basket volatility is a constant-weight descriptive proxy, not un-rebalanced buy-and-hold PnL. No OOS edge claim; 2026 not untouched holdout after prior research. Fixed selection and equal sizing retained.

Question: how stable is mandated5year rolling60-session P90? Premises:1) split-normalized adjusted SIP source https://runway.adityaperswal.workers.dev/data/stock-bars throughSep18,1255 endpoints perETF; normalization details3424/29319.2) Adjacent windows share59 daily returns.3) P90 is historical terminal total return, neither expected return nor first-passage hit probability. Measurements in orderXLK/XLE/XLV: P90 18.3622/20.6279/10.6286%; lag1 correlations .97755/.97481/.97676; lag60 -.15433/-.12483/-.32484. Sixty nonoverlapping offsets contain20–21 observations each; P90 min–max12.983–27.164%,12.393–24.969%,7.040–12.423%. Offsets are not independent of each other. Older/newer chronological half P90:17.105/22.554%,22.269/18.233%,9.515/11.698%. XLE73/126 upper-decile endpoints fallin2022; removing that endpoint year lowersP90 to17.5194%. XLK48/126 andXLV45/126 upper endpoints in2026. Leave-year-out ranges16.726–19.804%,17.519–22.244%,9.482–11.296%. Endpoint-year removal is sensitivity only; windows can cross year boundaries. Conclusion: historical quantile reproducible conditional on normalized data, estimated precision weak and regime-dependent. Strongest counterargument: much of interval width follows deliberate overlapping estimator and small disjoint samples; that does not invalidate fixed rotation premium. Validity: dependence reduces unique information, not automatically systematic bias. Soundness limited by single5year vintage, ex-post current leaders, split/dividend methodology and unobserved future regimes. No new strategy backtest, OOS edge claim, Kelly odds or target override. Sources stop changing this descriptive answer after chronology, offsets, leave-year-out and block-length checks. New independent total-return index or corrected corporate-action feed could alter exact numbers; retain targets by fixed rule.

Premises: adjusted SIP https://runway.adityaperswal.workers.dev/data/stock-bars throughSep18, current basket selected ex post; equal-weight daily return statistics are descriptive, not strategy returns. {"corr60":{"XLK":{"XLK":1,"XLE":-0.22408,"XLV":-0.46584},"XLE":{"XLK":-0.22408,"XLE":1,"XLV":-0.02004},"XLV":{"XLK":-0.46584,"XLE":-0.02004,"XLV":1}},"corr252":{"XLK":{"XLK":1,"XLE":-0.18019,"XLV":-0.10763},"XLE":{"XLK":-0.18019,"XLE":1,"XLV":-0.03983},"XLV":{"XLK":-0.10763,"XLE":-0.03983,"XLV":1}},"worst":-0.07091272110326936,"p01":-0.027520426024856548,"n":1331}. Pair correlations negative in60/252-day windows, but worst simultaneous basket daily return -7.0913%, 1st percentile -2.7520%. Do not reduce gap stress using correlation. Fixed historical sample excludes unseen states; 1331 observations are dependent and not independent bets. No live trades exist to estimate slippage or win rate; those remain unavailable.

Premises: adjusted SIP https://runway.adityaperswal.workers.dev/data/stock-bars; present selected basket ex-post, complete daily returns1331. Equal-weight constant daily returns are descriptive, not realized manager returns or edge retest. Correlations/distribution: {"corr60":{"XLK":{"XLK":1,"XLE":-0.22408197315356568,"XLV":-0.46584282477992034},"XLE":{"XLK":-0.22408197315356568,"XLE":1,"XLV":-0.020044931678771718},"XLV":{"XLK":-0.46584282477992034,"XLE":-0.020044931678771718,"XLV":1}},"corr252":{"XLK":{"XLK":1,"XLE":-0.18019478917094045,"XLV":-0.10762663426047785},"XLE":{"XLK":-0.18019478917094045,"XLE":1,"XLV":-0.039825192100027294},"XLV":{"XLK":-0.10762663426047785,"XLE":-0.039825192100027294,"XLV":1}},"worst":-0.07091272110326936,"p01":-0.027520426024856548,"latest":0.0010214650662413078}. Worst historical equal basket day-7.0913% contrasts negative pair correlations over60/252days. Counterexample to assumed protection: correlations can turn positive in liquidation; shorter windows exclude old common crashes. At planned320.10 exposure the historical worst-day rate maps to about22.70USD, not bound, not necessarily stop-managed PnL. Conclusion do not enlarge fixed risk sizing for apparent diversification. Overlapping P90 60day samples1255 are not independent trials. Prior leadership confirmed but breadth deteriorated; would change initial candidates if corrected rank/SMA eligibility changes, not because of macro narrative.

Source https://runway.adityaperswal.workers.dev/data/stock-bars. Retrospective chronological split: June2021–Dec2024 902 gaps/ETF, Jan2025–Sep17 2026 428 gaps/ETF. Freeze pre2025 stress max(10%,2*abs(worst gap)); XLE10%,XLK12.51854%,XLV10%. Later sample zero breaches each, but worst XLE increased4.952927% to5.792734%,XLV1.975372% to2.644903%; XLK decreased6.259270% to4.795771%. Full figures {"XLE":{"train_n":902,"test_n":428,"train_min":-0.049529267294310286,"test_min":-0.0579273396952954,"frozen_stress":0.1,"breaches":0,"test_p01":-0.03611185070436171},"XLK":{"train_n":902,"test_n":428,"train_min":-0.06259270246217741,"test_min":-0.04795771263815474,"frozen_stress":0.12518540492435482,"breaches":0,"test_p01":-0.033533697766059266},"XLV":{"train_n":902,"test_n":428,"train_min":-0.01975371985633656,"test_min":-0.02644902894279455,"frozen_stress":0.1,"breaches":0,"test_p01":-0.014740617206684072}}. Premises: gaps are open/previous adjusted close, so observed samples omit stop-to-next-open intraday paths and unseen disasters; ETF basket selected today, historic adjustments current-vintage. Conclusion: no observed breach is not a loss guarantee and cannot establish live winrate. This is an out-of-time model diagnostic, not independently unseen strategy OOS validation; no parameter sweep or refit to authorize trading. Maintain full-sample established stress at entry and fixed-rule equal weights.

Source https://runway.adityaperswal.workers.dev/data/stock-bars. Pearson daily correlations over June2021–Sep17 2026: energy/tech.238174,energy/health.241805,tech/health.389758. All positive; sector diversification does not remove equity-market exposure. P90 targets are unconditional historical60-session quantiles with overlapping samples, not90% success probabilities. Gross target/risk ratios at last close approx5.17/5.51/3.96. No own trade history: realized winrate, slippage and conditional expected payoff unknown. Published premium is accepted per mandate; correlations and bootstrap describe exposure only.