Stats

1. Retain supply cash $278.67, proposed $0. Live ledger has two closed operator liquidations, zero open/pending. Own mean win2.27871%, no loss sample; quarter-Kelly undefined. FRO frozen N41 mean interval[-0.874362,1.958541]% and negative AA/CENX actual-exit holdouts do not qualify. Prior CLI/vectorbt/indicator/simulation work retained, no new backtest claimed. Source https://runway.adityaperswal.workers.dev/data/stock-bars . New analyses651/652/656/657 record statistics, fundamentals, consequences and technical gate. 2. Sep8: Census Q2QFR10ET https://www.census.gov/economic-indicators/calendar-listview.html ; ASPI10–12:30ET https://ir.aspisotopes.com/news-events/press-releases/detail/112/asp-isotopes-to-hold-capital-markets-day-on-september-8 ; NYFedSCE11ET https://www.newyorkfed.org/research/calendars/i-sep26.html ; FedConsumerCredit15ET https://www.federalreserve.gov/newsevents/2026-september.htm . BRZEafterclose/call16:30ET and GMEexpectedcompleteQ2 dateonly verified in observations8923/8924 with issuerURLs. CASYafterclose retained from prior461/contextmonitor18; PDFrefreshfailed, not newly verified. 3. Sep9: CHWYbeforeopen/call08ET, ASObeforeopen/call10ET, Inditexhalfyeardateonly: primaryURLs observations8921/8922/8925. June ECEC10ET https://www.bls.gov/schedule/2026/09_sched.htm ; Q2services10ET Censusabove; monthlySTEOdateonly https://www.eia.gov/outlooks/steo/outlook.php . WeeklyoilSep10noonET https://www.eia.gov/petroleum/supply/weekly/schedule.php ; PPI/CPIoutsidewindow. No verified supply-relevantSep8/9FDAdecision found, not comprehensiveabsence https://www.fda.gov/advisory-committees . Otherfundclinicalwebcasts are not FDAdecisions. 4. New monitor58 QFR Sep8 12:30ET review after10ETrelease, bothbranches researchonly. Inspect2026Q2disclosurechange and same-vintage comparisons; home/PDF DAOidentifier discrepancy unresolved. Sources https://www.census.gov/econ/qfr/index.html and https://www.census.gov/econ/qfr/documents/QFR_Methodology.pdf . Existing43ASPI/49STEO/50weeklyoil retained. No supplymonitor due now or before nextSep7 12:30UTC holidayreview; otherfunddue6/55 not ours. Closure https://www.nyse.com/trade/hours-calendars . 5. Change-mind gates: accepted profitable ASPIshipments/newfirmofftake or demonstrated carrier incrementalnetcash -> recordprospectively and require independentcostedstockedge. Delays,dilution,lower shippedvolume orrestoration -> weaken scarcitythesis. QFRcan corroborate Q2margins,not Septemberweather. Neither branch automatically authorizestrade. Futurequalifyingorder needs freshwhole-sharequote,realistic>=1.5R,nominalrisk<=8.5099,position<=212.7475,fund<=278.67 and+1Rgainplan. No defensibleorderlevels now. 6. Fivejobs12/18/36/39/44 retained. Job39nextSep7 10:00:56UTC before nextrun; local08:03fourHTTP200 hashesmatch06UTC. Helium36next12:45after nextreview; old5success/3failedpanels remain gap. No duplicatejob. No eligiblepremisechanged. Gaps: actualnetbacks/valuation,15minexitdepth, GSCPIrenderedvalue,CASYrefresh andforeignsubsidiarycoverage; observation8946 retracts unsupported overseasplantclaim. Conclusions valid conditional on edge gate; soundness remains limited by these gaps. Scarcity may reprice before cash, but that counterexample does not calibrate payoff. 7. Approximately7minutes used; no thread overran. Finish records/lessons239–241 completed. Initial finalstrategy write exceeded4000characters and was rejected; this compact replacement is the actual saved decision. No trading tools used.

Existing job12 September4 IEX completed bars: FRO46.15 versus prior20close45.43, ADR20=1.23575; STNG82.345 versus81.46, ADR20=1.67. Both simple breakouts true but eligible false. Existing job44 six-name prior20close AND SMA50 signals all false. These are readbacks of previously recorded data, not fresh session prices; holiday does not add outcomes. Source https://runway.adityaperswal.workers.dev/data/stock-bars . Relevant validation retained: FRO frozen backtest456 N41 return20.242565%, drawdown24.419466%, win41.463415%, average win5.462704% and loss2.963577%; mean95% block interval[-0.874362,1.958541]% prevents sizing. AA/CENX exact-exit analyses615/616 lose3.940576%/21.602593% on22/26 holdout entries, unlike five-close base returns. Monte Carlo618 and indicator612 retained: resampling adds no independent observations and selecting favorable neighbor after holdout is invalid. CLI and vectorbt tests were already recorded; no repeated/new backtest or simulation claimed this run. A defensible future order must pass physical evidence, firm cash, independently validated costed stop/target/horizon and fresh executable prices together. One-to-two measured daily ranges bounds plausible target selection but does not create a setup. No current stop or target recommendation, no proposed capital. Historical15-minute monitoring and extended-session slippage gaps remain.

Premises: (1) September 3 Hapag and September 4 Maersk notices document selected port delays; four selected source hashes are unchanged at September 7 08:03 versus 06UTC. https://www.hapag-lloyd.com/en/services-information/operational-updates/updates/2026/09/ops-update-middle-east-week36.html ; https://www.maersk.com/news/articles/2026/09/04/east-asia-typhoon-season-impacting-logistics . (2) Same Gemini voyages and Sea-Intelligence citations are not independent corroboration. (3) Carrier notice does not disclose incremental collected net cash. Prior fundamental437, including ZIM annual supplier/charter risks and Q2 debt-service bridge, is retained as prior work, not newly verified financial results. First order: fewer available vessel-days may raise freight and storage charges; a two-to-three-day port delay is a local range, not a global weighted duration. Second order: fuel, leases, insurance and lost cargo can redistribute the extra revenue to suppliers or erase it; shippers may need more working capital. Third order: substitution, reduced ordering, port recovery and new capacity can reverse the scarcity premium. These are conditional mechanisms, not measured stock alpha. Strongest counterargument: carriers could collect attractive cash before it appears in guidance. That would improve the physical/commercial premise but still needs a separately validated entry and realistic exit. ASPI qualified, accepted profitable shipments would similarly strengthen supply monetization; delay or dilution weakens it. Prior point estimates of project revenue are not probabilities. Start state flat $278.67, goal positive costed payoff, binding constraint issuer netbacks plus independent trade evidence. Valid conclusion under mandate is no order; soundness limited by selected pages, absent AIS/invoices, unavailable helium panels, ownership/FX and actual exit execution. No new downstream short or crypto proxy established.

1. ASPI Aug20 issuer forecast https://www.sec.gov/Archives/edgar/data/1921865/000119312526358072/aspi-ex99_1.htm gives70Mcf/day helium,$600/Mcf modelingprice,15%contracted volume. Python full365day nameplate=$42000/day,$15.33m/year; contracted-volume equivalent=$2.2995m/year at same assumedprice, not contracted economics or recognized cash. 85%equivalent$13.0305m remains dependent on future contracting at this static assumption. LNG2500GJ/day*$15–18 gives$13.6875–16.425m/year; combined$29.0175–31.755m consistent with over$27m forecast, not a guidance increase. No costs/uptime/acceptance bridge establishes profit. Sep8 issuer event10–12:30ET verified https://ir.aspisotopes.com/news-events/press-releases/detail/112/asp-isotopes-to-hold-capital-markets-day-on-september-8 . Strongbull firstacceptedprofitablecommercialshipments; counterexample delays/purityloss/costs/financing dilution. Positive shipment alone does not validate stock payoff. 2. New broad crosscheck Census QFR Sep8 10ET,Q2period https://www.census.gov/economic-indicators/calendar-listview.html . Sampling thresholds manufacturingassets5m,othercoveredindustries50m; noncertainty adjacentpanels7/8overlap; imputation and samplingerrors matter. https://www.census.gov/econ/qfr/documents/QFR_Methodology.pdf datedJuly27 and https://www.census.gov/econ/qfr/index.html announce2026Q2 disclosuremethodtransition but differ DAOidentifier216-16/216-26; unresolved publisher inconsistency. Do not automatically label economicjump artifact or real: inspect same-vintage revisions,suppression,TableOerrors and response. Foreignsubsidiary treatment unverified,correctedobservation follows8917. 3. Startcash278.67;goal measurable margins;bindingconstraint acceptedcash and comparable observation. QFR cannot date Septemberweather impact; old annual risk review437 remains relevant to charter/supplierpass-through, not reread claimed today. Monitor new QFR review; existing43 intact. Facts end at sources and arithmetic; margin/valuation conclusions remain incomplete.

1. Recalculate all supply trades from live query_trades in Python/numpy: FRO3.805162958%,STNG0.752256770%,meanwin2.278709864%,N2,zero observed losses. Both exited by operator liquidation onSep4 afterSep2 entries, not strategy stop/target outcomes. Source venue ledger https://runway.adityaperswal.workers.dev . No open or pending rows. 2. Illustrative IID Wilson95% lower win-probability bound34.2380% for2/2 already exposes uncertainty; shared tanker exposure and discretionary exits violate the IID/setup premise. Do not adopt this bound as calibrated probability. Average loss, execution slippage versus synchronous quotes, and stop/target-hit frequencies are unidentifiable. 3. Freeze existing empirical evidence instead of treating repeated readings as new data: FRO analysis456 N41 meanCI[-0.874362,1.958541]%; aluminum615/616 actual-ADR exits negative. https://runway.adityaperswal.workers.dev/data/stock-bars . These are distinct cohorts, not pooled experiments. No fresh backtest claimed. Conclusion valid under required numeric-edge gate: approvedsize0; soundness limited by finite selected histories and unmodeled15minfills. Counterexample profitable scarcityrepricing ahead of data may occur; it cannot supply missing probability/loss estimates.

1. Decision: retain all $278.67 supply cash. No PROPOSE, EXIT, ADJUST or FUND. There are no holdings or queued orders to manage. Equity850.99 implies212.7475 positioncap and8.5099 nominalriskcap, but both stocks fail entry and net-edge requirements. The two historical fund wins were operator liquidations, not a usable win/loss or execution distribution. 2. Fundamental588 and consequences587 find a real operating mechanism: regional aluminum premiums can benefit Century's US output, while LME-linked inputs, supplier concentration and recovering Gulf/domestic capacity limit rents. EGA315/1262cells restored onAug26 is operator evidence of a slow physical recovery, not proof of future stockreturn; CENXQ2actual margin improvement is already public. Sources https://media.ega.ae/egas-al-taweelah-smelter-restoration-reaches-25-completion-milestone/ ; https://www.sec.gov/Archives/edgar/data/949157/000162828026054300/a20260630q2ex991earningsre.htm ; https://www.sec.gov/Archives/edgar/data/949157/000162828026013788/cenx-20251231.htm . 3. Frozen generic technical holdouts600/601 are positive16.16%/4.59% over22/26 trades, but block mean intervals includezero. Indicator612 rejects picking L40 after seeing the holdout. Actual-exit sensitivity615/616 loses3.94%/21.60%, stress8.08%/25.58%. Statistical610 shows shared metal exposure without significant daily intercepts. Simulation618 gives model-conditioned target frequency26.72%/19.18%, not calibrated future probabilities. Correct Kelly units separately; relevant ADR net means are negative and long-only approvedsize iszero. Technical611 confirms both below prior20closinghighs. Source https://runway.adityaperswal.workers.dev/data/stock-bars . 4. Five active collectors retained. Passed local extension of43 then replaced it with44 daily27initialruns. Remote06:15 output verified11lines,6barrows,4JSON,3HTTP200;26remain,nextSep8 06:15. EGAindex https://media.ega.ae/en/ monitors changed restoration titles alongside unchanged UAMY/LRV source coverage. Existing12/18/36/39 persist. Carrier39 new06UTC success has unchangedhashes; USDA18 refreshed04:30but Aug23–29period predatesquota; helium36 still3sourcefailures. No additional datedcatalyst verified, no inventedmonitor needed. Existing supply monitors unchanged, none due before next08UTC research. 5. Next run: inspect genuinely new deliveredtonnes, power costs, premium realization and completed restarts; if rents improve, recordprospectively and require a new independentlyvalidated trade rule, not a rerun/retune of this holdout. If recovery accelerates or inputs rise, weaken scarcitywatch. Full China/global inventory balance, hedge-cohort netbacks, restrictedcash reconciliation, historical first-news labels and15-minute executable exit model remain gaps. No assertion those missing facts are false. Primary collection ceased changing the answer; no thread overran its allowance. Research used about17minutes of42allowed, finishreserve intact. All eight analysis kinds, measured outputs, individual holdout trades and three lessons recorded.

1. Position-sizing illustration uses each main-rule holdout's chronological net trade returns, resampling circular four-trade blocks into 10,000 paths of20 trades at hypothetical10% capital allocation, seed90607. AA ending return5/50/95 percentiles -2.9641/+1.5116/+7.6594%; loss frequency31.87%,95th-percentile trade-mark drawdown3.7797%. CENX -5.2178/+.8558/+7.6125%; loss41%,DD95=6.3986%. These are model-conditioned empirical scenarios, not future probability estimates; trade-mark drawdown omits adverse movement within trades and joint same-sector risk. 2. Separate exit simulation samples whole observed five-session OHLC path units from the22/26 matched entry cohorts, seed90707,10,000 draws. With priorADR stop and2ADR target, simulated stop/target/timeout percentages are AA49.74/26.72/23.54 and CENX57.03/19.18/23.79. IID bootstrap target-frequency95% intervals are9.09–45.45% and3.85–34.62%; underlying observed counts are AA11stops/6targets/5timeouts andCENX15/5/6. Sampling repeats evidence;10,000 paths are not10,000 independent events. Stop-first conflict policy and opening-gap fills; no ambiguous historicalbars. Actual 15-minute polling and extended-session depth remain unmodeled. 3. Exact fee-inclusive ADR mean returns are AA-.0762645%,CENX-.7722893%; bootstrap mean95% intervals[-1.9802,1.9385]%/[-2.9093,1.4994]%. Baseline compounded returns-3.9406%/-21.6026%, stress-8.0756%/-25.5751%. Exact figures supersede earlier approximate subtract40bp means; individual trades recorded. 4. Main-rule point quarterKelly numbers.036369/.018490 are binary unit-risk fractions, not stock allocation fractions. Stock-capital conversion under a two-outcome approximation would divide by average fractional loss; those optimistic figures are not adopted. The relevant ADR distribution has negative net mean, so long-only concave expected-log optimum is atzero; moreover base mean intervals spanzero and currententryfails. ApprovedcapitalUSD0, despite fundcash278.67, perpositioncap212.7475, equityrisk8.5099. Two actual operator-liquidation wins with no loss sample cannot substitute. Source https://runway.adityaperswal.workers.dev/data/stock-bars ; analyses600/601,615/616 and exact exit observations8553/8555. No probability assigned to unverified future tariff/war paths. Preserve cash.

Frozen note593 requested ADR exit sensitivity. Preserve all 26 main-rule holdout entries and original next eligibility; early exits do not enable additional entries. At next open set stop one prior20 average daily high-low range below entry and target two ranges above. Exit fifth session close if untouched. Model an opening gap at its opening price and a same-bar conflict stop-first; none of the observed cohort had ambiguous touches. This is a matched-entry holdout sensitivity, not a separately selected strategy with a fresh independent validation set. SIP all-adjusted2025–Sep4 2026 source https://runway.adityaperswal.workers.dev/data/stock-bars . Vectorbt long-only from_orders with synthetic open/exit accounting steps allows same-day entry and exit. Mark pre-exit at actual daily open, close after exit; this is an accounting proxy, not observed intraday timestamps. Entry/exit counts and compounded net equity returns reconcile. Fees10bp each side plus10bp slippage baseline,20bp stress. Exact baseline {"slippage":0.001,"return_pct":-21.602592602959675,"drawdown_pct":22.82503128442278,"n":26,"win_pct":38.46153846153847,"avg_win_pct":5.814500996130267,"avg_loss_pct":4.889033270116533,"mean_pct":-0.7722893215600712}. Stress {"slippage":0.002,"return_pct":-25.575087102594818,"drawdown_pct":25.575087102594818,"n":26,"win_pct":38.46153846153847,"avg_win_pct":5.603082989308801,"avg_loss_pct":5.079065551359626,"mean_pct":-0.97054688187177}. Returns use both entry and exit fees in the denominator and numerator; supersede initial approximate ADR mean figures. Drawdown is sampled open/close, omits unknown between-sample excursions. No calibrated15min sampling/executablequotes means true venue performance could be worse. Conclusion: required stop-target economics have negative expectation and compounded return. The positive five-session-close backtest cannot be transplanted. Long-only Kelly optimum is zero for the observed negative-mean exit distribution; no proposal.

Frozen note593 requested ADR exit sensitivity. Preserve all 22 main-rule holdout entries and original next eligibility; early exits do not enable additional entries. At next open set stop one prior20 average daily high-low range below entry and target two ranges above. Exit fifth session close if untouched. Model an opening gap at its opening price and a same-bar conflict stop-first; none of the observed cohort had ambiguous touches. This is a matched-entry holdout sensitivity, not a separately selected strategy with a fresh independent validation set. SIP all-adjusted2025–Sep4 2026 source https://runway.adityaperswal.workers.dev/data/stock-bars . Vectorbt long-only from_orders with synthetic open/exit accounting steps allows same-day entry and exit. Mark pre-exit at actual daily open, close after exit; this is an accounting proxy, not observed intraday timestamps. Entry/exit counts and compounded net equity returns reconcile. Fees10bp each side plus10bp slippage baseline,20bp stress. Exact baseline {"slippage":0.001,"return_pct":-3.940575738369189,"drawdown_pct":23.249139218614435,"n":22,"win_pct":40.909090909090914,"avg_win_pct":5.161940272449477,"avg_loss_pct":3.70271398132087,"mean_pct":-0.07626451386936436}. Stress {"slippage":0.002,"return_pct":-8.075563237851439,"drawdown_pct":24.46738388717066,"n":22,"win_pct":40.909090909090914,"avg_win_pct":4.9518260858627805,"avg_loss_pct":3.8951165356090667,"mean_pct":-0.2759127359160381}. Returns use both entry and exit fees in the denominator and numerator; supersede initial approximate ADR mean figures. Drawdown is sampled open/close, omits unknown between-sample excursions. No calibrated15min sampling/executablequotes means true venue performance could be worse. Conclusion: required stop-target economics have negative expectation and compounded return. The positive five-session-close backtest cannot be transplanted. Long-only Kelly optimum is zero for the observed negative-mean exit distribution; no proposal.

1. Freeze rule and parameter roles before prices (note593). Removing SMA50 from L20 lowers holdout return from16.16 to12.83% for AA and4.59 to2.49% for CENX; this is descriptive, not an independent filter experiment. Neighbor L10 loses16.47%/5.93%, while L40 gains30.70%/28.24%. Selecting L40 now would reuse test information. 2. The economically relevant combination is physical restoration, realized margins, a valid entry, and a net payoff for the actual exit rule. Century's measured Q2 margin improvement does not supply a first-news price label or validate the generic breakout. Neither current price passes L20 even though both exceed SMA50. 3. Changing only the exit on the same holdout entry cohort to priorADR stop/2ADR target/five-session timeout gives AA-3.94% and CENX-21.60%. Therefore base-rule win rates cannot support a stop-target proposal. Volatility subgroup means in statistical analysis do not authorize switching to a low-vol variant. Sources: https://runway.adityaperswal.workers.dev/data/stock-bars ; https://www.sec.gov/Archives/edgar/data/949157/000162828026054300/a20260630q2ex991earningsre.htm . Valid decision logic requires all gates; counterexample is a rally before documentary cash arrives, but its probability is not measured. Keep size0.

Completed September4 SIP daily bars, retrieved September7, not live executable quotes. AA close49.97 is above SMA50=48.6814 but below prior20 closing high54.08; CENX46.775 is above45.6459 but below52.29. Both conjunctive entries fail. AA/CENX20-session returns are -.1997%/-8.6961%, ADR20 is2.01799/2.496905 or4.0384%/5.3381% of price. Trailing20 annualized volatility48.9472%/43.6170%; mean traded dollars237.786m/90.245m per day says nothing about holiday quote depth. Hypothetical stop/target scales, not proposals: AA47.95201/54.00598, CENX44.278095/51.76881. Positioncap212.7475 permits4 shares each, but nominalADR risk8.5099 permits4 AA or3 CENX. Actual gaps and 15-minute sampling can exceed the nominal1% equity loss. Whole-share arithmetic is an upper bound, not edge-based authorization. Binding entry and payoff gates both fail, so approved quantity0. Source https://runway.adityaperswal.workers.dev/data/stock-bars .

1. The frozen 2025–September 4, 2026 holdout has only 22 AA and 26 CENX trades. Net mean returns are 0.8684% and 0.4291%; 10,000 circular four-trade block resamples yield 95% mean intervals [-1.7738,4.2172]% and [-2.7594,3.7542]%. Both include zero. Two operator-liquidation wins in our own book provide no loss distribution or independent calibration. 2. On matched five-session windows, AA and CENX underperform uncosted XME by 1.3465 and 0.3836 percentage points per trade. This compares net strategy to gross benchmark, not clean alpha. Daily AA/CENX correlation is 0.6652. Equal initial strategy sleeves returned 10.3778% with 17.6426% close-mark drawdown and 0.3842 strategy-return correlation, but this is virtual full deployment, not authorized sizing. 3. NumPy OLS with Newey-West Bartlett lag5 covariance on 420 holdout daily returns: AA XME beta1.0600, daily intercept -0.08397%, asymptotic HAC p=.5051, R²=.4705; CENX beta1.1907, intercept+.07418%, p=.6039, R²=.3901. Descriptive exposures, no predictive stock alpha established. statsmodels unavailable; direct matrix implementation used. 4. Train-median volatility splits were determined without holdout data: AA threshold55.95% annualized, CENX69.08%. High-vol holdout means -.4450% N7 and -1.2889% N6; low-vol +1.4813% N15 and +.9446% N20. Small selected subgroups cannot justify a new low-vol rule. Two current surviving candidates, corporate changes, adjusted histories, concurrent metal exposure and no historical first-news labels limit inference. Source https://runway.adityaperswal.workers.dev/data/stock-bars ; observations8047–8078 and individual trade observations8430 onward. Conclusion: retain zero approved size; neither diversification nor a positive average repairs unresolved uncertainty.

Preregistered note593 beforeprices: prior20closebreakout ANDclose>SMA50, nextsessionopenentry, fifthsessioncloseexit,nooverlaps, no boundarycrossingtrades, longonly. Training2020–2024,test2025–Sep4 2026. SIP alladjusted1678uniqueOHLCVbars/symbolJan2 2020–Sep4 2026;positivevolume andOHLCboundsasserted,paginationcomplete. Source https://runway.adityaperswal.workers.dev/data/stock-bars . Vectorbtfees10bp+slippage10bpeachside,fullvirtualsingle-namesleeve;exitcount/longonly/compoundedfee-inclusivereturnreconciled. Assumedcostsnotmeasuredquotes;alladjustedprices dividendproxy, no exactcashdividend/15minfills. AdjacentL10/L40andtrendoff preregistereddiagnosticsonly,allresults observations8047–8064; noholdoutwinnerselected. Train={"symbol":"CENX","lookback":20,"trend":true,"part":"train","slippage":0.001,"n":65,"return_pct":118.10994452529171,"dd_pct":40.21957140118801,"win_pct":52.307692307692314,"avg_win_pct":9.301176735584354,"avg_loss_pct":6.615987911525149,"mean_pct":1.7099135961936671,"buyhold_pct":138.16993464052283}. Test={"symbol":"CENX","lookback":20,"trend":true,"part":"test","slippage":0.001,"n":26,"return_pct":4.592402599209526,"dd_pct":25.851935531208014,"win_pct":53.84615384615385,"avg_win_pct":5.8023551441461745,"avg_loss_pct":5.839605130793615,"mean_pct":0.42914270955857975,"buyhold_pct":152.5647948164147,"mean_ci_pct":[-2.759395918534919,3.7542347334998523],"point_quarter_kelly":0.01849002253815543}. Doubleslippage={"symbol":"CENX","lookback":20,"trend":true,"part":"test","slippage":0.002,"n":26,"return_pct":-0.70742500764929,"dd_pct":26.295492950057287,"win_pct":50,"avg_win_pct":6.023894703560419,"avg_loss_pct":5.566925341856218,"mean_pct":0.2284846808521005,"buyhold_pct":152.5647948164147}. Four-trade circularblockmeanCI10000resamplesseed90607spanszero. Rejectpositionapproval: positivehistoricalreturnalone doesnotestablishnetedge, currentlynoentrysignal; thisisgenerictechnicalrule,nothistoricaldisruption-eventbacktest. SeparateADRexits requiretheirownpayoffs; noportingthiswinrate tostop-targettrade. Selectionbiasfromtwosurvivingresearchnames, regimechangesandonly22/26testtradeslimitgeneralization.

Preregistered note593 beforeprices: prior20closebreakout ANDclose>SMA50, nextsessionopenentry, fifthsessioncloseexit,nooverlaps, no boundarycrossingtrades, longonly. Training2020–2024,test2025–Sep4 2026. SIP alladjusted1678uniqueOHLCVbars/symbolJan2 2020–Sep4 2026;positivevolume andOHLCboundsasserted,paginationcomplete. Source https://runway.adityaperswal.workers.dev/data/stock-bars . Vectorbtfees10bp+slippage10bpeachside,fullvirtualsingle-namesleeve;exitcount/longonly/compoundedfee-inclusivereturnreconciled. Assumedcostsnotmeasuredquotes;alladjustedprices dividendproxy, no exactcashdividend/15minfills. AdjacentL10/L40andtrendoff preregistereddiagnosticsonly,allresults observations8047–8064; noholdoutwinnerselected. Train={"symbol":"AA","lookback":20,"trend":true,"part":"train","slippage":0.001,"n":63,"return_pct":187.05010052263157,"dd_pct":32.25732078014123,"win_pct":61.904761904761905,"avg_win_pct":6.8207803493174275,"avg_loss_pct":5.829182623979365,"mean_pct":2.001746835680554,"buyhold_pct":78.67116032739527}. Test={"symbol":"AA","lookback":20,"trend":true,"part":"test","slippage":0.001,"n":22,"return_pct":16.163185209217573,"dd_pct":18.62111378205926,"win_pct":50,"avg_win_pct":5.969249060937278,"avg_loss_pct":4.232482434679379,"mean_pct":0.8683833131289493,"buyhold_pct":33.288877033875686,"mean_ci_pct":[-1.773844512645932,4.217241047401682],"point_quarter_kelly":0.036369035043773076}. Doubleslippage={"symbol":"AA","lookback":20,"trend":true,"part":"test","slippage":0.002,"n":22,"return_pct":11.162808385889111,"dd_pct":19.752482767067704,"win_pct":50,"avg_win_pct":5.75752186655539,"avg_loss_pct":4.4238265084244395,"mean_pct":0.6668476790654753,"buyhold_pct":33.288877033875686}. Four-trade circularblockmeanCI10000resamplesseed90607spanszero. Rejectpositionapproval: positivehistoricalreturnalone doesnotestablishnetedge, currentlynoentrysignal; thisisgenerictechnicalrule,nothistoricaldisruption-eventbacktest. SeparateADRexits requiretheirownpayoffs; noportingthiswinrate tostop-targettrade. Selectionbiasfromtwosurvivingresearchnames, regimechangesandonly22/26testtradeslimitgeneralization.

Ran node /app/agent/dist/backtest.js --symbols AA,CENX --strategy breakout --lookback20 --days1460 --stop-pct5 --target-pct10 (actualcommand spacesbeforevalues). Source https://runway.adityaperswal.workers.dev/data/stock-bars ;CLI usesIEXdaily,1002bars/symbol,all-window2022–Sep4 2026,itsownsignal/5%stop/10%target logic. AA71trades,-29.87397%return,43.12439%DD,33.80282%wins,mean-.41949%,BH2.24064%;CENX84trades,+14.87024%,27.47976%DD,35.71429%wins,mean.27465%,BH511.35770%. No independentholdout inCLI output; causaldisruptionexposure andvenuesamplingnotmodeled. These are diagnostic only andcannotbe usedtosize. Separate frozenSIPfee-inclusivevectorbt2020–24training/2025+holdout suppliesmainstudy; do not mixtheirfeed,exitsorwindow.

Premises: (1) Actual CENX Q2shipments130632t,primaryrevenue693.5m=>5308.806USD/t versus4733.651Q1. US6020.598/t vsIceland3406.887/t showsregionalandproductmix, not purecommodityexposure. Q2adjEBITDA326.9m;Q3guide325–345m midpointonly+2.48%, so Q2acceleration cannot be extended blindly. Source issuerAug6 https://www.sec.gov/Archives/edgar/data/949157/000162828026054300/a20260630q2ex991earningsre.htm . (2) H1CFO236m minuscapex134.4m=101.6mconsolidatedcashproxy;Hawesvillesalegain287.9m boostsnetincome;July94.3mtaxrefundnotQ2runrate. (3) FY2025Glencore54%salesand36.4%yearendownership, limited coke/pitch/cathode suppliers; 70%IcelandpowerfullyLMEindexed,majoraluminafractionLMElinked, Sebreemarketelectricity, MtHollycostofservicepowerthrough2031. Cannot passnon-indexedcostshocks automatically;take-or-pay persistsduringoutage. ActualItem1A andNote4 read at https://www.sec.gov/Archives/edgar/data/949157/000162828026013788/cenx-20251231.htm . CLI section labeledRiskFactors actuallyMD&A; originalHTML corrected provenance. (4) June30shares98,993,295 timesSep4close46.775=>4.6304bnmarketcap;plus480mdebtless343.4mcash=>4.7670bnroughEV,3.65x four-timesQ2adjEBITDA. Two-timesH1CFO-capex/marketcap=>4.39%illustrativeyield. Not normalizedEV/multiple: ignoresNCI/leases/pensions/dilution andperiodchange, earnings normalization andnewsmelter funding. Restrictedcash46.3mbalancesheet vs44.8mliquiditynarrative unreconciled; exclude fromproxy. Julycash>debt can coexist withJunenetdebt dueperioddifference. (5) AAAluminaproductionguidecutto9.5–9.6Mt duePinjarra/gas; aluminumguideunchanged2.4–2.6Mt. Q3tariffcostimprovement10m offsetbyaluminacostincrease10m; source https://news.alcoa.com/press-releases/press-release-details/2026/Alcoa-Corporation-Reports-Second-Quarter-2026-Results/default.aspx . AA is not uniformly shortagebeneficiary. Conclusion: operatingrent evidence exists, but headlinevaluation doesnot boundfairvalue or5dayreturn. Soundness requires recurringnetbacks, hedgecohorts andfuturecash; no stockentry solely fromoperatorforecast. Prior durable scarcity weakened by observedrestarts; new deliveredmargins could upgrade operatingwatch, only independentlyvalidatedstockedge upgradeszero size.

1. Start state: EGA damaged AlTaweelah March28; Aug26 315/1262 cells restarted vs227 asofAug10. Primary operator https://media.ega.ae/egas-al-taweelah-smelter-restoration-reaches-25-completion-milestone/ ; https://media.ega.ae/ega-delivers-resilient-h1-2026-performance-maintaining-operational-and-supply-chain-continuity-amid-regional-disruption/ . This is88additionalcells/16days=5.5/day;947remaining implies172days at unchanged rate, a scenario consistent broadly with Q1 2027 goal, not a completion forecast. Cells take aboutweek to stabilize and feed later restarts; cells are not deliveredtonnes. Goal profitable delivered metal; binding physicalconstraint repaired/stablecells and outboundlogistics, rather than ore scarcity alone. 2. First order: reduced Gulf output/shipments can favor unaffected domestic metal and premium realization; damaged smelter loses volume and cash. EGA H1cash-57%despiteadjustedEBITDA+11% is actual counterexample to profit=cash. Alternative routes and inventories soften shortage. One operator acrossmultiple releases is one independent origin. 3. Second order: Century USrestart andIcelandrecovery add saleablevolume. CENX Q2 unitrevenue+12.15%,tonnes+6.32%,EBITDA+41.27%; https://www.sec.gov/Archives/edgar/data/949157/000162828026054300/a20260630q2ex991earningsre.htm . Electricity/alumina counterparties share LME rises; Glencore captures trading/financing exposure, not quantified independentstockbenefit. Downstreambuyers unableto pass premium may suffer, but no buyer-specificmargin data acquired and no short proposal. 4. Third order: domestic/Gulf restarts, recycling and eventual newcapacity can erode premiums; output gains can offset pricecompression for a restarter. EGA refiner at50%doesnotbindsmelterrecovery; imports/China cap andglobaldeliveredtonnebalance unverified, so no aggregate deficitforecast. Strongcounterargument: prolonged war or slowercells preserves rents; answered with conditional operatingwatch, not assumed durablealpha. Conclusion: mechanism valid conditionally, soundness limited by netback/hedge mix, realizedpowercosts and independentdeliveredvolumes. Fresh research discovery is not fresh news: sourceAug26 already public, doesnotprove unpricedSeptemberedge.

1. Fundamental540 and consequences537 identify an operating opportunity: already-issued DLA orders lock prices while input costs can fall. Commercial GP/lb collapsed to0.36; acceptedJuly2.6m and orderpriceformula do not establish cohortcost/paidcash. Sources https://www.sec.gov/Archives/edgar/data/101538/000110465926094035/uamy-20260630x10q.htm ; https://www.sec.gov/Archives/edgar/data/101538/000110465925110313/uamy-20250930xex10d1.htm . 57.3morderclaim vs12mfiling remainsunreconciled; cannot useceilingasguarantee. 2. Fixedbacktest543 rejects: train-41.66%,holdout35trades-7.12%,DD58.25%,40%wins,avgwin12.96375%,avgloss8.05117%;blockmeanCI[-3.9867,4.7908]%. Sep4close5.20fails6.85breakout/SMA5.82. https://runway.adityaperswal.workers.dev/data/stock-bars . PointquarterKellynotaccepted underuncertainty; approved0. Simulation546notfutureprobability; twoactualtankerwinsnoloss samplecannotreplaceedge. 3. Cash278.67fund,0holdings/queues. Equity850.99,positioncap212.7475,risk8.5099. Nothingtoexit oradjust. No PROPOSE, EXIT, ADJUST ornewFUND. Operatorfundinggap doesnotalterpremises. 4. Next06UTCresearch: inspectnewacceptedshipment/batchcostandorderreconciliation only; strongeroperatingevidence requires independentstockedge beforetrade; delays/costoverruns/relaxation retainrejection. Noexactearningsdateverified, so noinventedearningsmonitor. Policyreview54setNov25beforeMOFCOMNov27clause2suspensionend, bothbranchesresearchonly. https://www.mofcom.gov.cn/zfxxgk/gkml/art/2025/art_43489ca6f2094b70954c25249702b4cc.html . 5. Sixthcollector denied5activecap; testedcombinedscript2735chars preservesjob16sixbarwatchesandaddsUAMY/LRVPDFselection. Replaced16with43,daily27initialruns. Firstremote04:15verified10lines1740chars,6barlines+3JSON,2HTTP200,26left,nextSep8 04:15. Existing12/18/36/39retained;carrier39actual01:45success4sources. SelectedPDFhashes notnewprofits. https://usantimony.com/investors/ ; https://www.larvottoresources.com/investors/asx-announcements/ . Noarmedreviewbefore06UTC. Conclusion validunderedgegate; soundnesslimits batchcosts,governmentpaymentvouchers,priceclauseredactions,actualrecoveries,anddailyvs15minexecution. Broadsourcecollectionchangedtheinitialshortagethesis; lastprimarychecksconfirmedconditionalspread ratherthanguaranteedrent. Saturationreachedwithin17minutes; nooverrun orborrowedfinish. Lessons217–219recorded; allanalysis kinds anddatapoints persisted.

Ran required CLI: node /app/agent/dist/backtest.js --symbols UAMY --strategy breakout --lookback20 --days1460 --stop-pct5 --target-pct10 (actual command uses spaces between flag and20). Source https://runway.adityaperswal.workers.dev/data/stock-bars . Output1002bars,68trades,+65.4721%,34.1623%DD,39.7059%wins,mean0.9560%,buyhold1185.9267%. This all-window diagnostic has different signal exits and fixed percentage stops; it is not the subsequent preregistered ADR/five-session long-only rule, has no independently held-out execution model in this output, and cannot justify entry. The costed vectorbt train/holdout test543 supplies the decision and rejects. Do not confuse attractive gross CLI performance with validated net stock edge.

Premises: (1) Aug25 hauled ore and Aug31 Hillgrove first production show physical progress, but tonnage is not recovered and accepted product. https://usantimony.com/wp-content/uploads/gravity_forms/2-dbc5eb4a0dee7afbc4a7138457de0fdb/2026/08/2026-08-25-US-Antimony-Corp-Continued-Mining-Success.pdf ; https://www.larvottoresources.com/wp-content/uploads/2026/08/61341464.pdf. (2) July DLA acceptance flips the first-shipment gate, but cohort costs and the57.3m/12m order reconciliation remain missing. https://www.sec.gov/Archives/edgar/data/101538/000110465926094035/uamy-20260630x10q.htm ; https://www.sec.gov/Archives/edgar/data/101538/000110465926097708/tm2622899d2_ex99-1.htm. (3) Sep4close5.20 is below SMA50=5.82 and prior20high6.85; fixed35tradeholdout loses7.1161% with58.2452%DD. https://runway.adityaperswal.workers.dev/data/stock-bars . Conclusion: a physical milestone and accepted invoice are useful, but do not jointly establish repeatable stock payoff. Review logic requires dated physical evidence, verified commercial cash margins, independently costed payoff and executable exits. This is a decision checklist, not an independently backtested new combined strategy. Strong counterexample is price rally before cash evidence; opportunity cost cannot identify its probability. Keep zero entry, retain conditional operating watch.

1. Resample corrected net-equity returns from 35 stressed historical holdout trades into 10,000 IID paths of20 trades at hypothetical10% capital allocation. Seed90407. End-return5/50/95 percentiles -7.0841/+0.1911/+8.6712%; loss frequency48.38%,95th-percentile trade-mark drawdown8.7577%. These are model-conditioned frequencies, not future win probabilities. Missing regime shifts, dependency and between-exit drawdowns make them optimistic risk descriptions. Four-trade block mean interval[-3.9867,4.7908]% reinforces uncertainty. Source https://runway.adityaperswal.workers.dev/data/stock-bars . 2. Separate deterministic economic grid, no assigned probabilities: DLA implied31.7073/lb from accepted82k/$2.6m; assumed all-in costs13.34,20,28/lb. Accepted82k yields GP1.50612/0.960/0.304m; hypothetical163k yields2.99387/1.90829/0.60429m; hypothetical289k yields5.30815/3.38341/1.07141m. Highest volume includes shipments not yet accepted as ofAug11 and some discussed forQ4, so notQ3forecast. At fixed existing7m quarterly operating deficit, evenbest illustrated289k case leaves1.69185m deficit, though othergrowth andcostchanges could alterthis. Cost13.34 is commercial proxy, not measured military-ingot cost; samepriceacrosscohorts approximate. https://www.sec.gov/Archives/edgar/data/101538/000110465926094035/uamy-20260630x10q.htm ; https://www.sec.gov/Archives/edgar/data/101538/000110465926097708/tm2622899d2_ex99-1.htm . 3. Hardlimits: equity850.99,position212.7475,fund278.67,risk8.5099. Point quarterKelly0.6842% capital(base),0.2811%(stress) assumes calibrated historicaldistribution; lower meanbounds negative, trainfails, entryfalse =>approved0. Two actualoperatorliquidations have no loss sample. No MC result repairs those gaps; no stop/target order proposed.

Source https://runway.adityaperswal.workers.dev/data/stock-bars , adjustedSIPdailycompletedSep4. Close5.20<prior20closinghigh6.85 andSMA50=5.82; fixedconjunctiveentryfalse.20sessionreturn-24.0876%,samplehighdrawdown-70.2347%,ADR20=.391045(7.5201%close),20daymeantradednotional42.988m. A largepricefalldoesnotprovecheapness orentirefundamentalbadnewspriced; corporate/regimechangesconfoundhistoricaltape. Hypothetical1ADRstop/2ADRtarget from5.20 wouldbe4.808955/5.98209; onlyscaleillustration, notqueuedexitsbecauseentryandedgefail. Single-share nominalADRrisk0.391045<equityriskcap8.5099,butovernightgapand15minsamplingcanexceedit. No executablebidask ortruecurrentextendedsessiondepthverified;marketclosed. Independentlycostedrulefailedso nofalling-priceentryorparameterchange.

1. Using issuerQ2roundedunitprices,costsandpounds, decomposeGPchange atpriorunitspread: volume(428425-340305)*8.47=+746376.4;price428425*(13.70-28.32)=-6263573.5;cost428425*(19.85-13.34)=+2789046.75;net-2728150.35,within75.65ofexact-2728226fromrounding. Sourcehttps://www.sec.gov/Archives/edgar/data/101538/000110465926094035/uamy-20260630x10q.htm. Identity notcausalregression: mix/NRV/processingtiming canaffectunitcosts. FourQ1/Q2comparatorsarenotindependentstatisticalevidence, no quarterlyregressionclaimed. Inventory13.964876mfinished /21.380103mSb=65.3172%; endfinishedvalue /Q2pureSbCOGS*91=222.37days descriptivecoverage, not turnoverforecast. 2. https://runway.adityaperswal.workers.dev/data/stock-bars SIPdaily2022-Jan3–2026-Sep4:1172returns. Dailyquantiles1/5/50/95/99=-14.3081/-8.75544/-0.266698/11.42564/19.10013%;SPYcorrelation0.2198245,beta1.24149. Poolednonstationarysampleincludingpennystockperiod; relationshipnotcausalantimonyexposure. 3. Fixed35tradeholdoutIIDmeanCI[-3.1859,4.1258]% widens4-tradeblock[-3.9867,4.7908]%; resamplingpreservesonlyshortlocaldependence, notfuturepolicyregime. Rejectedrulecannotbequalifiedbypositivearithmeticmeanwhilecompoundedreturnnegative. 4. Live query_trades fund=supply ids6/3 onlytwooperatorliquidations,FRO3.80516% andSTNG0.75226%,mean2.27871%,zeroobservedloss. Reuseanalysis469; noindependentnewoutcomes. MissinglossandcensoredexitsmakefundrecordKellyundefined. Conclusion: sourcesestablishcommercialsqueeze, notstockpayoffprobability;approvedsize0.

Preregistered note552 before customtest; UAMY selected after today'sfilings, so historicalholdout notpristineprospective. Datahttps://runway.adityaperswal.workers.dev/data/stock-bars, SIPadjusted daily2022-Jan3–Sep4 2026,1173bars, no pagination. Fixedclose>prior20closehigh ANDSMA50;nextopen; frozenprior20ADR,1ADRstop2ADRtarget,fifthsessionclose;stopfirstonambiguousdailybar,stopgapworseopen;flatatsplit/end,nooverlap. Fees5bp+slippage10bp eachside;stress20bp. vectorbt longonly withtwomarks/day; all35holdouttradeslong,35exits,compoundedentry-fee-inclusivePNL matchesportfolio1e-8. DailyOHLCnot15mintrigger/intradayliquidityvalidation; DD sampled twice/day mayunderstateintradayDD. Train2022–24:39trades,-41.6594%return,57.9309%DD,30.7692%wins. Held2025–Sep4 2026:35trades,-7.11606%return,58.2452%DD,40%wins,avgwin12.96375%,avgloss8.05117%,mean0.35480%;IIDbootstrap95%mean[-3.18592,4.12585]%;4-tradecircularblock[-3.98671,4.79083]%. Stress return-13.39560%,58.8669%DD,37.1429%wins,avgwin13.72262%/loss7.86336%. Primary netquarterKellypoint0.6842%capital,stress0.2811%; uncertainty-adjustedapproved0 becausemeanintervalcrosseszero,trainingfails,drawdownlargeandcurrententryfalse. Neighbors10/40 sessions testedasrobustnessonly:train-54.2387/-38.8928%;holdout+17.8054/+36.7124%,N41/25, bothmeanintervalscrosszero; notswitching tobestholdout. This rejectsfixedstockrule, notprovesnegativefundamentals. Numericaledgecannotbeimportedfromtwooperator-directed tankerwins. Internaldefaultboth-directionversiondiscarded; laterrawtrade-return denominatorcorrectedbeforeanalysis. Earlierobservationaverages supersededbycorrectedobservations; portfoliooutcomesunchanged. No trade orpaperprofitclaim.

Start flat, supplycash278.67; goal validated profitable exposure. Binding operating constraint: accept economically priced shipments and collect cash. Prior scarcity thesis narrows to order-date price locks. 1. Q2 pure-antimony volume428425lb vs340305; ASP13.70 vs28.32, cost13.34 vs19.85; GP152857 vs2881083. GP/lb falls95.75% to0.36 despite volume+25.89%. H1 OCF-20.720183m; inventorycashuse9.601887m and APpayment4.473542m together67.93% ofburn. Inventory alone46.34%, not majority. JuneSbfinishedinventory13.964876m is65.32% ofSbvalue; ore7.415227m. https://www.sec.gov/Archives/edgar/data/101538/000110465926094035/uamy-20260630x10q.htm . Counterexample: stock can precede profitable contracted sales; inventory growth alone isn't distress. 2. 82klb acceptedJuly and2.6m revenue recognizedJuly imply31.7073/lb. Later81k/2.6m intransit,126k/3.97m pendinginspectionAug11; call phases some intoQ4, not all9.17mQ3. https://www.sec.gov/Archives/edgar/data/101538/000110465926094035/uamy-20260630x10q.htm ; https://www.sec.gov/Archives/edgar/data/101538/000110465926097708/tm2622899d2_ex99-1.htm . Commercial cost13.34 is not proven military-ingot cost; cohortmargin remains unknown. 3. Executed DLAcontract245m ceiling,5000original guaranteedminimum; laterorders ArgusRotterdam*(1+redactedprocessingfactor)+1/lbhandling, fixed for each order after issuance. No verifiedfiveyearfloor. Higherpriced orders may monetize cheaper feedstock: inference, not auditedcohortprofit. https://www.sec.gov/Archives/edgar/data/101538/000110465925110313/uamy-20250930xex10d1.htm . July1release andAug11call report57.3m orders,10Qnote5retains12m. Cannot fully reconcile by period:57.3m is issuerreported, not independentgovernmentconfirmation. https://usantimony.com/wp-content/uploads/gravity_forms/2-dbc5eb4a0dee7afbc4a7138457de0fdb/2026/07/2026-0701_DoW-Shipments-Update-PR_Final.pdf . 4. Riskfactors: historically mostore fromoneCanadiansupplier, top3customers80%2025revenue; suppliers tookgreater marketpriceshareH22025; ownedmine seasonal. https://www.sec.gov/Archives/edgar/data/101538/000110465926032049/uamy-20251231x10k.htm . June4msupplierconvertibleincludes356570prepaidundeliveredstock; belowmarketinputsclaimed buttotalcostunknown. Successfulownedore andhydromet couldlowercost. https://www.sec.gov/Archives/edgar/data/101538/000110465926094035/uamy-20260630x10q.htm 5. Sep4close5.20 *Aug7shares149669384=778.2808m indicativecap,10.38–12.97x60–75mguide. Sharecountdated, subsequentdilution/buybackunverified. Sourceshttps://www.sec.gov/Archives/edgar/data/101538/000110465926094035/uamy-20260630x10q.htm;https://runway.adityaperswal.workers.dev/data/stock-bars . SubtractingJune cash,Treasuries,Larvottomarks gives672.896m proxy, NOT formalEV: no liability,tax,illiquidityadjustments. Buyback100m merelyauthorization https://usantimony.com/wp-content/uploads/gravity_forms/2-dbc5eb4a0dee7afbc4a7138457de0fdb/2026/08/2026-08-19-Share-repurchase-program.pdf . Conclusion: reject automatic scarcity-buy, watch acceptedordercohortspread. Valid conditionalreasoning: unknown cash/payoff cannot passrequirededgegate. Soundnesslimits: issuerinterested; PR/transcript/10Q oneorigin; no batchcostledger,governmentpaymentvouchers,currentArgusinvoice or independentstockeventedge. Strongestbullcase is highpricedorders+cheapinventory+grant yieldingH2cash; preserveit. Upgrade operatingconfidence when acceptedcohortGP coversincrementaloverhead andOCF convertswithoutnewinventory/financingbuild. That supplies no calibratedstockwinprobability; size0 pendingseparateedge.