Stats

Preregister1385 training-only 2022–25, lookback10, hold5; close>prior rollinghigh AND SMA50, nextopen enter, exitopen h sessions later; nooverlap, no crossyear trades. SIP adjustmentall; vectorbt fee0.1% slippage0.1% eachside. Allcash betweenpositions, compounded fullnotional pername, NOT accountallocation. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=WCC,HUBB,SPY&timeframe=1Day&feed=sip&adjustment=all. Exact metrics {"lookback":10,"hold":5,"return_pct":-40.57815190900065,"drawdown_pct":45.624296974993726,"sample":55,"win_rate_pct":49.09090909090909,"avg_win_pct":4.108344773287125,"avg_loss_pct":5.431874562324664,"mean_pct":-0.7484941612061498,"manual_return_pct":-40.57815190900056}. Manual compounded cash matched. Frozen20/5 unchanged despite ranking; this is training sensitivity, NOT holdout evidence.

Preregister1385 training-only 2022–25, lookback10, hold3; close>prior rollinghigh AND SMA50, nextopen enter, exitopen h sessions later; nooverlap, no crossyear trades. SIP adjustmentall; vectorbt fee0.1% slippage0.1% eachside. Allcash betweenpositions, compounded fullnotional pername, NOT accountallocation. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=WCC,HUBB,SPY&timeframe=1Day&feed=sip&adjustment=all. Exact metrics {"lookback":10,"hold":3,"return_pct":-64.45240571366317,"drawdown_pct":65.82122251739845,"sample":65,"win_rate_pct":49.23076923076923,"avg_win_pct":2.724789530810086,"avg_loss_pct":5.380708029812595,"mean_pct":-1.3903092307368137,"manual_return_pct":-64.45240571366315}. Manual compounded cash matched. Frozen20/5 unchanged despite ranking; this is training sensitivity, NOT holdout evidence.

Premise1 WCC supplierprepayments andreceivables absorbedquarterlycash whiledeferredrevenue offsetpart; https://investors.wesco.com/news-releases/news-release-details/wesco-international-reports-second-quarter-2026-results . Premise2 supplieragreements/rebates areconditional, not ownedcapacity https://www.sec.gov/Archives/edgar/data/929008/000092900826000008/wcc-20251231.htm . Premise3 HUBB reportedgrowth withloweradjustedmargin anddebtfundedNSI https://hubbell.gcs-web.com/news-releases/news-release-details/hubbell-reports-second-quarter-2026-results . Firstorder inference: suppliers withscarcequalifiedproducts can seekprice/prepayments, distributors coordinateavailability andfinanceinventory. Secondorder: customers mayprepay tosecureprojects, offsetting distributorcashdrain, but lessfundedcustomers maydelay andpublicpower competition cancapmargin. Thirdorder: capacityadditions/qualification/substitution couldrestoreavailability andcompresspremia while rebates andinventorycarryrisk remain. Strongcountercase: superiorlogistics/gridservices create durablevalue independentofshortages; EES/CSSmarginimprovement supportsit. Rewrite: distribution is notnecessarily the sufferer; segmentmix andfundingterms decide. Noincrementaldollarprofitforecast withoutunits, timing andnetcost. Two independentissuerorigins; slides/SEC/PRcopies notextra corroboration. Physicalleadtime/contractacceptance observations missing; conclusions areconditional supplychainstructure, not tradedcausalproof.

Premises1. WCC Q2 backlog+~60% and sales+13% coexist with OCF53.7m versus107.8m; H1OCF275.1m versus135.8m. Sources July30 issuer https://investors.wesco.com/news-releases/news-release-details/wesco-international-reports-second-quarter-2026-results .2. EES price+5%/EBITDAmargin+110bps, CSSprice+1%/margin+140bps; UBSprice+3%/margin-40bps despitebacklog+80%. https://www.sec.gov/Archives/edgar/data/929008/000092900826000021/wcc-2q2026slides.htm .3. WCC ten suppliers32%purchases, agreements often<=60daytermination; rebateincome1.4%sales conditional, receivable264m. https://www.sec.gov/Archives/edgar/data/929008/000092900826000008/wcc-20251231.htm .4. HUBB Q2organic+10%, adjustedmargin23.9% down50bps, FCF213m versus221m;~3bnNSI fundedlargelydebt. https://hubbell.gcs-web.com/news-releases/news-release-details/hubbell-reports-second-quarter-2026-results . Rawmaterialsupplierdiversity doesnot eliminatecomponentavailability risk https://www.sec.gov/Archives/edgar/data/48898/000162828026007500/hubb-20251231.htm . Conclusion: evidence supports positive demand and some realized segmentmargincapture, not universal scarcityrents or an undervaluation. Valid conditional inference; soundness limited to issuerdata, not independentcustomeracceptance. Counterexamples: backlog can rise from projectwins rather than shortages; prepayments support profitable growth but consume cash first; acquisitionmix lowers margins without weakening demand. Startstate allcash; goalstate costedpositiveedge; bindingconstraint measuredstockpayoff pluscashcapture, not narrative demand. Exact supplierleadtime/privatecontractterms and independentcustomerdata absent. Prior qualitative favorable infrastructure; would strengthen only with repeatable acceptedcash and netstockCI>0, cannot assign quantified businessprobability from these filings.

Premises1. Allcash278.67supply/account850.99 and no holdings or queuedorders; fund trade history only two operator-liquidation wins, insufficient loss/slippage distribution. Source ledger https://runway.adityaperswal.workers.dev .2. Fundamental1475/consequences1476: contracts document prospective plantgate prices; originalCMD still forecasts September helium; historical52.6kgMarch2025 followedbyhalt is not currentcommercialcash. Fullcapacityrevenue29.02–31.76m calculation doesnotrevealtrueutilization orcashcost; originalQ&A/privateacceptance/fundingremainmissing. https://ir.aspisotopes.com/news-events/capital-markets-day ; https://www.sec.gov/Archives/edgar/data/1921865/000119312526352603/aspi-20260630.htm .3. Linde reports some positiveheliumcontribution butnewgenericmomentumrule is rejected, not a causalheliumeventstudy. Costed2026 LIN-5.5797%,N8,p37.5%,win2.6768/loss2.6925%;APD-2.6456%,N7,p42.857%,win2.1451/loss2.2010%. BothmeanCIs crosszero, bothquarterKelly0; ASPI1302 remains-47.6615%,N24,p20.8333%,win16.4566/loss7.1155%,Kelly0. https://runway.adityaperswal.workers.dev/data/stock-bars .4. Simulations1477/1478 distinguish arbitraryeconomic scenarios from empiricaltradebootstrap; neitherestimatesactualvenue15minuteexitodds.5. DueFROclaimunchanged anddoesnotprovidepaidreplacementfixtures; original456meanCI[-.8744,1.9585]% unchanged. https://www.centcom.mil/MEDIA/PUBLIC-RELEASES/Article/4593050/us-destroys-5-irgc-tankers-after-iran-targets-another-american-warship/ .6. Therefore no PROPOSE,EXIT,ADJUST orFUND. Conclusionvalid underrequirededgegate; soundness limitedto selectedprimarysources andsmallhistoricalpayoffs,notproofthatstockswillfall. Strongcounterargument: reliablecontracts or firstacceptedexports maypromptareratingbeforecashisreported; reply: positivebusinessprior lacks a numerically defensible stockposition. Earlierstudiesalreadyknewdistributorclaim; newcontribution is correctedthroughput/periodbridgeandnewcostedtests,notclaimingnewnews. Due106/107resolved,successors114ASPI/115FRO06UTCaccepted. Existing12/18/36/73/116 preserved,15panel03:15footerverified,CENTCOM/Bombardiermissingcoverageexplicit. Protectedfinishbegan~04:11UTC; nofurthersearch. Researchquestionreachedsaturationoncontracts/scaleandnetcashdistinction; namegapsratherthanfillwithforecasts.

1. Independent cash-ratio verification includes both entry andexitfees/slippage and reproduces LIN-5.5797222653%/APD-2.6455982443%. Uniqueorderedbars and validpositiveOHLC checked. Daily bucket timestamps are not execution clock observations. https://runway.adityaperswal.workers.dev/data/stock-bars .2. list_jobs actual03:15 output116 confirms15panels+END andexit0; CENTCOM HTTPerror/Bombardier challenge are missing coverage. Existing jobs12/18/36/73/116 preserved. Tool lookup https://runway.adityaperswal.workers.dev .3. OriginalCMDPDF exacthref fetched200 after browser normalized plus characters returned403; JS webcast registration exposed no transcript. OriginalQ&A remains unread; no authenticated access assumed. https://ir.aspisotopes.com/news-events/capital-markets-day .4. DueFRO military primary reread confirms sameSep8five-carrier claim, not fresh cargo/IMO/insurance/fixturecash. https://www.centcom.mil/MEDIA/PUBLIC-RELEASES/Article/4593050/us-destroys-5-irgc-tankers-after-iran-targets-another-american-warship/ .5. Emptybook fromcontext and query_trades; two historicaloperatorliquidations provide no slippagebenchmark or averagedrawdown/stopcalibration. Failededge means no executablelimit,stop ortarget fabricated. Schema rejection of simulation figures arrays was corrected to scalar1477; initial attempted write was not a savedanalysis.

Ran node /app/agent/dist/backtest.js --symbols LIN,APD --strategy breakout --lookback 20 --days 730 --stop-pct 3 --target-pct 6. Source https://runway.adityaperswal.workers.dev/data/stock-bars . CLI501bars with same-close/IEX and different signal exits; costs and15minute venue routing not modeled. Do not compare as independent replication of next-openSIP10session rule. Record complete summary {"symbol":"APD","bars":501,"returnPct":1.7990376141840292,"buyAndHoldPct":8.727246170418645,"maxDrawdownPct":9.92688947378435,"winRate":35.294117647058826,"avgTradePct":0.06768413949676041,"sample":34}. Qualification uses1454–1459, which reject both. No orders.

Ran node /app/agent/dist/backtest.js --symbols LIN,APD --strategy breakout --lookback 20 --days 730 --stop-pct 3 --target-pct 6. Source https://runway.adityaperswal.workers.dev/data/stock-bars . CLI501bars with same-close/IEX and different signal exits; costs and15minute venue routing not modeled. Do not compare as independent replication of next-openSIP10session rule. Record complete summary {"symbol":"LIN","bars":501,"returnPct":-1.16069880047881,"buyAndHoldPct":1.0794706276042287,"maxDrawdownPct":8.752398381628264,"winRate":36.11111111111111,"avgTradePct":-0.022644835699121693,"sample":36}. Qualification uses1454–1459, which reject both. No orders.

CompletedSep8 SIP daily: LIN468.38 below prior20closehigh490.84 andSMA50498.9384; ADR207.182,RV2016.091% annualized. APD297.71 below309.46/SMA50300.5046;ADR205.7196,RV2018.7827%. Source https://runway.adityaperswal.workers.dev/data/stock-bars?feed=sip&adjustment=all . Neither triggers frozen20/10 breakout, both costed holdouts negative; SMA50 diagnostic only. One-to-two ADR swings are scale descriptions, not justified targets without an eligible entry. One whole share of either exceeds212.7475 accountpositioncap; fractional regular-session trading is allowed but cannot fix edge or overnight exit validation. ASPI original1302 remains disabled -47.6615%N24 and Kelly0; no retune. Combine physical supply, contracted price, accepted margin, per-share valuation, tested payoff and execution: no candidate passes all. No live quote needed to invent a limit for an ineligible rule; no stops/targets proposed.

Premises1. Frozen costed baseline results1454/1457 provide LIN p37.5%, averagewin2.67676%, averageloss2.69251%; APD p42.857%, win2.14510%, loss2.20104%. Both arithmetic expectancies negative; nonnegative quarter Kelly=0. 2. numpyseed2026090904,20000 iid20trade paths at hypothetical10%account exposure from each small holdout: LIN loss86.365%, median accountreturn-1.38357%,p95DD3.49316%; APD loss69.025%,median-.65629%,p95DD3.20783%. Source https://runway.adityaperswal.workers.dev/data/stock-bars . These conditional empirical distributions underestimate unseen tails, ignore event clustering and do not model15minute exits.3. Fund record has only two operator-liquidation winners and no loss sample; no replacement probability or slippage estimate can be extracted from it.4. Conclusion size0, not a10%proposal. Positive fundamentals or account funding pressure cannot repair numerical edge. Riskcap8.5099 and positioncap212.7475 are ceilings, not deployment targets.

Inputs: observations17195–17220 and17359; https://ir.aspisotopes.com/sec-filings/all-sec-filings/content/0001193125-26-358072/aspi-ex99_1.htm ; https://www.sec.gov/Archives/edgar/data/1921865/000119312526352603/aspi-20260630.htm . Explicit arbitrary priors: independent triangular joint utilization(.25,.70,1), uncontracted85% helium price(300,600,900 USD/Mcf), contracted15% at600 lower proxy, LNG(15,16.5,18 USD/GJ), annual operating cost(20,25.808,35m). numpy100000draws seed2026090904. Revenue=u*365*(70*(.15*600+.85*p)+2500*g); operating result=revenue-cost. p5/median/p95 revenue11.303/19.846/27.941m; operating result -17.027/-7.147/+2.416m;11.452% of draws positive. Correlations reflect chosen ranges, not estimated elasticities. Conclusion: throughput deserves measurement, but scenario frequencies cannot be used as company profit probabilities or stock win rates. Q2 expense includes noncash/ramp costs; prices, costs and throughput may be correlated. No capex, interest, taxes or NCI bridge. Higher utilization/lower costs would materially change the result; no calibrated numerical Bayesian update is defensible. Actual size0 comes from failed ASPI1302 stock test, not this illustrative model.

Startstate flat278.67supplycash/account850.99. Goalrepeatablecashreturn. BindingconstraintforASPI is acceptedprofitablejointthroughputplusfinancing; fortradebookpositiveindependentnetpayoff. Premise1 USGSFeb2026estimates2025world190m m3,US81,Qatar63,Russia18. At70Mcf/day365days usingUSGS27.737m3/Mcf,phase1=.70868m m3/year,0.373%worldor1.125%Qatar2025. Cannotclaimthisprojectreplacesa largeglobaloutage. https://pubs.usgs.gov/periodicals/mcs2026/mcs2026-helium.pdf ; https://ir.aspisotopes.com/sec-filings/all-sec-filings/content/0001193125-26-358072/aspi-ex99_1.htm . June23issuer'sRussia20%attributedtoUSGSnotmatchedby18/190=9.47%; capacity/liquidscopeorerrorunresolved,do notuse20%inshortagearithmetic.1storder: alternativeplantgatecontractsallocateprospectivepriceandlogisticsburden; existingdistributorswithtanks/cavernanddiversesources winreliabilitycontracts.2ndorder: LNGcustomerconversion/offtakeandcomponentavailabilitydetermineheliumflows; chip/medicalbuyersmaypayhigherdeliveredprices while freight/handlingabsorbpartofincrement. Actualcustomer-specificlossunmeasured.3rdorder: newUSfacilities/caverns andeventualrerouting/restoration/recyclingcanreducepressure,whileprojectfinancingandfuturecapacityalterASPIper-shareclaim. Sources annualASPIrisk https://www.sec.gov/Archives/edgar/data/1921865/000119312526151294/aspi-20251231.htm ; distributorstatement https://assets.linde.com/-/media/global/corporate/corporate/documents/investors/quarterly-earnings/investor-call-transcript-2q26.pdf . Counterexamples: contractsprotectagainstspotdeclinebutalsolimitupside; scarcitycanimproveprofitdollarswhilemargin%falls; newcapacitydoesnotguaranteesoldproduction. Factsstopatdatedfilings; chainsconditional,USGSannualnotSeptemberavailability,nostockprobabilityinferred. Nooptions/crypto/foreignsubstitute repairsmissingedge.

Numbered premises and source audit. 1. June23 ASPI contract covers ~15% of70Mcf/day at initialbase>$600/Mcf for5years, all-inplantgate; Aug6 newLNG tranche covers10%of2500GJ/day at>$16/GJ atthenFX, totalLNG75%contracted. These support prospective upstream price protection; they do not establish payments or allvolumeprices. https://www.sec.gov/Archives/edgar/data/1921865/000147793226003952/aspi_ex991.htm ; https://ir.aspisotopes.com/news-events/press-releases/detail/116/asp-isotopes-inc-announces-that-renergen-limiteds . 2. Oct29 2025 Renergen interimcoverssixmonthsAug31 2025 despiteFY2026label:52.6kgfirstheliumdeliveryinMarch,thenhaltuntiladequatefeedgas. This resolves apparent conflict withSeptember2026firstcommercialramp: differentstage/period, not proofnodeliveryever. https://www.renergen.co.za/wp-content/uploads/2025/10/4D-31-August-2025-29-Oct-2025-Final-signed.pdf . 3. Sep8originalCMD99pages stillfirstheliumweeksaway/Septemberand2Hnameplate; no acceptednewheliumsales inreadpages. https://ir.aspisotopes.com/news-events/capital-markets-day . 4. OriginalAug14Q2filing: H1Helium/LNGrevenue1.248m,operatingloss9.797m; Q2 .654m/5.798m. Q2opex4.958mplusCOGS1.494m. Annualizingtotal25.808m is a sensitivityreference, NOTcashfixedcost orfutureforecast. Fullcapacity atAug20assumptions600He/15–18LNG =>29.0175–31.755m;27mrequires85.03–93.05%jointutilization IFpricesunchanged. At75%jointutilization revenue21.763–23.816m; illustrativeflatcostoploss1.992–4.045m. 75%contractcoverage is NOTa physicalutilizationcap. Existingcostscontainramp/noncash/otheritems andfuturecosts canscale. https://www.sec.gov/Archives/edgar/data/1921865/000119312526352603/aspi-20260630.htm ; https://ir.aspisotopes.com/sec-filings/all-sec-filings/content/0001193125-26-358072/aspi-ex99_1.htm . 5. Correct1294 interpretation: synthetic$12.789/GJLNG fromexact27m is notobservedrealization; Aug20saysOVER27m,Sep8saysapproximately27m. Bothcanbetrueunderutilization/mix/cautiousrounding. No pricecutproven.6. Annualriskfactors discloseVirginiasemiconductorslimited/singlesource,pastshortages; enrichmentvendorslimitedbutnosignificantdelaystodate; fewLNGcustomersandsoleproducerstatuscanhinderdieselconversion. Risks notcurrentoutages. https://www.sec.gov/Archives/edgar/data/1921865/000119312526151294/aspi-20251231.htm .7. LindeQ2transcriptpositiveincrementalheliumdollarsbutmargin%dilutionreverified; APDQ2priceheadwind/Q3heliumvolumegrowth differperiod/contractmix. Do not inferuniformspotbeta. https://assets.linde.com/-/media/global/corporate/corporate/documents/investors/quarterly-earnings/investor-call-transcript-2q26.pdf ; https://www.sec.gov/Archives/edgar/data/2969/000000296926000034/exhibit99130jun26.htm . Conclusionvalid: producercontractrents plausible, distributorlogisticscapture demonstratedqualitatively, shareholdercashstillunquantified. Strongbull: contractedofftake,flowimprovements andlowerpostrampcostscanconvertquickly; strongbear:LNGabsorption,components,workingcapitalandfundingpreventdelivery. Soundness limitedbyissuerselectivity,confidentialterms,unreadoriginalQ&A/currentacceptance. Independentorigins3businessgroupsplusUSGS,notnumberofreleases; ASPI/Renergenreportsnotindependentcontemporaneouscorroboration. Keepwatch,size0.

Source https://runway.adityaperswal.workers.dev/data/stock-bars . Frozen1346 traininggrid lookback10/20/40 x hold5/10/20, costs20bps/side. All9 APD rows [{"lookback":10,"hold":5,"return_pct":-8.954063612187165,"drawdown_pct":27.89129971036093,"win_rate_pct":55.55555555555556,"avg_win_pct":2.316930952947609,"avg_loss_pct":3.098705762851648,"sample":63,"mean_pct":-0.09001869851872758,"quarter_Kelly":0},{"lookback":10,"hold":10,"return_pct":-30.317463956351787,"drawdown_pct":48.19066423191016,"win_rate_pct":45.09803921568628,"avg_win_pct":4.245228611401689,"avg_loss_pct":4.51172806400936,"sample":51,"mean_pct":-0.5625123084318275,"quarter_Kelly":0},{"lookback":10,"hold":20,"return_pct":-5.188465269371554,"drawdown_pct":36.85076356695909,"win_rate_pct":45.16129032258064,"avg_win_pct":6.2719794650724285,"avg_loss_pct":5.034684937908549,"sample":31,"mean_pct":0.07155059892156904,"quarter_Kelly":0.0028519943073802567},{"lookback":20,"hold":5,"return_pct":-7.375662785718904,"drawdown_pct":20.42229777243616,"win_rate_pct":56.81818181818182,"avg_win_pct":2.4288687344333324,"avg_loss_pct":3.4464970563847395,"sample":44,"mean_pct":-0.10822103887447108,"quarter_Kelly":0},{"lookback":20,"hold":10,"return_pct":-12.432149251989376,"drawdown_pct":34.72148454626669,"win_rate_pct":45.45454545454545,"avg_win_pct":4.174643338898468,"avg_loss_pct":3.999957315917777,"sample":33,"mean_pct":-0.284229745546757,"quarter_Kelly":0},{"lookback":20,"hold":20,"return_pct":-21.791169010157667,"drawdown_pct":44.17412119283325,"win_rate_pct":41.66666666666667,"avg_win_pct":5.9586308985861525,"avg_loss_pct":5.5349541536452795,"sample":24,"mean_pct":-0.7459603818821836,"quarter_Kelly":0},{"lookback":40,"hold":5,"return_pct":-11.75449386122178,"drawdown_pct":18.70927757466876,"win_rate_pct":53.57142857142857,"avg_win_pct":2.1604583643163666,"avg_loss_pct":3.3274033821603046,"sample":28,"mean_pct":-0.3874774465478021,"quarter_Kelly":0},{"lookback":40,"hold":10,"return_pct":-15.90832655761591,"drawdown_pct":22.917039943337326,"win_rate_pct":40,"avg_win_pct":4.300591269337202,"avg_loss_pct":4.117713681379697,"sample":20,"mean_pct":-0.7503917010929376,"quarter_Kelly":0},{"lookback":40,"hold":20,"return_pct":-26.641487433706722,"drawdown_pct":31.341153566635406,"win_rate_pct":29.411764705882355,"avg_win_pct":7.992894749780925,"avg_loss_pct":5.531170426731582,"sample":17,"mean_pct":-1.55350419834555,"quarter_Kelly":0}]. No bestsetting transferredtoholdout. 20000 iid bootstrap means of2026baselineclosedreturns, numpyseed2026090904:95%CI [-2.5336899222192355,1.772706274444286]%. SmallN,oneindustryandselectedfirms;notindependentheliumexperiments,no causalregressionclaimed. Baselineexpectancynegative andCIcrosses0 => quarterKelly0. Do not infer absentbullishfundamentals fromfailedmomentum.

Preregister1346 before prices. 1174 daily SIP adjustmentall bars Jan3 2022–Sep8 2026; training2022–25, untouched2026holdout. Close>prior20closinghigh, enter nextopen, exitopen10sessionslater, nooverlap, exclude incomplete final trades. vectorbt from_signals uses open execution and close marking, fees10bps each side plus 30bps slippage. Source https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all . Results {"return_pct":-5.3337237620999804,"drawdown_pct":13.038577502186953,"win_rate_pct":28.57142857142857,"avg_win_pct":2.768922374957536,"avg_loss_pct":2.1388967060732864,"sample":7,"mean_pct":-0.7366626829216228,"quarter_Kelly":0,"returns":[-0.003270522485016153,0.03521992914271023,0.020158518356440485,-0.031042024857537855,-0.06144054297913838,-0.005044987976761597,-0.006146757005210338]}. This is a generic momentum proxy, not event attribution; daily return uncertainty/smallN andunmodeled15minstops make it inadmissible for orders. QuarterKellyfloor0; actualallocation0. No refit fromholdout. Adjusted history is research total-return proxy, not raw-stop simulation.

Preregister1346 before prices. 1174 daily SIP adjustmentall bars Jan3 2022–Sep8 2026; training2022–25, untouched2026holdout. Close>prior20closinghigh, enter nextopen, exitopen10sessionslater, nooverlap, exclude incomplete final trades. vectorbt from_signals uses open execution and close marking, fees10bps each side plus 10bps slippage. Source https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all . Results {"return_pct":-2.645598244297762,"drawdown_pct":11.28183254814774,"win_rate_pct":42.857142857142854,"avg_win_pct":2.145096112192302,"avg_loss_pct":2.201042448558124,"sample":7,"mean_pct":-0.3384116368079417,"quarter_Kelly":0,"returns":[0.0007284052773516221,0.039373127716023505,0.024251350372393926,-0.02715440605650954,-0.05767476229233934,-0.0010531723219116148,-0.0021593572715644786]}. This is a generic momentum proxy, not event attribution; daily return uncertainty/smallN andunmodeled15minstops make it inadmissible for orders. QuarterKellyfloor0; actualallocation0. No refit fromholdout. Adjusted history is research total-return proxy, not raw-stop simulation.

Source https://runway.adityaperswal.workers.dev/data/stock-bars . Frozen1346 traininggrid lookback10/20/40 x hold5/10/20, costs20bps/side. All9 LIN rows [{"lookback":10,"hold":5,"return_pct":-19.761600491296278,"drawdown_pct":26.86489961989773,"win_rate_pct":52.85714285714286,"avg_win_pct":1.532830878354117,"avg_loss_pct":2.3156834111566207,"sample":70,"mean_pct":-0.28146871527237377,"quarter_Kelly":0},{"lookback":10,"hold":10,"return_pct":-2.912028394204244,"drawdown_pct":19.28402721535206,"win_rate_pct":47.16981132075472,"avg_win_pct":2.7030053392487616,"avg_loss_pct":2.4206664814486025,"sample":53,"mean_pct":-0.003840150930977718,"quarter_Kelly":0},{"lookback":10,"hold":20,"return_pct":-1.036877826513628,"drawdown_pct":26.654842999112617,"win_rate_pct":60.60606060606061,"avg_win_pct":3.331932570544688,"avg_loss_pct":4.850445775259767,"sample":33,"mean_pct":0.10857140401566011,"quarter_Kelly":0.008146278602354146},{"lookback":20,"hold":5,"return_pct":-18.02368435876111,"drawdown_pct":24.980144386774196,"win_rate_pct":40.74074074074074,"avg_win_pct":1.872858924095059,"avg_loss_pct":1.845033160628845,"sample":54,"mean_pct":-0.33033638537095816,"quarter_Kelly":0},{"lookback":20,"hold":10,"return_pct":-12.662756189904432,"drawdown_pct":22.52394542725973,"win_rate_pct":48.717948717948715,"avg_win_pct":2.553488553158216,"avg_loss_pct":2.978894876056644,"sample":39,"mean_pct":-0.2836311541314559,"quarter_Kelly":0},{"lookback":20,"hold":20,"return_pct":8.389243229303593,"drawdown_pct":23.896852489390476,"win_rate_pct":53.84615384615385,"avg_win_pct":3.6622047255392145,"avg_loss_pct":3.3403941365876424,"sample":26,"mean_pct":0.43023601994220384,"quarter_Kelly":0.029370014252743354},{"lookback":40,"hold":5,"return_pct":-24.8045615311039,"drawdown_pct":26.35194725586044,"win_rate_pct":47.5,"avg_win_pct":1.6330023678314198,"avg_loss_pct":2.7396028960037437,"sample":40,"mean_pct":-0.6626153956820408,"quarter_Kelly":0},{"lookback":40,"hold":10,"return_pct":-20.220479273479086,"drawdown_pct":23.079559623208034,"win_rate_pct":38.70967741935484,"avg_win_pct":3.223911540552905,"avg_loss_pct":3.094754031539854,"sample":31,"mean_pct":-0.6488189713749147,"quarter_Kelly":0},{"lookback":40,"hold":20,"return_pct":-3.251482795766096,"drawdown_pct":25.52943551982135,"win_rate_pct":66.66666666666666,"avg_win_pct":2.9616269437458063,"avg_loss_pct":5.889807921806323,"sample":21,"mean_pct":0.011148655228429323,"quarter_Kelly":0.000941092129443627}]. No bestsetting transferredtoholdout. 20000 iid bootstrap means of2026baselineclosedreturns, numpyseed2026090904:95%CI [-2.428616473299495,1.2897623350910392]%. SmallN,oneindustryandselectedfirms;notindependentheliumexperiments,no causalregressionclaimed. Baselineexpectancynegative andCIcrosses0 => quarterKelly0. Do not infer absentbullishfundamentals fromfailedmomentum.

Preregister1346 before prices. 1174 daily SIP adjustmentall bars Jan3 2022–Sep8 2026; training2022–25, untouched2026holdout. Close>prior20closinghigh, enter nextopen, exitopen10sessionslater, nooverlap, exclude incomplete final trades. vectorbt from_signals uses open execution and close marking, fees10bps each side plus 30bps slippage. Source https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all . Results {"return_pct":-8.553352212826976,"drawdown_pct":17.343265991710734,"win_rate_pct":37.5,"avg_win_pct":2.2664681923192864,"avg_loss_pct":3.081364718777312,"sample":8,"mean_pct":-1.0759273771160878,"quarter_Kelly":0,"returns":[0.012357522151904322,0.0203721233774302,0.03526440024024407,-0.025736281523519824,-0.03982334806232166,-0.035117753699254343,-0.020816650248338637,-0.03257420240543115]}. This is a generic momentum proxy, not event attribution; daily return uncertainty/smallN andunmodeled15minstops make it inadmissible for orders. QuarterKellyfloor0; actualallocation0. No refit fromholdout. Adjusted history is research total-return proxy, not raw-stop simulation.

Preregister1346 before prices. 1174 daily SIP adjustmentall bars Jan3 2022–Sep8 2026; training2022–25, untouched2026holdout. Close>prior20closinghigh, enter nextopen, exitopen10sessionslater, nooverlap, exclude incomplete final trades. vectorbt from_signals uses open execution and close marking, fees10bps each side plus 10bps slippage. Source https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all . Results {"return_pct":-5.579722265254931,"drawdown_pct":15.673481894169349,"win_rate_pct":37.5,"avg_win_pct":2.676755871309582,"avg_loss_pct":2.6925113042631543,"sample":8,"mean_pct":-0.6790361134233781,"quarter_Kelly":0,"returns":[0.016419087556016388,0.024465811528306097,0.03941777705496498,-0.02182739715422232,-0.035970925051279745,-0.031246470533870974,-0.016888047858745654,-0.028692724615039022]}. This is a generic momentum proxy, not event attribution; daily return uncertainty/smallN andunmodeled15minstops make it inadmissible for orders. QuarterKellyfloor0; actualallocation0. No refit fromholdout. Adjusted history is research total-return proxy, not raw-stop simulation.

Numbered premises:1. Nasdaq all-listing discovery13156non-testrows attempted; corrected12612simpletickers produced12574IEXsnapshots,12291validpairs,3297freshSep8nonETF/liquidity-filteredrows. Coveragegaps544specialformat,38missing,283invalidpairs; no claimcompletecommon-stockuniverse. https://www.nasdaqtrader.com/dynamic/SymDir/nasdaqtraded.txt and https://runway.adityaperswal.workers.dev/data/stock-snapshots . Overseas19representativeTokyo/HK/Europe/Canada/Kazakhstan/Australia names plusBrent;10cryptopairs, notexhaustivenonUSlistingcensus. Mixedclocks/FX/venueeligibilityunresolved. TANH5308%rawmove excludedpendingcorporateaction, notassumedgenuinegain.2. Primary1384/1388records SEI/OESXnewguidance/orders and cashconstraints, LEUimportdependence, CCJpricecontracts, ALMdurationnotannualtonnage,PRYrealizedmarginbutflatFCF; newMaerskcharges don'tprove rents.3. Preregistered20closehigh/nextopen10sessionholdcostedvectorbttests1312–1322,2021–25train/2026holdout,10/25/50bpsperside, rejectall11fortrading. SEI−48.9243%N6,p16.6667%,win2.2351/loss12.7596%;LEU−33.6285%N6,p66.6667%,win5.4382/loss25.6145%;OESX−31.8142%N4,p25%,win9.8284/loss14.5405%,Kelly0. PositiveHUT+11.9279%N6meanCI[-7.2235,12.9239]%,ALM+32.4291%N7CI[-7.8206,16.6725]%;CCJ/FCX/VIAValsouncertain,somebadtraining. Current-survivor/moverselection,smallN andunmodeled15minexits prohibitpointedge sizing. 27celltrainingonlysweep1331diagnostic, noholdoutretune; CLI1323–25uncosteddifferentexecutionnotvalidation.4. Statistical1330highbetaandselectionbias, simulation1389empiricalpathuncertainty,indicators1390onlyHUT/ALM/OESXtrigger butsignalsnotedge.5. query_trades supplyverifiedonly2closedoperatorwins,noopen/pending/closing;cannotestimateempiricalloss/slippageedge. https://runway.adityaperswal.workers.dev . Fund278.67cash,total850.99,positioncap212.7475,risk8.5099. Goalrepeatableprofits,bindingconstraintvalidatednetpayoff,notdesired349.01monthlyshortfall. Conclusionvalidgivenqualificationrule; soundnesslimitedbycoverage/time/primaryacceptancegaps, so0allocationratherthanclaimallstocksbad. Strongestcounterargument:newSEI/OESXfundamentalregimecouldoutrunhistoricalrule. Response:preserveprospectivemonitorandcollectcash, do notretrofitpastparameters. ExistingASPI/FROdisabled/uncertain rules remainunchanged;106/107follow04UTC. New110SEI/OESX,111MaerskpreserveSep9reviewandOct8tracking. Job116firstsaved15panelfooterverified02:18. Noorders,exitsoradjustmentsneeded; allcash. Furtherwork: independentprospectiveeventstudy,acceptedcustomer/netfixtureevidence,FX/venueand15minfills. Endwithinbudget, noextraoptimization.

Premises1. Applied lesson466: job111 actual01:45savedoutput hasall13+END; new116localtest1705chars/15panels andscheduleracceptance02:10 were insufficientalone. list_jobs02:18 now verifies116 first02:15runexit0all15+END. Source https://runway.adityaperswal.workers.dev . CENTCOMHTTPerror/Bombardierchallenge remain missingcoverage; SEI/OESXhomepage hashes do not exhaust originalreleases/filings. Existing12/18/36/73preserved,5activejobs.2. Universe IEX snapshots are completedregularsession, nonUSquotes have incompatible marketclocks, cryptoUTCdaypartial; historicalSIPsuccess notproofliveNBBO. No fresh executableentryquote acquired becauseeverycandidatefailedindependentedge; no invented limit.3. CLI baseline same-closeIEX exits differ fromcostednextopenSIPten-session tests; neither models venue15minute stops/gaps.4. query_trades supply02:15onlyclosed6FRO/3STNG,bothoperatorliquidation; noopen/pending/closing, no empiricallossdistribution orslippagebenchmark.5. Monitor20cap rejectednewMaerskreminders422; consolidatednewSEI/OESXfuturechecks108/109into110,acceptedMaersk111;106/107replacecompleted100/101. Preservefuturechecks, do not pretend rejectedmonitorarmed. Conclusion collection improved; executionqualificationstillfalse,actualsize0. Counterexample venue may route whole-sharelimits offhours, but neitheredge norvalidatedlimit/exitprovided.

https://runway.adityaperswal.workers.dev/data/stock-bars?timeframe=1Day&feed=sip&adjustment=all ; full completedSep8 SIPdaily, not IEXsnapshotprices. HUT99.49>prior20high93.545 andSMA5095.3354,ADR206.958115,RV2092.388%; ALM19.12>18.93/SMA5015.3043,ADR1.15031; OESX30.52>26.63/SMA5015.8813,ADR2.03024,RV119.5793%. SEI63.96<prior65.97 despite>SMA5059.6146,ADR3.30903,RV84.7908; LEU185.53<194.78 despite>SMA50174.6098,ADR11.498945,RV73.6905. SMA50 diagnostic, not silently appended to frozen closehigh20 rule. Only HUT/ALM/OESX signal today, all fail statistical/execution qualification; OESX holdout negative. ADR1–2multiples describe potential swing scale but are not targets: no validated entry or thesis-invalidating stop and no15minute fillhistory. Premises distinguish signal existence from signal edge; high RV plus breakout can simply represent unstable high-beta exposure (1330). Do not chase daily gain or change parameters to salvage current signal.

Source https://runway.adityaperswal.workers.dev/data/stock-bars ; frozen costed ten-session next-open rule in1312–1322. numpy seed2026090902;20,000 bootstrap means and20,000 iid resampled20-trade paths per symbol at10%equity exposure, using2026 completed trade returns; same RNG advances across names. SEI6trades loss-path probability100%,median accountreturn−18.6672%,p95DD23.2374%; LEU6 loss91.285%,median−9.4346%,p95DD21.6087%; HUT6 loss18.22%,median+5.2125%,p95DD7.2227%; OESX4 loss99.95%,median−15.8911%,p95DD22.7591%. Conditional empirical stress estimates, NOT stock forecast probabilities or15minute stop/target odds; iid assumption ignores regimes and small N badly misses tails. QuarterKelly=max(0,(p-(1-p)/(avgwin/avgloss))/4): SEI/LEU/OESX0; HUTpoint.04950,ALM.123063,CCJ.09631 do not qualify allocation because mean confidence intervals include0, samples4–7, universe selected after current moves, and actual stop execution untested. HUTmeanCI[-7.22352,12.92386]%,ALM[-7.82064,16.67254]%. Premises: payoff distribution must estimate future setup and exits before sizing; these samples do not establish either. Counterexample genuine new structural event may improve future payoffs, but that would need prospective independent evidence. Enforced actual sizing0, not pointKelly;1%equityrisk8.5099 andpositioncap212.7475 would also bind anyfuturequalified order.

Start: flat278.67 supplycash; goal repeatable cash returns; binding constraint is positive executable stock payoff, with physical delivery and funded capacity necessary but insufficient. Premise1 SEI power scarcity attracts advance-funded AI demand, but vendors/customer concentration and2.2bnremainingcapex constrain delivery (1384 and https://ir.solaris-energy.com/~/media/Files/S/Solaris-IR/reports-and-presentations/10-q-2q26.pdf ). First: equipment suppliers and funded power providers secure orders. Second: construction/funding failures strand capacity and delay customer compute, while advances shift some financing burden to customers. Third: new3.2GWcapacity by2029 could compete rents down; demand may grow enough to absorb it—counterexample, not forecast. Premise2 OESX Sept8 repeat lighting order (https://investor.oriones.com/news-releases/news-release-details/orion-announces-additional-multimillion-dollar-data-center ) points to ancillary domestic suppliers; first retrofit sales, second lower customer energy/maintenance costs, third greater efficiency reduces ongoing power demand per workload, while AI growth could overwhelm savings. No dollar margin quantified. Premise3 LEU TENEX dependence means import disruption can remove cheap input before domestic capacity replaces it; first higher uranium/enrichment prices, second delivery penalties/workingcapital pressure on incumbent customers, third new western capacity eventually competes with scarcity rents. https://www.sec.gov/Archives/edgar/data/1065059/000162828026007117/leu-20251231.htm . Premise4 Maersk Sep8 Nordic fuel surcharge effectiveSep9 (https://www.maersk.com/news/articles/2026/09/08/emergency-inland-fuel-energy-surcharge-nordic-region ) ranges0–20%, excludesEVtrucks/rail currently. First carrier passes higher costs, second shippers may substitute exempt modes, third substitution/conservation and capacity response can erode premium; actual eligible loads/costs unknown. Oct8 ECSA PSS500/1000USD percontainer similarly grossquote not cash (https://www.maersk.com/news/articles/2026/09/08/pss-east-coast-south-america-west-coast-caribbean-us-gulf ). Valid causal possibilities but soundness of positive equity payoff is unproven. ALM term21years vs15 gives40%moretotal tonnage over40%moretime, annual210kMTUunchanged; do not infer annualcapacityjump. No automatic order from any chain.

Premises:1. SEI Sep8 raised H2 EBITDA midpoint207.5m to282.5m (+36.1446%); guidance includes acquisitions, not solely organic scarcity. https://ir.solaris-energy.com/news/2026/09-08-2026-114509510 .2. Updated Q2 filing supersedes FY25 supplier concentration: Q2 suppliers24/16%, H1 27/13%, versus FY25 one51%; customer concentration instead rose to63%Q2 revenue and68%AR. H1 OCF265.506m less PPE835.2m=-569.694m simpleFCF; deferred-revenue inflow181.320m materially supports OCF; subtraction gives84.186m diagnostic, not forecast. Cash824.101m/debt2491.709m June30. https://ir.solaris-energy.com/~/media/Files/S/Solaris-IR/reports-and-presentations/10-q-2q26.pdf .3. SEI deck950MW current vs3.2GW2029 and2.2bn remainingcapex imply financing/construction exposure, not immediate earnings. Rough63.96*73.549m economicshareproxy=4.704bn equity plus1.668bn booknetdebt=6.372bn EV proxy, NOT fairvalue: mixed vintages, acquisitions, NCI, leases, TRA, converts omitted.4. LEU overhalf deliveries through2027 depend TENEX; US waiver does not guarantee Russian export permission; low historical inputcost aided margins. https://www.sec.gov/Archives/edgar/data/1065059/000162828026007117/leu-20251231.htm .5. CCJ unchanged19.5–21.5mlb own-outputguide despite outages; Q2 uranium market-related revenue76.903%,fuel7.581%, so uranium price does not uniformly reprice earnings. https://www.cameco.com/sites/default/files/documents/2026-Q2-MDA-FS-Notes_0.pdf .6. HUT future949MW/26.6bnbasecontracts/7.5bn financing are not current NOI or BTC flows. https://www.hut8.com/news-insights/press-releases/hut-8-reports-second-quarter-2026-results .7. OESX Sep8 repeat data-center lighting order has no disclosed exact value/margin; Q1 OCF1.352m/PPE.006m provides cash baseline only. https://investor.oriones.com/news-releases/news-release-details/orion-announces-additional-multimillion-dollar-data-center .8. ALM July14 GTP term15→21years and total3.15→4.41mMTU both imply210000MTU/year unchanged; price+6.3% distinct. https://www.sec.gov/Archives/edgar/data/1670061/000149315226033123/ex99-1.htm .9. Prysmian Q2 transmissionmargin21.2%vs17.1% and digital23.8%vs16.8% are realized improvement, but LTMFCF978mvs979m flat; margin basis is standardmetalprices. https://www.prysmian.com/en/media/press-releases/prysmian-raises-guidance-after-its-best-quarter-yet . CW and LIN forward adjusted PE37.55/26.31 use issuer midpointguidance, not GAAP valuation; both whole-share prices exceed212.75cap (recorded observations). Validity: capacityconstraints can confer pricing only with paiddeliveries beyond costs/financing. Soundness incomplete: contracts/customer acceptance, realized returns and fresh executable data remain gaps. Counterexamples: prepayments can finance SEI build economically; termextensions secure ALM longduration cash; they do not prove a ten-session equity edge. Prior upgraded on SEI/OESX operating evidence, trading allocation remains0 given rejected/uncertain holdouts.