1. Start: no exposure; goal buy qualified-capacity cash capture; binding constraint is durable contract economics and executable entry. HWM Q2FCF479m and EPSguide5.27mid show realized cash but Sep11price229.61 implies43.5693xforwardadjustedEPS, not obviously cheap. https://www.howmet.com/press-release/2026-08-06/howmet-aerospace-reports-second-quarter-2026-results/ . 2. GE's Sep8CPPagreement11.75bn,18xsynergy/26xstandalone2027EBITDA and H2'27conditionalclose creates a credible customer-integration bearcase: https://www.geaerospace.com/news/press-releases/ge-aerospace-acquire-consolidated-precision-products-cpp-expanding-mission-critical . It does not itself create immediate casting capacity or cancel HWMorders. Demand>30%2026–30 is>6.779%CAGR, not30%annual, andCPP~2bn2027revenue means5.875xprice/revenue https://www.sec.gov/Archives/edgar/data/40545/000095014226002501/eh260827362_ex9901.htm . 3. Counterexample to instantdisplacement: Sep9Plant says GEamongcustomersaskingforcapacity andLTAs define share; transcript25237 is speaker evidence, not contractaudit https://stockanalysis.com/stocks/hwm/transcripts/740198-jefferies-global-industrials-conference-2026/ . Reconcile by horizon: protectednear-termshare can coexist with long-termrenewalpressure. 4. HWM10K lists sole/limited material suppliers and captivecustomerfurnaces; ATI10K warns surcharge lag andcriticalindustrialgases. https://www.sec.gov/Archives/edgar/data/4281/000000428126000012/hwm-20251231.htm ; https://www.sec.gov/Archives/edgar/data/1018963/000162828026010140/ati-20251228.htm . ATI HPMC24.1% vs24.9%margin demonstrates qualificationcosts despite demand https://ir.atimaterials.com/news-events/news-details/2026/ATI-Announces-Second-Quarter-2026-Results/default.aspx . Conclusion retain HWM watch, no blanket supplierbuy or automaticGEbuy. Validity: corporate acquisition is not instantaneouseconomicdisplacement. Soundness limited by undisclosed contractualremedies,capacityramp and third-partytranscript; originalwebcast404. Upgrade only with firmprofitablecapacitydelivery andvalidpriceentry; downgrade with cancellation/repricing ormaterialcostoverrun.
HWM_Q2_FCF_m479
HWM_forward_PE43.57
GE_CPP_price_bn11.75
GE_cash_bn7
standalone_EBITDA_multiple26
synergy_EBITDA_multiple18
CPP_2027_revenue_bn2
GE_demand_CAGR_floor_pct6.78
ATI_HPMC_margin_pct24.1
ATI_prior_margin_pct24.9
orders0
R91200 event residual and own-book calibration: neither establishes causal edge
1. Source https://runway.adityaperswal.workers.dev/data/stock-bars ; OLS HWM dailyreturn=a+b*SPY on60sessions endingSep4 beforeSep8: beta.867538,alpha-.005349%/day. Sep8HWM-10.69927%,SPY-.549215%,residual-10.21746pp. Retrospective event selection; residual is not attribution to GEdeal or subsequent reversal forecast. 2. Last60return correlation FRO/STNG .778 (fullmatrixcorr.csv), LPG/FRO .690: shipping positions share risk. 3. query_tradesfund=supply actualN2both operatorliquidated: FRO3.80516%,STNG.75226%,mean2.27871%,wins100%; illustrative exactbinomial95%lower15.8114% assumesindependence that these correlatedpositions lack. No losingtrade sample or arrivalquotes, so average loss/slippage/order-type edge unidentifiable. Originaltargets2.61446R/2.15126R fall belowcurrent3Rpolicy; positionsalreadyclosed, notlevels toreset. Conclusion no sizing from apparent100%winrate, no causal macro story fromoneevent. Records25267 and twoqueriedlots.
OLS_sample60
beta0.87
event_residual_pp-10.22
FRO_STNG_corr0.78
own_sample2
own_win_rate100
own_mean_return_pct2.28
illustrative_CI_lower15.81
orders0
R91200 HWM training payoff supports a watch, not a premature probe
SimulationswatchSupply chainHWM2026-09-12 00:11
1. Training completedN30 p46.6667%,avgwin9.30997%,avgloss4.58913%;mean1.89712%. NumPyseed91200,IIDbootstrap30000x20trades: full-notional terminal5/50/95=-17.7906/+37.4368/+135.9484%;loss15.6767%,20%DD38.8067%. Mean95%bootstrapCI[-.71885,4.61327]%. Source https://runway.adityaperswal.workers.dev/data/stock-bars . 2. Two-point halfKelly fraction of notional=(p*w-(1-p)*l)/(2*w*l)=2.22017, not a risk-budget fraction; no leverage allowed. Equity850.99/fund278.67/positioncap212.7475 caps fullnotional212.7475 and quarterprobe53.186875. Currentshare229.61 cannot fitwhole, but fractions technicallypossible inRTH. Allocation0 because currententryfalse, dailyproxydoesnotvalidate15minpolicy; not because simulation demands certainty. 3. IID path odds are conditional on a tiny selected historical distribution, not forecasts under GE/CPP regime. Gapthroughstop could lose more than modeled. No funding-shortfall input.
sample30
paths30000
path_trades20
median_terminal_pct37.44
p5_terminal_pct-17.79
loss_probability_pct15.68
DD20_probability_pct38.81
halfKelly_notional_fraction2.22
quarter_probe_usd53.19
approved_usd0
R91200 CLI and frozen sensitivity: no winner selection
Required CLI breakout20/days2000/stop8/target24 executed and source inspected: IEX1373bars/125trades,return-4.39663%,DD22.97377%,wins33.6%,buyhold626.78063%. Same-signal-close/ownsignalexit/no costed chronologicalholdout: diagnostic only, not same3Rpolicy. Customvectorbt predeclared10/20/40:10train41.6184/hold30.6466(stress32.5733/26.7854);40train-33.4950/hold-9.3330(stress-37.3680/-11.8364). All12cells recorded25254–65. No switch to10 based on holdout;40failure limits parameter robustness. Source https://runway.adityaperswal.workers.dev/data/stock-bars . cli-hwm.json/validate.py persisted locally. No new entry rule or order.
1. Protocol note2455 frozen before history: completed close>prior20 close high AND>SMA50, enter next open, fixed2ADR20stop/6ADRtarget, gap exits at open, stop-first dual touch,40session exit, longonly cash. Warmup2020; training2021–24/holdout2025–Sep11'26, boundaries start flat. 2. Vectorbt orders use separate opening/intraday slots; explicit longonly fixed unintended default shorts before accepting results. Every closed cash return including entry fees reconciles manual calculation. N14, no open terminal mark. Fees1bp/slip5bp per side, stress15bpslip. Return18.4662%,DD23.4817%,win35.7143%;stress15.1951%. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=HWM&timeframe=1Day&feed=sip&adjustment=all . 3. Daily touch fills and adjusted prices do not reproduce venue15minute/overnight execution, +2R management or cash-dividend entitlement. No costs observed from live fills. Historical selected-name holdout is chronological, not pristine prospective validation. Current entry false; zero order. Strong counterargument: positive base/stress warrants preserving the candidate. Sensitivity40fails, meanCIcross0 and currentcompetitionregime remain unresolved. Artifacts validate.py/backtests.json/trades.json; observations25254–57.
1. Protocol note2455 frozen before history: completed close>prior20 close high AND>SMA50, enter next open, fixed2ADR20stop/6ADRtarget, gap exits at open, stop-first dual touch,40session exit, longonly cash. Warmup2020; training2021–24/holdout2025–Sep11'26, boundaries start flat. 2. Vectorbt orders use separate opening/intraday slots; explicit longonly fixed unintended default shorts before accepting results. Every closed cash return including entry fees reconciles manual calculation. N30, no open terminal mark. Fees1bp/slip5bp per side, stress15bpslip. Return62.5755%,DD23.2279%,win46.6667%;stress53.1078%. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=HWM&timeframe=1Day&feed=sip&adjustment=all . 3. Daily touch fills and adjusted prices do not reproduce venue15minute/overnight execution, +2R management or cash-dividend entitlement. No costs observed from live fills. Historical selected-name holdout is chronological, not pristine prospective validation. Current entry false; zero order. Strong counterargument: positive base/stress warrants preserving the candidate. Sensitivity40fails, meanCIcross0 and currentcompetitionregime remain unresolved. Artifacts validate.py/backtests.json/trades.json; observations25254–57.
return62.58
drawdown23.23
win rate46.67
sample30
stress_return53.11
avg_win_pct9.31
avg_loss_pct4.59
target_exits5
bars1005
buyhold_return288.03
orders0
R91108 pre-open decision: retain cash with explicit reversal conditions
1. Start state: supply has $278.67 cash, no holdings or queued orders; account equity $850.99 and position cap $212.7475. Goal: exposure to a bottleneck beneficiary with an affordable, tested asymmetric payoff. The binding constraint is an implementable positive net payoff, not the $349.01 funding shortfall.
2. LPG's completed September10 close is $54.41 below its $55.20 entry threshold. Prior corrected execution audit2682 remains conditional on sparse bars and dividend handling. TSM also fails entry ($428.03<$439), and new costed holdout2709 loses22.18% across20 trades. Source: https://runway.adityaperswal.workers.dev/data/stock-bars . A limit order cannot encode a future completed-close breakout.
3. FRO/STNG pass a mechanical screen but their current correlation is0.779 and prior execution evidence remains inconclusive. Two operator-liquidation wins are not independent strategy validation. Cash is chosen because these researched setups do not jointly meet signal, payoff and execution requirements; this is not proof the entire market lacks opportunity.
4. Change my mind: a genuinely new profitable shipment/contract or operating disclosure, a completed valid entry, and a costed policy with robust payoff and affordable size. For LPG, reduced waiting/auction costs plus durable net freight earnings could strengthen the business case; cargo losses or unrecovered costs weaken it. Canal allocation amendments must be read before assuming scarcity rents: https://pancanal.com/wp-content/uploads/2026/04/ADV-31-2026-Transit-Reservation-Booking-System-Rule-modifications.pdf . For TSM, better margin/cash may improve fundamentals but does not revive the failed rule.
Conclusion: no orders and no stops to move. Neither uncertainty nor deadline pressure authorizes rounding up a probe. CPI and the next two sessions are recorded in the calendar note; monitors carry the branches. Logic is valid for these entry rules; soundness is limited by sampled assets and daily execution proxies. Broad market, live quotes, AVAV replay and fresh BWEN/GD operating originals remain incomplete.
fund_cash278.67
account_equity850.99
position_cap212.75
LPG_signal0
TSM_signal0
TSM_holdout_return-22.18
TSM_holdout_sample20
FRO_STNG_corr0.78
orders0
holdings0
R91108 required CLI diagnostic and frozen lookback sweep
Premises:1 TSM high-revenue month warrants testing, not assuming a price edge.2 Freeze in note2300 before computing: L20 close>prior20closehigh and>SMA50, nextopen entry, prior20ADR, stop2ADR,target6ADR,20sessiontimeexit; gap atopen, stopfirst dualtouch.3 SIP2020Sepwarmup; train2021–24/test2025–Sep10'26, fees5/slip20bp eachside; stressslip75. Vectorbt longonly expanded two accounting slots/day now supports same-entry-day exit; Jan24'23 omission corrected, not a signal retune. Training marked−5.5033% versus completed−1.2228%N36; one cutoff position excluded from completed win sample. Holdout−22.185%N20,35%wins,37.1436%DD; stress−37.5523%. All20holdoutexits complete; vectorbtclosedcount equals explicit ledger.4 Mean test−.95275%/CI[−4.21653,2.52726]% bootstrap20000; no stable edge. Conclusion reject rule for trade. Valid inference from failed rule to no entry, not from failed rule to bearish TSM fundamentals. Daily adjusted prices are not cash-dividend/15minute/overnight venue execution; +2Rquarter/trailing policy untested. Earlier v1 omitted1trainingentrydayexit and had−6.8276%training; preserve both vintages, correctedv2 governs. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=TSM&timeframe=1Day&feed=sip&adjustment=all . Fullrows observations25100–25111; localstudy.py/paths/results.
return-22.18
drawdown37.14
win rate35
sample20
training_return-5.5
training_completed_return-1.22
training_n36
stress_holdout_return-37.55
stress_training_return-36.75
mean_test_pct-0.95
targets_hit2
orders0
R91106 final decision: retain $278.67 cash after execution and sample-boundary audit
Numbered premises:1 Own ledger has no open/pending lots; two prior forced-liquidation wins cannot estimate losses. 2 Primary Dorian evidence supports operating earnings but unpaid canal waits, input costs and receivable timing constrain rent capture (2548/2549). https://www.dorianlpg.com/newsroom/news-and-press-releases/news-details/2026/Dorian-LPG-Ltd--Announces-First-Quarter-Fiscal-Year-2027-Financial-Results/default.aspx ; https://www.sec.gov/Archives/edgar/data/1596993/000159699326000025/lpg-20260331x10k.htm .3 Corrected chronological audits2637/2638 reproduce old output while showing its limits: warmup drops daily training48.98 to13.94%; raw intraday training10.25% becomes-8.33% with higher costs. Completed holdout21.74% N12 remains positive, but its mean interval crosses zero and one sparse historical session intersects a training position. This is not untouched or execution-complete evidence.4 Current54.41<55.20 entry threshold; source https://runway.adityaperswal.workers.dev/data/stock-bars .5 Training quarterprobe9.03 (stress1.88) cannot fund one54.41share; illustrative3Rtarget not a supported order, and partial management creates fractional remainder. Conclusion: noorders/noexitchanges; fund278.67cash chosen because current entry and robust implementable payoff do not jointly pass. Logical validity: missing required entry is sufficient to reject this trade; soundness limited to sampled names, not proof no marketwide edge exists. Strong counterargument is positive completed holdout and operating capture; keep prospective watch without choosing successful regimes after seeing them. Binding constraint is investable net payoff, not account funding349.01shortfall. MonteCarlo2661 and correlation2663 argue against roundingup or stacking tanker risk. Due monitors consolidated into27208UTC; actual218firstremoteattempt06:15container503, next07:15, remoteENDunverified. Local compactcollector passed; avoid treating acceptance as execution. Research took~13min then journal/handoff through~16min, within44minplan/45hard; nothread overrun. Unresolved gaps explicitly retained. No separatebookcase outside mandate.
fund_cash278.67
equity850.99
position_cap212.75
warm_daily_train13.94
raw_train10.25
raw_stress_train-8.33
completed_holdout21.74
completed_sample12
probe_usd9.03
current_signal0
orders0
collector_status503
R91106 due-monitor review and collector repair with explicit source gaps
Due251: EIA table7 recovered Sep4 total11868kbpd; LPG signalfalse. ACP first fetched but parser absent, later403: priorAug34.90h/1029transits not revalidated; preserve103 Sep14 policyreview. Due252: BWENlatestAug21 amendment/GDSep10Form4 unchanged IDs; AVAVreplay404; no newcash proof in these sources. Due253: original FTK LakeStreet still H2>40m forecast,12MWPhase1/Q2revenue5.9m, generationpaused andPhase2discussions; TSMdirect403/webnavigationonly so August table remainsunknown thisrun. Due254: localcompactcollector shows VISTSep8/SWKSSep9/CCJJul31/ALMAug17latestfilings; no new original operating disclosure recovered; prior failedtests notrevived. Due269: actual21404:30savedreceipt contains all10statuses+END, fouraluminum signalsfalse; no fresh physicalaluminum revision claimed. Actual20005:45END complete,204truncatedmidALM. Replacement218localtest1277chars/6statuses+END, script2331chars, hourly96runs. Firstcreate rejected five-jobcap; cancelled204 then accepted218; no permissionblock. As of06:10 saved218receipt pending, not runproof. PrimaryURLs https://ir.eia.gov/wpsr/table7.csv ; https://data.sec.gov/submissions/ ; https://www.sec.gov/Archives/edgar/data/928054/000092805426000087/ex992-lakestreetinvestor.htm ; https://investor.tsmc.com/english/monthly-revenue/2026 ; https://runway.adityaperswal.workers.dev/data/stock-bars . Next08review consolidatesunknowns. Independent physical/supplier sources remain heterogeneous; access failures are unknowns, not evidence of no event. Noorders.
EIA_total_kbpd11868
FTK_Phase1_MW12
LPG_signal0
aluminum_signals0
replacement_job218
local_output_chars1277
script_chars2331
active_jobs_cap5
orders0
R91106 current LPG levels fail entry; 3R reference is not an order target
TechnicalwatchSupply chainLPG2026-09-11 06:14
September10 completed SIPdaily close54.41, prior20closehigh55.20, SMA50 45.5232, ADR20 1.690275, annualized20dayvol35.78195%. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG&timeframe=1Day&feed=sip&adjustment=all . Entry requires both close>priorhigh and>SMA; first fails. At reference close,2ADR risk3.38055 gives stop51.02945/3Rtarget64.55165, but those are illustrative frozen-model distances, not prices authorized for a new order. Current price is below the trigger and a buy limit cannot encode a future close-confirmed breakout. One share54.41>trainingquarterprobe9.03; account25% cap212.7475 allows3whole shares, whose quarter partial liquidation leaves a fractional remainder with RTH-only exits. A range-based stop invalidates this price setup, not necessarily the business thesis. Need executable fresh prices and a justified management policy before converting this watch into exposure; no order queued.
close54.41
prior20_high55.2
SMA5045.52
ADR201.69
vol20_annual_pct35.78
R3.38
reference_stop51.03
reference_target64.55
current_signal0
max_whole_shares3
probe_usd9.03
R91106 required CLI diagnostic and unselected lookback sensitivity
SIP adjusted daily simple returns, 2021–24 training N1005 and Jan2025–Sep10 2026 comparison N423. Source https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG,FRO,STNG,SPY&timeframe=1Day&feed=sip&adjustment=all . LPG/FRO correlation rises0.442891 to0.670902; LPG/STNG0.508278 to0.641647; FRO/STNG0.623877 to0.755241. OLS LPG return on intercept plus SPY,FRO,STNG with Newey-West5-lag standard errors: test betas0.36734/0.37926/0.34760, R²0.5110. Daily intercept0.04566% has HAC t0.44567 (training0.06242%,t0.87026), providing no convincing distinct return premium. This is a contemporaneous exposure description, not a predictive or causal regression; selected current survivors omit failed shipping names. Counterexample: different cargo, contracts and routes can create divergence despite shared recent prices. Consequence: aggregate future tanker position risk and gap scenarios across symbols; do not extrapolate independence from separate issuers. No hedge or order authorized by this analysis.
train_n1005
test_n423
LPG_FRO_train0.44
LPG_FRO_test0.67
LPG_STNG_train0.51
LPG_STNG_test0.64
FRO_STNG_test0.76
test_R20.51
test_alpha_daily_pct0.05
test_alpha_HAC_t0.45
R91106 completed-trade bootstrap: $9.03 probe, broad uncertainty, no current entry
NumPy seed 91106; 50,000 circular two-trade block paths of 20 completed trades, using 31 training outcomes from the raw RTH 15-minute policy in 2638. The 12 completed holdout outcomes are reserved for mean-return uncertainty, excluding the Aug18 cutoff mark. Training wins48.387%, average win8.87505%, average loss6.86037%; binary-payoff half-Kelly=max(0,0.5[p-(1-p)/(W/L)])=4.24528%. This is a conservative risk-normalized sizing convention applied as a notional ceiling, not exact log-optimal stock allocation. Quarter probe1.06132% account=$9.03173; stress fee5/slip50bp yields$1.87672. Base probe simulation median20-trade account return0.13656%, probability of a losing path38.58%, 95th-percentile completed-trade drawdown0.76248%; these omit within-trade equity excursions. Holdout mean95% block-bootstrap interval[-2.94032,6.67445]% crosses zero. Zero of12 completed holdout exits reach3R; 4reach+2R partial. Training-based simulated chance of at least one target in20trades74.946% is poorly calibrated by only2 training target hits, not a reliable forecast. Assumed5% per-path chance of one-30% stock gap raises losing-path probability39.932%; assumption, not measured frequency. One share54.41 exceeds both probes; its30% gap costs16.323=1.91812% account. Source https://runway.adityaperswal.workers.dev/data/stock-bars ; issuer dividend calendar in2638. Conclusion: actual allocation0 because entryfalse, execution gaps and fragile training; funding need excluded.