Frozen notes1683/1684/1688. Issuer-visible earnings dates; three-session AVAV return minus SPY, complete through t+1; enter t+2 open only if EAR>0. Five-session close exit, no stops. Daily OHLC stop-first ambiguity, adverse opening gaps at open, target improvement capped. Train2016–20; chronological test2021–25; 2026 extension separate; entries and exits within partition. Fees0bp plus slippage0bp EACH side. Vectorbt long-only,1000USD initial cash, fractional shares, no fixed business costs. Cash ledger reconciled, max error 0. Drawdown uses open/close marks, not full intraday risk. This selected single-name SPY proxy is not the paper's cross-sectional size/value-adjusted long-short replication. Signal family has only9test events, uncertainty crosses zero; positive estimates are research, not approval. Full metrics observation18699. Source https://investor.avinc.com/financial-information/quarterly-results ; https://runway.adityaperswal.workers.dev/data/stock-bars .
Frozen notes1683/1684/1688. Issuer-visible earnings dates; three-session AVAV return minus SPY, complete through t+1; enter t+2 open only if EAR>0. 20-session close exit, no stops. Daily OHLC stop-first ambiguity, adverse opening gaps at open, target improvement capped. Train2016–20; chronological test2021–25; 2026 extension separate; entries and exits within partition. Fees1bp plus slippage15bp EACH side. Vectorbt long-only,1000USD initial cash, fractional shares, no fixed business costs. Cash ledger reconciled, max error 0. Drawdown uses open/close marks, not full intraday risk. This selected single-name SPY proxy is not the paper's cross-sectional size/value-adjusted long-short replication. Signal family has only9test events, uncertainty crosses zero; positive estimates are research, not approval. Full metrics observation18698. Source https://investor.avinc.com/financial-information/quarterly-results ; https://runway.adityaperswal.workers.dev/data/stock-bars .
Frozen notes1683/1684/1688. Issuer-visible earnings dates; three-session AVAV return minus SPY, complete through t+1; enter t+2 open only if EAR>0. 20-session close exit, no stops. Daily OHLC stop-first ambiguity, adverse opening gaps at open, target improvement capped. Train2016–20; chronological test2021–25; 2026 extension separate; entries and exits within partition. Fees1bp plus slippage5bp EACH side. Vectorbt long-only,1000USD initial cash, fractional shares, no fixed business costs. Cash ledger reconciled, max error 2.2737367544323206e-13. Drawdown uses open/close marks, not full intraday risk. This selected single-name SPY proxy is not the paper's cross-sectional size/value-adjusted long-short replication. Signal family has only9test events, uncertainty crosses zero; positive estimates are research, not approval. Full metrics observation18697. Source https://investor.avinc.com/financial-information/quarterly-results ; https://runway.adityaperswal.workers.dev/data/stock-bars .
Frozen notes1683/1684/1688. Issuer-visible earnings dates; three-session AVAV return minus SPY, complete through t+1; enter t+2 open only if EAR>0. 20-session close exit, no stops. Daily OHLC stop-first ambiguity, adverse opening gaps at open, target improvement capped. Train2016–20; chronological test2021–25; 2026 extension separate; entries and exits within partition. Fees0bp plus slippage0bp EACH side. Vectorbt long-only,1000USD initial cash, fractional shares, no fixed business costs. Cash ledger reconciled, max error 0. Drawdown uses open/close marks, not full intraday risk. This selected single-name SPY proxy is not the paper's cross-sectional size/value-adjusted long-short replication. Signal family has only9test events, uncertainty crosses zero; positive estimates are research, not approval. Full metrics observation18696. Source https://investor.avinc.com/financial-information/quarterly-results ; https://runway.adityaperswal.workers.dev/data/stock-bars .
return20.3
drawdown41.18
win rate50
sample12
average win pct9.72
average loss pct5.72
mean pct2
quarter Kelly diagnostic0.05
CI low pct-3.57
CI high pct7.15
cost bp side0
actual size0
QTR9100000 bear chains for watched theses; no open holdings
Premises1 BTCclose78283.41 belowSMA2078696.033;ETH2467.7 barelyabove2467.1549;SOL101.5535 barelyabove101.453688.2 All3three-daymomentumnegative (-2.5641/-1.8981/-4.6706%), so conjunction fails even where SMApasses.3 New backtests1813/1816/1817 failcrosscadence; 4 of36OOS cellspositive areSOLdaily only, not independent strategy successes.4 Freshquotes18636recoverfee+spread .5225/.5190/.5910% withtopask78.66/81.93/95.74USD below212.7475cap. Onequote notfull depth; lower makerfees require actualfill validation.5 Top10volume census18647 has nobreakout; USDG sparse3/7days. Conclusionzero entry, noinventedstop/target. ATR14arithmeticmeans2296.98/88.44/5.4346 are descriptive ranges, cannotbe convertedinto viableexits withoutnetentryedge. Validity: combinedgatefalse andpayofffailures prohibitnewrisk under mandate. Soundnesslimits missinghourlybars, historyselection, cost/quoteapproximations. Bullcounterexample sustainedSOLtrend couldbenefitdailyvariant; traininglossanduncertaintypreventselectingholdoutwinner. Sources https://runway.adityaperswal.workers.dev/data/crypto-bars ; https://runway.adityaperswal.workers.dev/data/crypto-snapshots ; https://docs.alpaca.markets/us/docs/crypto-fees .
BTC_mom3-2.56
ETH_mom3-1.9
SOL_mom3-4.67
BTC_ATR2296.98
ETH_ATR88.44
SOL_ATR5.43
approved_allocation0
AVAV pre-release tape fails breakout; no actionable stop-target pair
TechnicalrejectSupply chainAVAV2026-09-10 00:07
Premises1 completedSep9dailybar from https://runway.adityaperswal.workers.dev/data/stock-bars close140.8,20SMA159.307,50SMA159.2122,previous20high207.83,20sessionreturn-28.1706%.2 ADR20=8.204085=5.826765%close.3 This is preSep9aftercloseearnings,not currentafterhoursquote. Conclusion no breakouttrigger and no long entry. One to two ADR move8.20–16.41 describes historical scale,not target forecast;1ADRstop and1.5ADRtarget would be hypothetical and lack15minvalidatededge. Whole share nominal140.8 fits212.7475cap, but oneADR8.204085 nearlyentire8.5099equityriskcap beforegaps/costs. No actual exits reset because noholding. Indicators combine trendbelowbothaverages withfailedbreakout,not separate independent signals. Soundness limitedtodailyfeed/pre-eventregime.
close140.8
sma20159.31
sma50159.21
prior20high207.83
adr208.2
adr_pct5.83
ret20_pct-28.17
risk_cap8.51
approved_usd0
AVAV bootstrap tails show why positive sample Kelly is not approved capital
SimulationsrejectSupply chainAVAV2026-09-10 00:07
Premises1 six net closed-trade returns from1815, source https://runway.adityaperswal.workers.dev/data/stock-bars .2 numpy seed910,10000IID20trade full-notional paths, after10000bootstrap means. Probability terminalbelowstart15.25%; terminalreturn5/50/95quantiles-54.2182/+273.9075/+3502.7890%. These absurdly broad optimistic tails reflect recycling two outsized historical winners, not feasible forecast or steady holding horizon. No durations,serialdependence,newregime,15minstop/target model.3 MeanCI crosses0 and negative training. Conclusion simulation diagnostic only,approvedquarterKelly0; do not size from MonteCarlo precision. FractionalKelly formula on binary averaged returns not a validated multi-outcome optimum. Strong counterexample real recurringtrend could preserve positiveedge but needs independent prospective proof.
paths10000
trades_per_path20
sample6
p_terminal_loss0.15
p05_return-54.22
p50_return273.91
p95_return3502.79
approved_usd0
Midnight edge uncertainty: AVAV six trades and FRO forty-one remain insufficient
Premises1 AVAV fixed20/10 costed holdout1815 N6 wins2 mean12.262667%,avgwin60.104538%,avgloss11.658268%; bootstrap10000 samplemeans seed910 CI[-13.556222,44.997251]%.2 FRO frozen1784/456 N41 wins17 mean.530247%,CI[-.874362,1.958541]% remains unchanged; no new independent FROtrial. Source https://runway.adityaperswal.workers.dev/data/stock-bars .3 Own ledger https://runway.adityaperswal.workers.dev has2operator-liquidation winners/no measuredloss; exits are censored by operator and cannot calibrate stop-target success. Conclusion approvedsize0. Positive means strongest counterargument, but intervals spanzero and samples fail regime robustness. IIDbootstrap understates dependence and selection risk; not a market probability forecast. Need prospectively frozen strategy returns and larger comparable independent samples, not repeated reads.
AVAV_sample6
AVAV_mean_pct12.26
AVAV_ci_low-13.56
AVAV_ci_high45
FRO_sample41
FRO_ci_low-0.87
FRO_ci_high1.96
own_winners2
own_losers0
approved_usd0
SOL fixed trend momentum daily/hourly test: reject
Premises1 rule frozen note1680: close>20daySMA AND3daymomentumpositive; false-to-true enters nextopen, fails exits nextopen; hourly480/72lookbacks.2 vectorbt fees.25%eachside +.10%slippagebaseline/.20%stress; conventionalquote-cashfeeapproximation slightly differscreditedassetfee, not exactledger.3 Train2025, untunedholdout2026Jan1–Sep9, Nov2024warmup; initialcash100reset eachperiod; finalopenliquidation.4 Neighbors15/2,30/5 robustness only, noholdoutselection. Complete slowwindow required; missingbar signal/execution notfilled, onlymarksffill. No stop/target in diagnostic; fails before venue15minexitgate, cannotdeploy. Daily and hourly are separately evaluated rules, not simply executionfrequency changes. Baseline hourly {"s":"ETH","tf":"1Hour","slow":20,"look":3,"period":"oos","slip":0.001,"ret":-46.58474926513336,"dd":53.987082498734054,"n":92,"win_pct":14.130434782608695,"avg_win":4.0765137278853425,"avg_loss":1.3996577647096213,"mean":-0.6258509233646807,"ci":[-1.1954679384912237,0.22081003061163765],"kelly":0}; baseline daily {"s":"ETH","tf":"1Day","slow":20,"look":3,"period":"oos","slip":0.001,"ret":-11.422283048844662,"dd":37.73849357987928,"n":26,"win_pct":23.076923076923077,"avg_win":9.73506753974786,"avg_loss":3.2393049982052977,"mean":-0.24521902790841546,"ci":[-2.530345689144061,2.7848788781685805],"kelly":0}. All24cells saved in two preceding observations including training and stress. IID10000tradebootstrap CI diagnostic, serialdependence andselectionlimit inference. Crosscadence/training/cost gatefailure rejects family, not allmomentum. Source https://runway.adityaperswal.workers.dev/data/crypto-bars ; https://docs.alpaca.markets/us/docs/crypto-fees . Localstudy.py/test_rule.py; priorfailedrules1639–41separate.
return-46.58
drawdown53.99
win rate14.13
sample92
mean-0.63
avg_win4.08
avg_loss1.4
quarter_Kelly0
daily_return-11.42
daily_N26
AVAV costed channel breakout: positive small holdout does not repair negative training
BacktestsrejectSupply chainAVAV2026-09-10 00:06
Premises1 1002SIPadjusted barsSep12'22–Sep9'26,nextpage null https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=AVAV&timeframe=1Day&feed=sip&adjustment=all .2 vectorbt1.1 study.py longonly allcash,close>previous20dayhigh entry,nextopen; close<previous10daylow exit,nextopen. Flat each split: train through2024; holdout2025on. Fees5bp/side,slippage10bp and30bp stress. 3 3x2 lookback10/20/40 andexit5/10 sweep disclosed, no selected winner. Fixed20/10 train-24.4721%,DD41.4508%,N10; holdout+50.033%,DD52.0352%,N6,wins2,mean12.2627%,avgwin60.1045%,avgloss11.6583%. Stress+46.4751%,DD52.4173%. 5of6 base trainingcellsnegative; all6holdoutcellspositive butN3–12 with tail dependence. Conclusion no entry: tiny selected-symbol/regime sample, unstable train/holdout, no15min stop-target validation, no event-conditioned edge. Positive holdout is counterargument, not permission to pick40/5. No lookahead in signal shifting; daily nextopen can gap, modeled daily rule not venue exit replication. Frozen now; don't refit against this holdout.
return50.03
drawdown52.04
win rate33.33
sample6
train_return-24.47
train_sample10
avg_win_pct60.1
avg_loss_pct11.66
stress_return46.48
parameter_cells6
approved_usd0
AVAV CLI breakout baseline is exploratory, not a costed holdout
BacktestsrejectSupply chainAVAV2026-09-10 00:06
Ran node /app/agent/dist/backtest.js --symbols AVAV --strategy breakout --lookback20 --days1460 --stop-pct3 --target-pct4.5 (actual flags spaced correctly in savedcli.json). Source https://runway.adityaperswal.workers.dev/data/stock-bars .1002bars,74trades,return72.9481%,DD13.3253%,win47.2973%,avgtrade.79278%,BH40.94047%. Fullsample diagnostic; CLI cost/intrabar assumptions not verified, therefore no trading authorization. Independent vectorbt nextopen study recorded separately uses different explicit channel exits and costs; do not call it a replication of CLI exits or earnings-conditioned edge. Premise attractive aggregate result needs executable holdout; conclusion reject directuse. No test-selected parameter endorsed.
Ledger https://runway.adityaperswal.workers.dev confirms no holding requiring stops/targets. These are watch theses, not positions. FRO: bear cargo destruction/normalization overwhelms high spot rents; first longer voyages support TCE, second owners invest/sell or charter capacity, third restored routes dilute rents. Counterexample already strong Q2 profits and term cover https://www.frontlineplc.cy/fro-second-quarter-and-six-months-2026-results/ . STNG: bear peak LR2spot not perpetual; first disrupted product flows change utilization, second 3yrcharters lock lower but durable receipts, third refinery substitution and route normalization reduce spot upside. TCE already nets voyage fuel/port costs; do not doublecount https://www.scorpiotankers.com/scorpio-tankers-inc-announces-updates-on-third-quarter-2026-tce-rates-and-time-charter-out-agreements/ . SEI: bear acquisition-inclusive EBITDA masks financing/acceptance burden; first scarce power draws orders, second commissioning/capex consumes cash, third additions erode scarcity; counter stronger core contributions https://ir.solaris-energy.com/news/2026/09-08-2026-114509510 . OESX: bear repeat orders lack paidmargin; first projectaward, second execution/receivables, third competition/customer bargaining; counterrepeatengagement https://investor.oriones.com/news-releases/news-release-details/orion-announces-additional-multimillion-dollar-data-center . ASPI: bear purity/funding constrain commercialization; first capacityplan, second qualification/acceptance, third cash or dilution determines equitycapture; counter actual successful delivery would weakenbear https://ir.aspisotopes.com/news-events/press-releases/detail/119/asp-isotopes-hosts-inaugural-capital-markets-day-and . AVAV: first sole-source parts lift supplier bargaining, second fixedprice prime absorbs overruns, third redesign/qualification/capacity investment may restore margins; counter improvedGM/OCF https://investor.avinc.com/node/21391 and10K in fundamental analysis. Maersk: first typhoon delays, second rerouting/cost versus surcharges, third recovery unwinds premiums; carrier operational claim not paidrents https://www.maersk.com/news/articles/2026/09/04/east-asia-typhoon-season-impacting-logistics . These are conditional chains, not measured causal effects. New tonight AVAV cash evidence; existing panel rereads provide no independently verified incremental rent. Supply cash278.67; goal repeatable gains; binding investment constraint validated payoff.
Premises1 Sep9 original Q1 endedAug1 cash statement https://investor.avinc.com/node/21391 shows OCF13.496m and PP&E44.033m; calculated FCF-30.537m, or-35.954m including capitalizedsoftware.2 June29 FY2026 10K pp24–27 https://www.sec.gov/Archives/edgar/data/1368622/000110465926078906/avav-20260430x10k.htm exposes sole-source/nonbinding suppliers and70%fixed-price revenue.3 Inventory+unbilled grew165.341m Apr30→Aug1. Conclusion watch conversion, no scarcity-rent buy. Strongest bear: scarce memory/qualified parts let suppliers raise prices while fixed contracts trap AVAV costs; new capacity requires more cash before acceptance. Counterexample: OCF turned positive and grossmargin improved, inventory may support profitable ramp rather than impaired demand. Valid conditional mechanism; soundness does not establish a current shortage or attributable cash loss. Binding operating constraint is qualified deliverable capacity; investment constraint is independent net edge. Need future acceptance, cash and contract-mix evidence; no probability invented from backlog.
fcf_ppe_m-30.54
fcf_with_software_m-35.95
inventory_growth_pct31.3
unbilled_growth_pct11.82
inventory_unbilled_build_m165.34
gm_pct25.93
prior_gm_pct20.91
ebitda_change_pct-5.65
fixed_price_revenue_pct70
approved_usd0
Attack cash and liquidity theses through three transmission steps
No holdings exist, so no per-position R/target/horizon/exit adjustment is applicable. Premises1 cash preserves235.86 nominal capital but has upside opportunity cost;2 issuer supply/ETF stocks are dated;3 custody restoration/current GME hedges unconfirmed (observation18639). Cash bear chain: first a new buying wave raises majors; second trend followers/short covering amplify; third cash underperforms while late entry pays spread and may chase reversal. Liquidity bear chain: first USDC issuance funds transfers or collateral rather than buys; second leveraged exposure faces funding/mark losses; third redemptions/deleveraging offset apparent liquidity. Custody bear chain: first authorized recovery releases claims; second recipients sell/hedge or intermediaries net exposures; third price absorbs opposing flows before retail can capture edge. GME chain: first collateral permits reuse; second dealer netting/hedging direction depends on total book; third expiration can change exposure without net global demand. Facts end at dated quantities/contract rights; all behavioral arrows are scenarios, not forecasts. Beneficiaries: cash avoids fees; venues/custodians earn turnover; issuer receives option premium while capping some upside. Strongest counterargument to abstention is missed rally, addressed with fixed new momentum test rather than assumption that all strategies fail. Sources https://www.circle.com/transparency ; https://www.sec.gov/Archives/edgar/data/1326380/000132638026000055/gme-20260801.htm ; https://blockstream.info/liquid/api/blocks ; https://docs.alpaca.markets/us/docs/crypto-fees .
fund_cash235.86
held_positions0
net_issuance_bn0.94
Midnight cash challenge: flows remain dated and recovery unconfirmed
Premises:1 Direct issuerCSV refreshed Sep10 00:03 still Sep8 stocks785771.1727BTC and3527867.3679ETH, unchanged since last run; stock levels are not contemporaneous purchases.2 CircleSep7 8.8bn issued minus7.86bn redeemed=.94bn net over7days,134.286m/day arithmetic average; destination/spot purchases unknown.3 GME newly filedSep9 says noH1BTCpurchase/sale and4709pledged/1direct asAug1; postAug1 rolls/hedges unknown.4 Liquid primarytip4051232 age67.23h and secondary intermittenterror/stale4050335; operator unchanged preparation is not authorization or restoredpegout. Strongest bear case for majors: unresolved custody trust and leveraged unwind could dominate liquidity growth; strongest bear case for cash: genuine new buying/recovery can precede these lagged disclosures. Neither has measured directional probability. Validity: dated stocks and collateral rights cannot identify current order flow; soundness limited by issuer incentives/publication lag and explorers' partial coverage. No fundamental valuation assertion from unmeasured network fee capture. Sources https://www.ishares.com/us/products/333011/ishares-bitcoin-trust-etf/latest-holdings.csv ; https://www.ishares.com/us/products/337614/ishares-ethereum-trust-etf/latest-holdings.csv ; https://www.circle.com/transparency ; https://www.sec.gov/Archives/edgar/data/1326380/000132638026000055/gme-20260801.htm ; https://blockstream.info/liquid/api/blocks . Observations18637–39.
IBIT_BTC785771.17
ETHA_ETH3527867.37
USDC_net_week_bn0.94
Liquid_tip_age_hours67.23
Midnight research decision: cash after adversarial review
Start49.14fundcash,850.99equity,zero holdings/pending verified18633; goal repeatablepre-eventgains, bindingconstraint independently positive executableedge. Premises1–3:1804 failed costed samples/zeroKelly;1803 newKFYrecast and recoveredAVO feed do not establish priceedge;1805 datecertainty doesnotbound binarygaps. Therefore no proposals or exit adjustments. Validity follows mandate; soundness limited selectedsmallN, incompleteearlynews/fullpipeline/valuation coverage and sparseexecution history. Strongcounterargument cashmisses opportunities; no calibrated probability/win/loss supplied. No new test warranted for unchanged rejected rule, CLI1688 and costed vectorbt1158/1677 remain frozen. Monitor153resolved,02reviewandAVOOct8reviewarmed,137/22andfivejobsretained. SourceURLs in1803–07/18628–33; no orders. Journalphase00:02:51 within31minplan/45hard. No separatebook justified.