job: 22 supply panels plus LPG signal; replaces173 (exit 0)
Supply chain2026-09-11 07:00
fund=supply 2026-09-11T07:00:19.420301+00:00 research only; eligible=0; URL map in job script
Maersk OK c1c2759bea0f
Typhoon OK 4672d4314359
Hapag OK 8ab284858970
SeaIntel OK 0eeef616fcf3
WH OK 1b8ce73b697a
Bombardier CHALLENGE ff2ec65dacb6
GD 2026-08-07 8-K 0001193125-26-340490/d129075d8k.htm | 2026-08-04 8-K 0001193125-
CENTCOM ERROR HTTPError
FRO OK 3aaf7244dc4a
STNG OK c54b428d736e
ASPI Tue, 08 Sep 26 09:15:00 -0400 asp-isotopes-hosts-inaugural-capital-markets-da
ACP OK 111fd2450a7e adv-34-2026-monthly-canal-operations-summary-august-2026.pdf
SEI OK 140860f05cb5
OESX OK f7316d1affda
Canada OK 09084e6193e2 news.html
FTK OK b2455970b637 [('2026-09-09-Flotek-Raises-2026-Guidance-Again-on-Continued-S
FTKslides OK 32911d7c68cb [('Lake-Street-Investor-Presentation.pdf', 'Lake Street
AVAV OK bc31162c3404 [('f4f2cf5b-081f-4a73-848b-52666a7e9bcc', 'AeroVironment Q1 F
SSL OK cd1cb3d694c5
SSLnews OK e90322db91dd
ACSArelay OK b8abed2463e0
END 21/21 attempted; hash is not news; ACSA baseline Aug28 not live.
fund=supply LPG DAILY research only {"source": "https://runway.adityaperswal.workers.dev/data/stock-bars", "asof": "2026-09-11T07:00:25.771577+00:00", "bar": "2026-09-10T04:00:00Z", "close": 54.41, "prior20closehigh": 55.2, "sma50": 45.523199999999996, "adr20": 1.6902749999999997, "signal": false, "eligible": false, "note": "current UTC date excluded; no live execution validation or automatic order"}
END LPG panel attempted; missing data are not a false signal
job: Crypto quotes and Liquid to08UTC; replaces207 (exit 0)
Crypto majors2026-09-11 07:00
fund=crypto research only eligible=0 observed=2026-09-11T07:00:26.579092+00:00
src=https://runway.adityaperswal.workers.dev/data/crypto-snapshots ; https://blockstream.info/liquid/api/blocks
BTC b/a=77254.881/77270.058 age=15 ok=1 depthUSD=76.9/77.3 hurdle%=0.522
ETH b/a=2465.563/2466.2 age=144 ok=1 depthUSD=81.4/81.5 hurdle%=0.528
SOL b/a=99.7266/99.81 age=705 ok=0 depthUSD=95.3/94.9 hurdle%=0.586
USDC b/a=0.9995/1 age=8438 ok=0 depthUSD=9995000.0/45999.0 hurdle%=0.552
USDG b/a=0.99999/1.00001 age=97704 ok=0 depthUSD=9999900.0/10000100.0 hurdle%=0.504
XRP b/a=1.3499/1.3525618 age=85 ok=1 depthUSD=5586.8/7.4 hurdle%=0.700
UNI b/a=6.147/6.1596 age=51 ok=1 depthUSD=4328.1/4317.3 hurdle%=0.708
LTC b/a=52.82/53.1375 age=12 ok=1 depthUSD=3561.1/3632.4 hurdle%=1.106
HYPE b/a=79.794/80.2 age=62 ok=1 depthUSD=62047.0/62212.7 hurdle%=1.013
LINK b/a=11.5072/11.528 age=162 ok=1 depthUSD=13296.0/13390.6 hurdle%=0.684
fresh_quotes 7 /10
recipient BTC 598.50049115 mempool_net 0.00015661 tx_n 297
reserve BTC 3601.47203772 mempool_net 3.384e-05 tx_n 590
recentpage 59 controlled 0 not_full_history=1
tip 4052353 age_s 17 t 1789110010
second 4052353 age_s 17 t 1789110010
operator_notes 3
operator_hash 1e913157cc811e629b1b86d4db894412f924f07ea0928c7aa1a3e183681de0ca claim To those responsible for the theft of bitcoin from the Liquid Network:
Blockstream will not pay a r
END fund=crypto panels_attempted=7/7 eligible=0; balances not sales or recovery
fund=catalyst RESEARCH ONLY size=0 UTC=2026-09-11T07:00:21.107597+00:00
CNXC ['CNXC'] [{"filingDate": "2026-08-14", "form": "SCHEDULE 13G", "accessionNumber": "0000930413-26-002515", "primaryDocument": "xslSCHEDULE_13G_X02/primary_doc.xml"}, {"filingDate": "2026-08-13", "form": "SCHEDULE 13G/A", "accessionNumber": "0001167557-26-000183", "primaryDocument": "xslSCHEDULE_13G_X02/primary_doc.xml"}, {"filingDate": "2026-08-06", "form": "SCHEDULE 13G/A", "accessionNumber": "0000315066-26-001965", "primaryDocument": "xslSCHEDULE_13G_X02/primary_doc.xml"}]
TBNnews Tue, 08 Sep 26 08:00:00 -0400 Tamboran Delivers First Gas Sales From the Beetaloo Basin https://ir.tamboran.com/news-events/press-releases/detail/49/tamboran-delivers-first-gas-sales-from-the-beetaloo-basin
TBNnews Tue, 01 Sep 26 08:00:00 -0400 Tamboran Resources Corporation Announces Commencement of SPCF Commissioning Activities https://ir.tamboran.com/news-events/press-releases/detail/48/tamboran-resources-corporation-announces-commencement-of-spcf-commissioning-activities
TBNnews Mon, 20 Jul 26 07:00:00 -0400 Tamboran Successfully Completes Largest Beetaloo Basin Stimulation Program https://ir.tamboran.com/news-events/press-releases/detail/47/tamboran-successfully-completes-largest-beetaloo-basin-stimulation-program
TBN https://www.tamboran.com/corporate-calendar/ sha=e2dc8bbfdce213e0 US Australian Event February 11, 2026 (PM) February 12, 2026 (AM) FY26 Second Quarter Earnings May 13, 2026 (PM) May 14, 2026 (AM) FY26 Third Quarter Earnings September 25, 2026 (AM) September 25, 2026 (PM) FY26 Annual Earnings November 13, 2026 (AM) November 13, 2026 (AM) FY27 First Quarter Earnings November 2026 November 2026 Annual Meeting
NEOV 2026-09-09T15:47:09.043+00:00 https://neovolta.com/blog/home-battery-safety-what-every-homeowner-should-know sha=352ea8322de7b03c Home battery systems are becoming an increasingly important part of how homeowners manage their energy. They can store solar energy for later use, provide
job: PEAD financial-context discovery; replaces8 (exit 0)
Quant replication2026-09-11 07:00
{"utc":"2026-09-11T07:00:18.593610+00:00","seen":26,"pages":1,"more":false,"matches":2,"omitted":0,"items":[{"t":"2026-09-11T05:50:28Z","s":["RH"],"h":"RH Reports Q2 2026 Results: Full Earnings Call Transcript","u":"https://www.benzinga.com/news/26/09/61731550/rh-reports-q2-2026-results-full-earnings-call-transcript"},{"t":"2026-09-11T05:44:24Z","s":["REF"],"h":"Reformation Reports Q2 2026 Results: Full Earnings Call Transcript","u":"https://www.benzinga.com/news/26/09/61731490/reformation-reports-q2-2026-results-full-earnings-call-transcript"}],"coverage":"5h bounded news discovery; issuer verification required; no orders"}
VM9110600 final handoff: price audit precedes further strategy searches
Quant replication2026-09-11 06:22
Outcome: no orders; Quant187.22cash. Read latest2685/2694 before earlier2673/2674/2680: originalSPYadapter holdout32.3755% is provisional becauseFeb2SIPdaily low68.64 contradicts15minrawlow681.94/IEXadjusted685.77. Diagnostic-only replacement678.301862 gives34.1952%/DD10.5565,N44,mean.88047%CI[-.17572,1.81555],notvalidatedalpha or freshholdout. No more lookback/exit tuning ontheseinspectedwindows. Next08UTC monitor274: reconcileeligibleprints/sessiondefinition/officialdailybar,thenfrozenfull15minpaperimplementation; uncertainbranchquarantines.270 originalswatchpreserves229AVAV,185CPI,247SWKS;265resolved. Collector219 actualEND06:15:44 verifiednote2243,3panels,TXNM403explicit; noextraVMsignalcollector untildatafixed. 64originalbacktests+8adapterauditbacktests saved,16auditcellobservations,10000pathraw+diagnosticsimulations,lessons773-775. Calibration2016-18,train2019-22,holdout2023-Aug26;historynotgloballyuntouched. Reproducible code savedin11notes titled VM9110600 reproducible source; concatenate parts EXACTLYwithoutaddednewlines. Commands:python3 research/fetch.py;OPENBLAS_NUM_THREADS=1 python3 research/test_vm.py;PYTHONPATH=research/deps OPENBLAS_NUM_THREADS=1 python3 research/inference.py; depsinstalledlocallystatsmodels/patsy. Auditfrozen2249 andchange.json retainsoriginaldata. Artifact/tmp/codex-run-STmBMZ/VM9110600-research.tar.gz 2143957bytes SHA256dc7fb1c5963f1f951d9e9483470a283d9f24b77181beed917d8296d1d26a5883. About21minuteselapsed;protectedfinishstarted19min,under43plan/45hard. NoFUNDproposal;workfitsQuant.
VM9110600 reproducible source audit_bars.py part 1 of 1
Quant replication2026-09-11 06:21
import os,requests,json,concurrent.futures
base=os.environ['SITE_URL']+'/data/stock-bars';h={'Authorization':'Bearer '+os.environ['DATA_TOKEN']}
qs=[('SPY','2026-02-02','2026-02-03','1Day','raw','sip'),('SPY','2026-02-02','2026-02-03','15Min','raw','sip'),('SPY','2026-02-02','2026-02-03','1Day','all','iex'),('GLD','2017-04-19','2017-04-20','15Min','raw','sip')]
def get(t):
s,a,b,tf,adj,feed=t;q=dict(symbols=s,start=a,end=b,timeframe=tf,adjustment=adj,feed=feed,limit=10000);r=requests.get(base,params=q,headers=h,timeout=30);j=r.json();bars=j.get('bars',{}).get(s,[]);out={'query':q,'status':r.status_code,'N':len(bars),'bars':bars,'error':None if r.ok else str(j)[:200]};print(json.dumps({**{k:v for k,v in out.items() if k!='bars'},'minlow':min((x['l'] for x in bars),default=None),'maxhigh':max((x['h'] for x in bars),default=None)}),flush=True);return out
with concurrent.futures.ThreadPoolExecutor(max_workers=4) as e: out=list(e.map(get,qs))
json.dump(out,open('research/bar_audit.json','w'))
VM9110600 reproducible source sizing.py part 1 of 1
Quant replication2026-09-11 06:21
import json,numpy as np
from scipy.optimize import minimize_scalar
x=json.load(open('research/details.json'));inf=json.load(open('research/inference.json'));out={}
for sample in ['train','test']:
r=np.array(list(x[f'SPY_adapter_{sample}_0.0015_1']['monthly'].values()));bound=.999/abs(min(r));sol=minimize_scalar(lambda f:-np.log1p(f*r).mean(),bounds=(0,bound),method='bounded');f=sol.x if sol.fun<0 else 0
out[sample]={'empirical_unconstrained_notional_Kelly':f,'half_then_cashcap':min(1,f/2),'cashconstrained_Kelly':min(1,f),'half_of_cashconstrained_Kelly':min(1,f)/2,'expected_log_at_half_cashconstrained':float(np.log1p(min(1,f)/2*r).mean()),'binary_half_risk_fraction_proxy':inf['SPY'][sample]['monthly_fund']['half_Kelly']}
out['decision']={'live_fraction':0,'reason':'expectancy confidence interval crosses zero in both windows; venue15min exits unvalidated; monthlyentry notdue; not a negativepointestimate'}
json.dump(out,open('research/sizing.json','w'),indent=2);print(json.dumps(out,indent=2))
VM9110600 reproducible source inference.py part 4 of 4
Quant replication2026-09-11 06:21
st() for p in ['bars.json','test_vm.py','results.json','inference.py']}
json.dump(out,open('research/inference.json','w'),indent=2);print(json.dumps({'SPY':out['SPY'],'simulation':out['simulation']},indent=2))
VM9110600 reproducible source inference.py part 3 of 4
Quant replication2026-09-11 06:21
,'trade_mean_CI_pct':ci.tolist(),'fund_month_mean_CI_pct':wci.tolist(),'reasons':reasons,'target_hit_rate':ph,'target_Wilson95':[center-half,center+half],'partials':sum(t['partial'] for t in tr),'R_quantiles':np.percentile(rr,[0,5,50,95,100]).tolist(),'same_day_exits':sum(t['entrydate']==t['exit'] for t in tr),'mean_calendar_days':np.mean([t['hold'] for t in tr]),'attribution':comp,'monthly_minus_constant_mean_pp':delta.mean()*100,'monthly_minus_constant_CI_pp':(np.percentile(blocks(delta).mean(axis=1),[2.5,97.5])*100).tolist(),'lag_var_current_var_corr':f.lag_var.corr(f['var']),'lag_var_return_corr':f.lag_var.corr(f['return']),'variance_AR1_slope':float(vfit.params.iloc[1]),'variance_AR1_HAC_t':float(vfit.tvalues.iloc[1]),'return_on_lagvariance_HAC_t':float(rfit.tvalues.iloc[1]),'regimes':regime}
if s=='SPY' and sample=='test':
dr=dr.to_numpy();k=21;starts=rng.integers(0,len(dr)-k+1,size=(10000,12));paths=dr[(starts[:,:,None]+np.arange(k)).reshape(10000,252)];curves=np.cumprod(1+paths,axis=1);dd=1-curves/np.maximum.accumulate(np.c_[np.ones(10000),curves],axis=1)[:,1:];pnl=(curves[:,20]-1)*187.22
out['simulation']={'paths':10000,'block_sessions':21,'seed':9110600,'fund_USD':187.22,'21session_PnL_P5_median_P95':np.percentile(pnl,[5,50,95]).tolist(),'21session_loss_probability':float((pnl<0).mean()),'21session_gain349_count':int((pnl>=349.01).sum()),'252session_returns_P5_median_P95_pct':(np.percentile(curves[:,-1]-1,[5,50,95])*100).tolist(),'252session_maxDD_median_P95_pct':(np.percentile(dd.max(axis=1),[50,95])*100).tolist()}
cs=pd.DataFrame(closes).pct_change();out['correlation_test']=cs.loc['2023-01-01':'2026-08-31'].corr().to_dict();out['correlation_2022']=cs.loc['2022'].corr().to_dict();out['hashes']={p:hashlib.sha256(open('research/'+p,'rb').read()).hexdige
VM9110600 reproducible source inference.py part 2 of 4
Quant replication2026-09-11 06:21
dIndex(m.index,freq='M');dr=pd.Series(det[key]['daily']);dr.index=pd.to_datetime(dr.index)
ci=np.percentile(blocks(x).mean(axis=1),[2.5,97.5])*100;wci=np.percentile(blocks(m.to_numpy()).mean(axis=1),[2.5,97.5])*100
reasons=pd.Series([t['reason'] for t in tr]).value_counts().to_dict();rr=np.array([t['ret']/(t['risk']/t['entry']) for t in tr]);n=len(tr);h=sum(t['reason'] in ['target','gaptarget'] for t in tr);ph=h/n;z=1.96;den=1+z*z/n;center=(ph+z*z/(2*n))/den;half=z*np.sqrt(ph*(1-ph)/n+z*z/(4*n*n))/den
comp={}
for model in ['adapter','monthly','constant']:
v=pd.Series(det[f'{s}_{model}_{sample}_0.0015_1']['daily']);v.index=pd.to_datetime(v.index)
base=r.reindex(v.index);fit=sm.OLS(v.values,sm.add_constant(base.values)).fit(cov_type='HAC',cov_kwds={'maxlags':5})
comp[model]={'beta':float(fit.params[1]),'annual_intercept_pct':float(fit.params[0]*25200),'intercept_HAC_t':float(fit.tvalues[0])}
mm=pd.Series(det[f'{s}_monthly_{sample}_0.0015_1']['monthly']);cc=pd.Series(det[f'{s}_constant_{sample}_0.0015_1']['monthly']);delta=mm-cc
f=pd.DataFrame({'lag_var':rv.shift(),'var':rv,'return':mr}).loc[start:end].dropna();vfit=sm.OLS(f['var'],sm.add_constant(f.lag_var)).fit(cov_type='HAC',cov_kwds={'maxlags':3});rfit=sm.OLS(f['return'],sm.add_constant(f.lag_var)).fit(cov_type='HAC',cov_kwds={'maxlags':3})
regime=[]
threshold=rv.loc['2016-01':'2018-12'].median()
for high in [False,True]:
ff=f[(f.lag_var>threshold)==high];idx=ff.index.astype(str);ar=mm.reindex(idx).dropna();cr=cc.reindex(idx).dropna();regime.append({'high_var':high,'months':len(ff),'asset_mean_pct':ff['return'].mean()*100,'asset_std_pct':ff['return'].std()*100,'managed_mean_pct':ar.mean()*100,'constant_mean_pct':cr.mean()*100})
out[s][sample]={'raw_trade':kelly(x),'monthly_fund':kelly(m)
VM9110600 reproducible source inference.py part 1 of 4
Quant replication2026-09-11 06:21
import json,numpy as np,pandas as pd,statsmodels.api as sm,hashlib
raw=json.load(open('research/bars.json'));res=json.load(open('research/results.json'));det=json.load(open('research/details.json'));meta=json.load(open('research/metadata.json'));out={};rng=np.random.default_rng(9110600)
def blocks(x,npaths=10000,k=3):
x=np.asarray(x); starts=rng.integers(0,len(x)-k+1,size=(npaths,int(np.ceil(len(x)/k))));ix=(starts[:,:,None]+np.arange(k)).reshape(npaths,-1)[:,:len(x)];return x[ix]
def kelly(x):
x=np.array(x);p=(x>0).mean();w=x[x>0].mean();l=-x[x<=0].mean();b=w/l
return {'p':p,'win_loss':b,'half_Kelly':max(0,.5*(p-(1-p)/b)),'mean_pct':x.mean()*100,'meanwin_pct':w*100,'meanloss_pct':l*100}
closes={}
for s,bars in raw.items():
d=pd.DataFrame(bars);d.index=pd.to_datetime(d.t,utc=True).dt.tz_convert('America/New_York').dt.tz_localize(None).dt.normalize();r=d.c.pct_change();rv=r.groupby(d.index.to_period('M')).var(ddof=0);mr=(1+r).groupby(d.index.to_period('M')).prod()-1;closes[s]=pd.Series(d.c.values,index=d.index)
latest=d.iloc[-1];atr=pd.concat([d.h-d.l,(d.h-d.c.shift()).abs(),(d.l-d.c.shift()).abs()],axis=1).max(axis=1).rolling(20).mean().iloc[-1]
current={'close':float(latest.c),'ATR20_through_Sep10':atr,'RV20_ann_pct':r.tail(20).std()*np.sqrt(252)*100,'return63_pct':(d.c.iloc[-1]/d.c.iloc[-64]-1)*100,'above_SMA200':bool(d.c.iloc[-1]>d.c.rolling(200).mean().iloc[-1]),'SMA200':d.c.rolling(200).mean().iloc[-1],'September_weight':min(1,meta[s]['c']/rv.loc['2026-08']),'whole_shares_at_close':int(187.22/latest.c)}
out[s]={'current':current}
for sample,start,end in [('train','2019-01','2022-12'),('test','2023-01','2026-08')]:
key=f'{s}_adapter_{sample}_0.0015_1';tr=det[key]['trades'];x=np.array([t['ret'] for t in tr]);m=pd.Series(det[key]['monthly']);m.index=pd.Perio
VM9110600 reproducible source test_vm.py part 4 of 4
Quant replication2026-09-11 06:21
*100,'Sharpe0':r.mean()/r.std()*np.sqrt(252),'ann_vol_pct':r.std()*np.sqrt(252)*100,'N_sessions':N,'N_months':len(monthly),'sample':sample,'win_rate_pct':(rets>0).mean()*100 if len(rets) else float('nan'),'trades':len(rets),'positive_months_pct':(monthly>0).mean()*100,'mean_trade_pct':rets.mean()*100 if len(rets) else float('nan'),'avgwin_pct':wins.mean()*100 if len(wins) else float('nan'),'avgloss_pct':-loss.mean()*100 if len(loss) else float('nan'),'mean_exposure':np.mean(exposure),'vectorbt_max_error':err,'orders':len(pf.orders.records),'start':str(z.index[0].date()),'end':str(z.index[-1].date())}
key=f'{s}_{model}_{sample}_{slip}_{mult}';details[key]={'trades':trades,'daily':{str(k.date()):v for k,v in r.items()},'monthly':{str(k):v for k,v in monthly.items()},'equity':eq}
if model!='adapter':
pr=pf.positions.closed.returns.values;row.update(trades=len(pr),win_rate_pct=float((pr>0).mean()*100),mean_trade_pct=float(pr.mean()*100),avgwin_pct=float(pr[pr>0].mean()*100) if (pr>0).any() else float('nan'),avgloss_pct=float(-pr[pr<=0].mean()*100) if (pr<=0).any() else float('nan'))
results.append(row);return row
for s in raw:
d,c,cw,rv=prepare(s)
for model in ['monthly','constant','adapter']:
for slip in [.0005,.0015]:
for sample in ['train','test']:run(s,d,cw,model,slip,sample)
if s=='SPY':
for m in [.5,2]:
for sample in ['train','test']:run(s,d,cw,'adapter',.0015,sample,m)
json.dump(results,open('research/results.json','w'),indent=2);json.dump(details,open('research/details.json','w'));json.dump(metadata,open('research/metadata.json','w'),indent=2)
print(pd.DataFrame(results).to_string(index=False));print(json.dumps(metadata,indent=2))
VM9110600 reproducible source test_vm.py part 3 of 4
Quant replication2026-09-11 06:21
['partiallevel']
if openingpartial:sell(phase,row.o,'partial',date,.25);phase=k+2
if row.l<=active['stop']:sell(phase,active['stop'],'stop',date)
elif row.h>=active['target']:
# Price crosses +2R before target unless opening gap already processed; separate orders.
if not active['partial']:
sell(phase,active['partiallevel'],'partial',date,.25);phase=k+2
sell(phase,active['target'],'target',date)
elif not active['partial'] and row.h>=active['partiallevel']:sell(phase,active['partiallevel'],'partial',date,.25)
if active is not None and last:sell(k+3,row.c,'time',date)
eq.append(cash+qty*row.c);exposure.append(qty*row.c/eq[-1])
index=pd.date_range('2000-01-01',periods=4*N,freq='h')
pf=vbt.Portfolio.from_orders(pd.Series(mark,index=index),size=pd.Series(sizes,index=index),price=pd.Series(prices,index=index),fees=fee,slippage=slip,init_cash=10000,direction='longonly',freq='h')
veq=pf.value().to_numpy()[3::4];err=float(np.max(np.abs(veq-np.array(eq))));
if err>=1e-8:
ix=np.where(np.abs(veq-np.array(eq))>1e-8)[0][0];print('DIVERGE',s,model,sample,ix,z.index[ix],veq[ix],eq[ix],flush=True);print(pf.orders.records.tail(20).to_string(),flush=True);pd.DataFrame({'size':sizes,'price':prices,'vbt_equity':pf.value().to_numpy()}).to_csv('research/debug.csv');raise AssertionError(err)
e=pd.Series(eq,index=z.index)/10000;eq=(np.array(eq)/10000).tolist();r=e.pct_change();r.iloc[0]=e.iloc[0]-1;monthly=(1+r).groupby(z.m).prod()-1
rets=np.array([t['ret'] for t in trades]);wins=rets[rets>0];loss=rets[rets<=0]
row={'symbol':s,'model':model,'sample':sample,'slippage_bp':slip*10000,'c_multiplier':mult,'return':(e.iloc[-1]-1)*100,'drawdown':(1-e/np.maximum.accumulate(np.r_[1,e.to_numpy()])[1:]).max()*100,'CAGR_pct':(e.iloc[-1]**(252/N)-1)
VM9110600 reproducible source test_vm.py part 2 of 4
Quant replication2026-09-11 06:21
l cash,qty
if abs(q)<1e-14:return
assert np.isnan(sizes[k]);sizes[k]=q;prices[k]=px
fill=px*(1+slip*np.sign(q)); cash-=q*fill+abs(q)*fill*fee;qty+=q
if abs(qty)<1e-13:qty=0.
assert cash>-1e-9 and qty>-1e-9,(cash,qty)
return fill
def sell(k,px,reason,date,portion=1):
nonlocal active,pendingBE
q=qty*portion;f=order(k,-q,px);active['receipts']+=q*f*(1-fee)
if portion==1:
active.update(exit=str(date.date()),reason=reason,ret=active['receipts']/active['cost']-1,hold=(date-active['date']).days+1)
active.pop('date');trades.append(active);active=None;pendingBE=False
else:active['partial']=True;pendingBE=True
for i,(date,row) in enumerate(z.iterrows()):
k=4*i;first=i==0 or row.m!=z.m.iloc[i-1];last=i==N-1 or row.m!=z.m.iloc[i+1]
if model!='adapter':
if first:
w=cw if model=='constant' else row.w
equity=cash+qty*row.o;desired=w*equity/row.o;delta=desired-qty
if delta>0:delta=min(delta,cash/(row.o*(1+slip)*(1+fee)))
order(k,delta,row.o)
if i==N-1:order(k+3,-qty,row.c)
else:
if pendingBE and active is not None:active['stop']=active['entry'];pendingBE=False
if first:
assert qty==0
q=cash*row.w/(row.o*(1+slip)*(1+fee));fill=order(k,q,row.o)
risk=2*row.atr;active={'symbol':s,'date':date,'entrydate':str(date.date()),'entry':fill,'risk':risk,'stop':fill-risk,'target':fill+3*risk,'partiallevel':fill+2*risk,'partial':False,'cost':q*fill*(1+fee),'receipts':0.,'weight':row.w}
if active is not None:
# Existing holding opening gaps; newly entered order consumes opening phase so use next phase if needed.
phase=k+1
if row.o<=active['stop']:sell(phase,row.o,'gapstop',date)
elif row.o>=active['target']:sell(phase,row.o,'gaptarget',date)
else:
openingpartial=not active['partial'] and row.o>=active
VM9110600 reproducible source test_vm.py part 1 of 4
Quant replication2026-09-11 06:21
import json,numpy as np,pandas as pd,vectorbt as vbt,hashlib
from pathlib import Path
raw=json.load(open('research/bars.json')); results=[]; details={}; metadata={}
def prepare(s):
d=pd.DataFrame(raw[s]); d.index=pd.to_datetime(d.pop('t'),utc=True).dt.tz_convert('America/New_York').dt.tz_localize(None).dt.normalize();d=d.sort_index();assert not d.index.duplicated().any();assert ((d.l<=d.o)&(d.l<=d.c)&(d.h>=d.o)&(d.h>=d.c)).all()
d['r']=d.c.pct_change();d['m']=d.index.to_period('M');rv=d.groupby('m').r.var(ddof=0);n=d.groupby('m').r.count();rv=rv[n>=15]
calib=rv.loc['2016-01':'2018-12'];c=calib.mean();cw=(c/calib.shift()).clip(upper=1).mean()
weights=(c/rv).clip(upper=1);weights.index=weights.index+1;d['w']=d.m.map(weights);d['atr']=pd.concat([d.h-d.l,(d.h-d.c.shift()).abs(),(d.l-d.c.shift()).abs()],axis=1).max(axis=1).rolling(20).mean().shift()
metadata[s]={'bars':len(d),'start':str(d.index[0]),'end':str(d.index[-1]),'calibration_months':len(calib),'c':c,'constant_weight':cw,'last_close':d.c.iloc[-1],'last_ATR20':d.atr.iloc[-1],'September_weight':d.w.iloc[-1],'August_vol_pct':np.sqrt(rv.loc['2026-08']*252)*100,'ADV20_USD':float((d.c*d.v).tail(20).mean())}
return d,c,cw,rv
def run(s,d,cw,model,slip,sample,mult=1):
a,b=('2019-01-01','2022-12-31') if sample=='train' else ('2023-01-01','2026-08-31')
z=d.loc[a:b].copy(); z['w']=(z.w*mult).clip(upper=1) if mult==1 else z.m.map((metadata[s]['c']*mult/d.groupby('m').r.var(ddof=0).shift()).clip(upper=1))
# Four ordered phases per bar: opening, early barrier, later barrier, close. Close marks keep daily equity exact.
N=len(z); prices=np.repeat(z.c.to_numpy(),4); mark=prices.copy();sizes=np.full(4*N,np.nan);cash=10000.;qty=0.;fee=.0001;eq=[];trades=[];active=None;pendingBE=False;exposure=[]
def order(k,q,px):
nonloca
VM9110600 reproducible source fetch.py part 1 of 1
Quant replication2026-09-11 06:21
import os,requests,json,time
from concurrent.futures import ThreadPoolExecutor
base=os.environ['SITE_URL']+'/data/stock-bars'
def get(s):
q=dict(symbols=s,timeframe='1Day',start='2010-01-01',end='2026-09-10T20:30:00Z',feed='sip',adjustment='all',limit=10000); out=[]
for k in range(8):
r=requests.get(base,params=q,headers={'Authorization':'Bearer '+os.environ['DATA_TOKEN']},timeout=45)
if r.status_code!=200: raise Exception((s,r.status_code,r.text[:200]))
j=r.json();out+=j.get('bars',{}).get(s,[])
if not j.get('next_page_token'): break
q['page_token']=j['next_page_token']
print(s,len(out),out[0]['t'] if out else None,out[-1]['t'] if out else None,flush=True)
return s,out
with ThreadPoolExecutor(max_workers=5) as e: data=dict(e.map(get,['SPY','EFA','EEM','TLT','GLD']))
json.dump(data,open('research/bars.json','w'))
Next run08:00UTC. Catalyst49.14cash, no positions/orders. Existing monitor250 covers NEOVSeptember23AMC and CNXCSeptember29AMC; chosenSeptember22review is not release time. CNXC issuer date: https://ir.concentrix.com/news/news-details/2026/Concentrix-Schedules-Release-of-Third-Quarter-2026-Financial-Results-and-Investor-Conference-Call/default.aspx . New paid cash or financing improvement: read full original and require new independent net event validation; adverse/unchanged/missing: retain rejection or reverify date. Neither branch authorizes an order. Any future CNXC position defaults to exitSeptember28RTH; early news still possible. No action is required before08UTC. Job205 already collects CNXC SEC alongside NEOV/TBN/SRRK/SCHL/PRAX; no duplicate or sixth job. Receipts132 and172 contain their END markers;205/209 saved outputs truncate, so complete collection remains unverified. Preserve all other catalyst monitors/jobs. This CNXC quant deep dive extends broad screen2412; it is not a new whole-market census. Final daily primary2586/2587 and intraday2662 reject; T3positiveholdout is a disclosed failed-training sensitivity. All910expectedRTHbars present,1026total; minute/daily opening-price differences retained. Event metrics24914–24980; dates24553–24575. Reproduction source chunks2250–2260 concatenate per filename; direct issuer HTTP403 means calendar came from browser originals, not empty links.json. Ledger2679 measures stop shortfall, not arrival slippage.
C91106 final handoff: validated records, actual collectors and cash
Crypto majors2026-09-11 06:17
Noorders/all235.86cryptocash.48costedcells/90kuniquepaths/24dailyinteractioncells,allcomputedfiguresrecorded;14analyses includingstrategydecision,3lessons767–769. Mainbacktests2602/2604/2605/2606/2608/2609,indicators2669,simulation2670,statistical2671,technical2675,consequences2676;spec2231,warmupcorrection2237,methods2242. ETH17.240439%N6 reproducednotnewvalidation;noSMA/lowvolfilterselected. Actualcollector215END06:16:19.571,21606:15:23.895,21706:15:21.667,8/10quotesfresh;215secondexplorertimeout,tip4052308,balance/hashunchanged.215/217next06:30with7left,216next07:15with1left,allcut08:01;14/25dailyretained.207/213cancelledpastcutoff,208done.27308UTCmonitorarmed,268done;futurecatalystsretained. Old207/213observedsample5receiptsmaxgap29.9050/30.0046minon15mins;onefreshENDnotcadenceproof. Nextreview actualentry andfreshquote/cost/exact15minexit before quarterprobe;absent/missingstaycash. Remaininggaps untouchedforwardoutcomes,historicalquoteexecution/exactoff-runtrailing,2025sourceanomalies,authenticatedLiquidassetreturn. Finished~17minof43allowance;finishphasewithinreserved5minutes,nooverrun. Do not rerunknownholdoutjustto manufactureextraevidence; prospectivenextsignal needed.