Stats

A news scan found a report about common major-coin risk and funding distribution. The report is based on three premises. Premise 1 looks at the relationship between the prices of three cryptocurrencies, Bitcoin, Ethereum, and Solana, over 90 days. The correlation, or how closely the prices move together, is 0.8680598 for Bitcoin and Ethereum, 0.7972226 for Bitcoin and Solana, and 0.8025502 for Ethereum and Solana. Volatility, or how much the prices fluctuate, is 38.64% for Bitcoin, 52.07% for Ethereum, and 57.36% for Solana per year. Premise 2 examines the funding settlements on the OKX exchange from August 10 to September 12. The latest funding rates are 0.324729% for Bitcoin, 0.274278% for Ethereum, and 0.605550% for Solana, which are all within normal ranges. Premise 3 considers the current open interest, or the number of outstanding contracts, for each coin: 27112.401 for Bitcoin, 622879.194 for Ethereum, and 2717751.07 for Solana. The conclusion is that the three major coins have correlated exposure, but there is no extreme funding trade. However, this conclusion has limitations, such as the possibility of correlations changing during crashes and the fact that funding rates on one exchange may not represent all venues. The data used for this report comes from https://runway.adityaperswal.workers.dev/data/crypto-bars, https://www.okx.com/api/v5/public/funding-rate-history, and https://www.okx.com/api/v5/public/open-interest.

A news scan found the headline that on September 12, some stock prices changed, but they still did not meet the requirements. The stocks in question are BB, which had 4 out of 12 requirements met, HITI, which had 1 out of 9 requirements met, and GIS, which had 0 out of 9 requirements met. An edge test, which is a test to determine the effectiveness of a trading strategy, was performed on these stocks. For BB, the test showed that the median absolute return was 9.987505% from September 2023 to June 2026. On September 11, the stock price was 7.67 and the option price was 0.45, plus 0.05 in assumed costs, which requires a return of 10.821382% to be profitable, but this was only met 4 out of 12 times. For HITI, the median return was 5.494505%, and on September 11, the stock price was 2.53 and the option price was 0.21, which requires a return of 9.090909% to be profitable, but this was only met 1 out of 9 times. For GIS, the median return was 3.063603% from June 2024 to July 2026, and on September 10, the stock price was 35.96 and the option price was 0.38, which requires a return of 12.430478% to be profitable, but this was not met at all. The closing prices of the stocks are not the same as the executable prices, and the intrinsic value, which is the value of an option based on its underlying stock, excludes the residual time value, which is the value of an option based on the time until it expires. The dates for the earnings reports of these stocks are: BB on September 12, HITI on September 14, and GIS on September 23. The sources of this information include option contracts, stock bars, and historical issuer URLs. A review of the data shows that the returns of BB and HITI recovered with the original dates and URLs, and GIS had nine consecutive earnings added from the issuer's remarks and daily SIP bars. The prices of GIS moved by -4.5862, +0.6834, -3.0636, -2.0500, -5.1148, -0.7720, +3.3705, -2.9655, and +8.5355%. There was no selection of outcomes. Further analysis shows that a put option below the absolute median return of BB still only covers 3 out of 12 directional falls, so it is not possible to authorize entry or hold through. The costed premium ceiling of BB is 0.302576 at a fixed stock price of 7.67, which yields 5 out of 12, but this is a review trigger, not an order price. The option prices of GIS are unaffordable. The validity of the data is correctly computed, but there are limits to the soundness of the data, including small sample size, regime change, event timing, and option residual value. The rejection of the trades follows the failed mandate gate, and cash is not certainty seeking.

The fund started with $49.14 in cash and no active trades or pending orders. A review of past trades from the Catalyst fund showed a loss of $31.93, with a win rate of 0% and a mean return of -16.38%. The RARE trade lost 45.86% of its value, which is 6.44 times the intended stop risk. The IONS and MIRM trades were closed by the operator, so their returns do not reflect the natural outcome of the strategy. The planned target-to-risk ratios for these trades were 1.72, 1.40, and 1.08, respectively, all of which are below the current 3R mandate. Based on these results, it was concluded that there is no positive return from this record, and the allocation was set to 0 until a new valid setup is found. This does not mean that every FDA run-up has negative expectancy, as early approvals can produce upside, such as the recent SRRK announcement. The historical trades were fractional, limiting the ability to exit during off-hours. The entry quotes and timestamps are not available, so it is not possible to determine if there was slippage in stop loss execution. The current total loss for the fund would be 5.77% of the account, and a hypothetical 50% gap would result in a loss of 2.89%, regardless of monthly revenue needs.

The automated trading fund manager made the following premises: there are five liquid ETF proxies, each with 2687 daily bars from 2016 to September 10, 2026, and 925 common daily returns from 2023 forward. The correlation between EFA and EEM is 0.7863, between SPY and EFA is 0.7591, between SPY and TLT is 0.1199, and between SPY and GLD is 0.1527. This means that different tickers do not create independent equity bets. The latest 252 session returns are EEM 33.44%, EFA 18.27%, GLD 18.65%, SPY 17.83%, and TLT -5.32%. The current monthly job has 15 retained, with no change to its eligibility gates. A range census of over 20% found one SPY anomaly and zero others. The diagnostic threshold does not validate smaller anomalies. Whole-share affordabilities at 187.22 are SPY 0, EFA 1, EEM 2, TLT 2, and GLD 0. The fund recomputed three forced liquidations, with a win rate of 1/3, average win of 7.93994%, average loss of 5.26143%, mean of -0.860974%, and realized loss of -1.046053 USD. The raw half-Kelly risk proxy is -0.054218, clipped to 0. The conclusion is that no alpha or exact exit odds can be inferred from correlations, ranks, or three censored outcomes. No order was made, as evidence would change after a frozen costed rule and eligible execution pass, not merely favorable momentum. The source of this data is https://runway.adityaperswal.workers.dev/data/stock-bars and query_trades(fund=quant), with research from broad_risk.json and ledger.json.

The automated trading fund manager performed a live ledger audit on the Catalyst fund. The audit found that the fund has three closed trades and no current holdings or orders. The previous ledger, numbered 2679, still shows a profit and loss of -31.937258, with no winning trades out of three, and an undefined average winning return. Two trades were stopped early by the operator, which affected the intended holding periods. One trade, RARE, entered at 25.84, stopped at 24, and exited at 13.98976, resulted in a loss of 6.44035. This shows that a stop does not guarantee a limit on losses. The audit also found that it is impossible to estimate the probability of winning trades for the new Catalyst fund due to the lack of winning trades. The current feasibility of the fund is 49.14, with an account balance of 850.99, and a 50% fund position gap of 2.8872% of the account. The audit notes that the losses may be understated due to censored data, but this does not validate the performance of an untested final-session entry. The source of the data is a query of trades for the Catalyst fund, using prior measured figures from ledger 2679, which is not a new sample.

A news scan found the report 'R91108 TSM beta and tanker concentration: descriptive risk before CPI'. Premises are as follows: seven names have matched 1513 SIP daily bars through September 10. The last 60 log returns regress TSM on SPY with an intercept, which is a statistical method to measure the relationship between two variables, with a beta of 2.37655 and a correlation of 0.62808; the HAC, or Heteroskedasticity and Autocorrelation Consistent, 5 confidence interval is between 1.62562 and 3.12748. TSM daily mean is -0.05126%, standard deviation is 2.81771%, and the range is from -7.23683 to 7.36599%. The correlation between FRO and STNG is 0.77894, and between LPG and FRO is 0.68235, indicating that tanker positions are joint-risk exposure. The correlation between TSM and LPG is -0.0218, which is not stress independence, meaning they do not move independently during times of stress. In conclusion, macro reaction could dominate company news, but there is no forecast for the direction of CPI returns. The statistics are selected from a recent window and are not causal, and no filter was adopted. The source is https://runway.adityaperswal.workers.dev/data/stock-bars, with a query of trades for the fund with IDs 3 and 6, and there are no orders.

The daily edge was recomputed at 08UTC from stored samples dated by the issuer. For BB, the median absolute value was 9.987505%, with an option price of $0.40 for the call and $0.32 for the put at a stock price of $7.54, requiring a change of +/-11.405836% to be considered. With assumed costs of $0.05 per share, the required change was +/-12.068966%. For HITI, the median absolute value was 5.494505%, with an option price of $0.31 for the call and $0.16 for the put at a stock price of $2.56, requiring a change of +9.765625%/-8.593750%. For PLAY, the median absolute value was 10.624339%, with an option price of $0.40 for the call and $0.50 for the put at a stock price of $8.17, requiring a change of +27.294982%/-20.440636%. None of the options passed the 40% gate or had a documented 3x expected payout. The companies BB, HITI, and PLAY are scheduled to announce their financial results on September 12, September 14, and September 14, respectively. The options expire in 8 days for BB and 32 days for the others. The catalog and fresh SIP all closes were unchanged versus frozen data. Twelve BB and PLAY events, and nine HITI events, were retained without refit. Necessary contract price ceilings were calculated to reach 40% after assumed costs. The current prices exceed the ceilings substantially. Other options, such as HAIN Oct16P1, were also evaluated but failed to meet the criteria. A strong counterargument is that residual time value can exceed intrinsic response, but there is no measured 3x expected payout or current 40% coverage.

A news scan found the headline 'C91108 pre-open concentration and data continuity'. The analysis is based on three premises. Premise 1 looks at hourly data from August 19, 08:00 UTC, to September 11, 07:00 UTC, which includes 552 bars of data for each cryptocurrency, with no missing hours, and 551 aligned log returns. The correlation between Bitcoin and Ethereum is 0.844566, between Bitcoin and Solana is 0.732571, and between Ethereum and Solana is 0.726796. The hourly standard deviation is 0.480154%, 0.650954%, and 0.740780% for each respective pair. Premise 2 examines the latest 15-note sample, which has a maximum gap of 30.0107 minutes between quote receipts and 29.9372 minutes between entries on a 15-minute cadence. However, this sample is not exhaustive and does not prove there were no stop outages. Premise 3 looks at one closed cryptocurrency trade, where the operator liquidated after 47 minutes and 56 seconds, entering at 81614.8 and exiting at 81548.7, but it cannot estimate the strategy's exit odds or order-type slippage. The conclusion is that Bitcoin and Ethereum positions would be one correlated risk, so they should not be sized independently. The fresh complete bar window supports current signal measurement, not historical execution. The prior fear of missing tail remains separate from the correlation fact. The source of this information is https://runway.adityaperswal.workers.dev/data/crypto-bars, with queries from query_trades and query_research for the crypto fund.

A news scan found the headline 'VM9110600 replacement-low sensitivity does not clear the validation gate'. The details are as follows: Premises include Audit2685 and freeze2249 showing SIP Feb 2 low contradiction, with only low 68.64 to 678.30 diagnostic, and all rules fixed. The stress holdout return changes from 32.37547 to 34.19519 percent, and the drawdown changes from 12.33273 to 10.55650 percent, with 44 trades and 23 wins unchanged, and a mean of 0.84966 to 0.88047 percent, with a 95 percent 3-month block confidence interval of -0.17572 to 1.81555. Training remains at 5.55750 percent and the confidence interval crosses zero. Target 3 of 44 and 14 partials are unchanged, with holdout time exits changing from 23 to 22, stops from 16 to 17, and same-day exits from 3 to 2, and beta becomes 0.34605, with an annual intercept of 0.98628 percent, and HAC of 0.27589. There is no statistically demonstrated alpha. Diagnostic 10000 path 21 session PnL full 187.22 fund P5, median, and P95 are -6.54013, 1.19048, and 9.02180 USD, with a 41.19 percent loss probability, and an annual return median of 8.04715 percent and P95 drawdown of 13.18197 percent. The direct monthly wealth Kelly test is 10.45232 before halving and capping, which is positive but unreliable, and live is 0 due to uncertainty, data, and timing. Eight adapter audit backtests and eight unchanged continuous control observations are saved, with raw vintages retained. The conclusion is that the method needs price-eligibility reconciliation and prospective execution validation, not parameter retuning. Diagnostic improvement does not turn repair into independent out-of-sample evidence. The source is https://runway.adityaperswal.workers.dev/data/stock-bars, with research and audit results in results.json, inference.json, and sizing.json. Monitor 274 next 08 UTC.

A news scan found the headline 'VM9110600 data-quality override: quarantine raw adapter performance pending OHLC reconciliation'. This refers to a problem with the quality of data from a trading system. The system uses something called OHLC, which stands for Open, High, Low, Close, to track stock prices. On February 2, 2026, the raw data for the SPY stock showed a low of 68.64, a high of 693.21, and a close of 691.70, which is a 90.2949% range with a 0.4969% close return. However, the daily low was confirmed to be 69.005, which shows that the problem is not just with the adjustments. The exact date and time of the issue was on February 2, with a 15-minute extended session minimum of 681.94, compared to a daily low of 69.005. The IEX adjusted daily low was 685.77, which is not the same as the raw data. This discrepancy affects the entry-day barrier and 20-day ATR into March. The internal OHLC order checks and daily vector matches verify the arithmetic, but not the price truth. The conclusion is to override the original data, and the raw vendor results will remain recorded, but not accepted as edge or calibrated exit odds. A diagnostic replacement of the February 2 low with 678.301862 has been made, using a matched adjustment factor. This is a conservative extended session proxy, not an authoritative daily correction. The GLD 2017 low of 110.03 is also visible in the 15-minute data, but this is from the same feed and not independent validation. A gross outlier screen across 13435 rows flags one issue at a 20% threshold, but smaller bad prints remain possible. There are no live orders. The sources of this information include https://runway.adityaperswal.workers.dev/data/stock-bars, research/bar_audit.json, audit/change.json, and observations 24984, 24990, and 24991.

The CNXC91106 fund had three losses. The fund looked at the Catalyst fund and found three closed trades with no open positions or pending orders. The returns were: MIRM -0.439321%, IONS -2.842420%, and RARE -45.860062%. The total profit and loss was -31.937258 before system costs. The win rate was 0 out of 3 trades, and the average win was undefined. MIRM and IONS were sold by the operator, so their exits were not based on the original plan. RARE had a loss of -6.440348 times the risk, with an entry price of 25.84 and an exit price of 13.98976, which was 41.709333% below the intended stop price. This was a problem with the stop price, not with the execution of the order. The data needed to measure the slippage of orders is not available. The conclusion is that the sample is too small and mixed to estimate the chances of winning, but it shows that a stop price is not a guarantee of the maximum loss. The fund will keep its current sizing method and not make any new trades or repeat old lessons. The source of this information is a query of the trades in the Catalyst fund.

A news scan found the report on the VM9110600 attribution and uncertainty. The report states that the fund had positive returns, but this does not necessarily mean it can time volatility well. The fund's performance is based on several premises. Firstly, the primary SPY stress adapter had a return of 5.5575% from 2019 to 2022 and 32.3755% from 2023 to August 26, with 48 out of 44 trades being successful. Secondly, the raw trade mean was 0.62035% and 0.84966%, with a 95% confidence interval of -0.59212% to 1.80894% and -0.20030% to 1.78505% for the three-month moving-block mean. The capital-weighted fund monthly confidence interval was -0.39509% to 0.64125% and -0.20727% to 1.52007%. Thirdly, the daily SPY OLS, or ordinary least squares, and HAC, which stands for heteroskedasticity and autocorrelation, 5 adapter annual intercept was -0.90499% during training and 0.21390% during the holdout period, with a holdout beta of 0.36583. These are descriptive zero-yield cash regressions, not original excess-factor alpha. Fourthly, the continuous monthly managed stress holdout was 72.7201% with a drawdown of 12.6491% and a Sharpe ratio of 1.33907, while the constant was 89.0037% with a drawdown of 16.3501% and a Sharpe ratio of 1.41453. The managed-minus-constant mean was -0.20681 percentage points per month, with a three-month block confidence interval of -0.53573 to 0.08946. Less drawdown comes with less exposure, and there is no demonstrated incremental return premium. Fifthly, the external stress adapter had returns of -19.2499% and 4.2771% for EFA, -24.7622% and 12.8840% for EEM, -16.3080% and -14.8836% for TLT, and 8.5946% and 37.2622% for GLD, with no promotion of GLD. In conclusion, the primary finding is that it is still under prospective watch and not live adoption, with neither a negative universal volatility claim nor a positive alpha claim. The history reused elsewhere, with a three-month bootstrap that misses long regimes, and 64 cells are not 64 independent confirmations. The sources of the data are https://runway.adityaperswal.workers.dev/data/stock-bars, research/results.json, and inference.json, with a fixed 2230.

A news scan found the following data: The C91106 measured hourly tails and limitations of the live track record from September 11, 2025, to September 11, 2026, with 8757, 8755, and 8747 samples for Bitcoin, Ethereum, and Solana, respectively. The hourly standard deviation was 0.467913%, 0.646934%, and 0.717868% for the three cryptocurrencies. The 1st percentile was -1.39813%, -2.07364%, and -2.10302%, while the 99th percentile was 1.33560%, 1.84153%, and 2.11065%. The worst-case scenario was -4.91158%, -6.12774%, and -6.10547%. The standard deviation on weekends was 0.328454%, 0.503313%, and 0.604826%, which is lower than the 0.513047%, 0.695920%, and 0.758127% on weekdays, but this lower variance does not provide left-tail protection, meaning it does not protect against large losses. The mean log hour was -0.004500%, -0.006600%, and -0.008465%, which is dependent on the dataset and regime, and not a forecast. A sole live crypto trade for Bitcoin was made with an entry at 81614.8 and an exit at 81548.7, resulting in a win of 0/1, with the exit being neither a stop nor a target, and the trade was held for 47 minutes and 56 seconds, versus the expected 3-4 weeks. It is not possible to calibrate the setup win rate, stop odds, market versus limits, slippage, or session effects from this forced sample, and actual fee activities are unretrieved. A regression analysis using HAC7 and bootstrap CI showed that the results cross zero, but there is no claim of new independent 2026 holdout, and repeated data reuse is flagged. The conclusion is to use a useful tail distribution and source and coverage controls, and prospective outcomes are required to measure edge, not operator intervention. The validity of descriptive statistics is noted, but not causal claims, and the soundness of the provider price construction and unobserved bid-ask remains a limitation. The sources of the data are https://runway.adityaperswal.workers.dev/data/crypto-bars and query_trades fund crypto trade 10.

A news scan found the analysis of tanker correlation. The analysis used SIP adjusted daily simple returns from 2021 to 2024 for training and compared them to returns from January 2025 to September 10, 2026. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=LPG,FRO,STNG,SPY&timeframe=1Day&feed=sip&adjustment=all. The correlation between LPG and FRO rose from 0.442891 to 0.670902, between LPG and STNG from 0.508278 to 0.641647, and between FRO and STNG from 0.623877 to 0.755241. An ordinary least squares, or OLS, regression was run on LPG returns with SPY, FRO, and STNG, using Newey-West 5-lag standard errors. The test betas were 0.36734, 0.37926, and 0.34760, with an R² of 0.5110. The daily intercept was 0.04566%, which had a HAC t-statistic of 0.44567, and 0.06242% in training with a t-statistic of 0.87026. This analysis does not predict future returns, but rather describes current exposure. It also notes that different cargo, contracts, and routes can create divergence in prices despite shared recent prices. The consequence of this analysis is to consider aggregate future tanker position risk and gap scenarios across symbols, and not to assume independence from separate issuers. No hedge or order is authorized by this analysis.

A news scan found the text describing the analysis of the CNXC91106 event. The analysis is based on two premises. First, primary scheduling announcements for 22 historical events were recovered before measuring event returns, from 2021 to 2025, and June 2026, excluding January 2021 because it was not recovered, with the earliest study being March 2021, but not all listing history is available. The analysis also excludes 6 events whose dates were not announced before entry, including same-day unknown hour, referred to as T10. Second, a frozen T5, which is a type of statistical model, with a 1 ADR stop, meaning a 1 average daily range stop, and a 3 ADR target, meaning a 3 average daily range target, of 30 basis points, or 0.3 percent, per side, showed a mean train return of -0.480636 percent and a holdout return of -4.292546 percent. A circular block, which is a method of resampling data, with 2 bootstrap, meaning two iterations of resampling, and 20000 seed, meaning a random number generator seed, of 91106, showed a mean 95 percent train return of -2.215259 percent to 1.293171 percent and a holdout return of -5.773292 percent to -3.038524 percent. The analysis is conditional on 7 selected company events and notes that resampling cannot discover unseen tails, meaning extreme events. The full calendar T5 time only holdout showed a return of -11.1555 percent versus a barrier return of -26.5783 percent, which shows that a tighter stop policy hurts this sample and does not justify removing required stops. The conclusion is to reject the primary rule. The validity of the analysis is limited to calendar vintages, selection, early news, and regimes, and the soundness is limited because the training and holdout datasets failed to propose an entry gate. This is a chronological holdout for a new rule, not a pristine fund-wide untouched dataset. The analysis also notes that September 2023 overlaps with Web help and excluding it, the train still returns -4.86638 percent N14. The primary source URLs are observations 24553 to 24575, and the data can be found at https://runway.adityaperswal.workers.dev/data/stock-bars.

A query of trades for the supply fund returned two closed lots, both of which were liquidated by the operator on September 4, rather than being stopped or targeted by the strategy. The entry and exit prices for these trades were 44.49 and 46.18 for FRO, and 79.76 and 80.36 for STNG. These two gains do not provide enough information to estimate the size of potential losses or the likelihood of stop hits or future wins. Additionally, there are no benchmark quotes available to measure slippage from the entry and exit prices. The sources of this information are the fund's trade ledger query and trades 6 and 3. The original reasons for these trades are still claimed, but have not been freshly verified as physical evidence. The conclusion is that the size of trades should be determined from audited, independent history with explicit uncertainty, rather than from these two forced exits or funding pressure.

A fund manager analyzed BlackBerry's stock options. The options are for September 10 and have a strike price of $7.54. The options that expire on October 2 have a close price of $0.40 and require an 11.405836% increase in the stock price to be profitable, which has happened 3 out of 12 times, or 25%. The options that expire on October 2 also have a close price of $0.32 and require an 11.405836% decrease in the stock price to be profitable, which has happened 2 out of 12 times, or 16.67%. The manager used 12 consecutive earnings reports from September 2023 to June 2026 to make these predictions. A news scan found the headline that BlackBerry will announce its fiscal second-quarter 2027 results on September 12. The manager assumed a fixed chronological split and a roundtrip execution reserve of $0.05 per share. The manager also used a Jeffreys Beta prior, which is a type of statistical prior, to calculate the probability of true coverage being greater than or equal to 0.40, which is 55.4481%. The manager decided to retain the current rejection of the call threshold and will reconsider if the mark price changes to $0.15 at an unchanged stock price.

A question was raised about whether the disagreement in the daily and hourly performance of Ethereum, or ETH, is due to real timing or inconsistent bars. Prior to this, it was thought that the disagreement might be due to a measurement artifact. An analysis was done on 984 daily and 23612 hourly data points for Bitcoin and ETH, and 746 and 17881 data points for SOL, up to September 11 at 00 UTC. The analysis found that there were 981 full 24-hour days. For 2026, there were 252 matched days with no discrepancies in the open, high, low, and close prices. However, for 2024 and 2025, there were small mismatches in the close prices and larger anomalies in the open, high, and low prices in November and December. A requery on December 2 confirmed that the native open price of ETH was 3020.1375, which was different from the hourly price of 2798.84 at 00 UTC, a difference of 7.906758 percent. The cause of this discrepancy is still unresolved. The analysis also found that using identical price inputs for 2026 resulted in identical policy results. The daily performance of ETH was -17.632466 percent, but when only full hourly days were considered, the result was -6.406080 percent, which is an 11.226386 percent selection effect. The analysis also looked at different entry and exit strategies, including daily entry and hourly exit, hourly calendar entry and exit, and rolling 480-hour entry and exit. These strategies resulted in different performance results, including -16.866373 percent, 7.324178 percent, and 17.240439 percent. The validity of the results is limited to the frozen conditional experiment and stated coverage policies. The soundness of the results is also limited due to historical bar revisions, no executable historical quotes, and tiny reused 2026 samples. The strongest counterexample is the omitted March loss, which suggests that a data quality filter can manufacture performance. The conclusion is to stop using distinct daily strategy failure as a standalone veto of hourly edge and to freeze hourly watch while requiring its actual signal and executable policy. No new backtest parameter was selected. The sources used for this analysis include https://runway.adityaperswal.workers.dev/data/crypto-bars and https://docs.alpaca.markets/us/docs/real-time-crypto-pricing-data, as well as notes 2175 and 2180 and backtests 2523-25.

A news scan found the body of a research paper: Frozen2177 applies five separately normalized lag1..5 negative cross-sectional-return weights to nine ETFs and averages them. Individual overlays are long1/short1; offsets make aggregate gross less than 2, correcting shorthand in2177, with no renormalization. Cash-only control clips aggregate negatives and normalizes positives to 1. Next-open returns use weights known previous close; same-close reference is explicitly non-executable. Findings: long-short next open holdout gross +3.38%, mean 0.381bp/day; one-way turnover break-even 0.6897bp, versus 6bp baseline yielding -35.60% and 16bp -70.75%. Training gross +22.55%, 6bp -18.91%, break-even 3.164bp. Long-only baseline holdout +20.48% but stress -51.74%; this positive baseline retains market risk and does not validate short-term selection. High vol condition does not rescue holdout: long-short next open 6bp mean -4.61bp/day, long only -5.29. These are identification calculations with linear turnover costs, no borrow/financing/terminal-liquidation charge and no actual short order; do not rank as executable backtests. They are not the weekly ATR/3R portfolio, which separately reconciles with vectorbt. Daily industry results need not replicate monthly within-stock results. All 18 descriptive cells recorded; no parameter promoted. Source research/diagnostic.py/json and https://runway.adityaperswal.workers.dev/data/stock-bars; methodological primary https://www.nber.org/papers/w17653.

A news scan found the report titled REV9110400 market exposure and matched selection attribution. The report is based on two premises. First, all 20 weekly costed cells are reconciled to vectorbt, which is a backtesting framework, with a holdout period from September 10, 2021, to September 10, 2026, excluding the full current week. Second, the historical window is not globally untouched, meaning it has been adjusted in some way. The primary baseline holdout had a return of 34.9556%, a compound annual growth rate of 5.43259%, a market beta of 0.453798, and an annual simple ordinary least squares intercept of -0.051691%. The stress beta was 0.453772, the annual intercept was -8.00789%, and the t-value was -1.27233. Fixed-beta ordinary least squares, a statistical method, cannot eliminate all conditional exposures, which are risks that are dependent on certain conditions. On 220 training weekly opportunities, the time-only loser mean was 0.038393% versus an equal 9 basket mean of 0.192879%, resulting in a selection delta of -0.154486 percentage points. On 255 holdout weeks, the loser mean was 0.128452% versus an equal 9 mean of 0.172813% versus an SPY mean of 0.212909%, resulting in a delta of -0.044361 percentage points. The costs were identical at 6 basis points per side, and there is no claim of significant negative delta, just an absence of demonstrated incremental selection. The full-window uncosted SPY buy-hold return was 103.0918% in the training period and 121.2061% in the holdout period, versus an equal 9 return of 82.8131% and 113.6551%, respectively. These returns have different exposures and cannot substitute for a matched control. The conclusion is that the positive baseline total is compatible with equity exposure, but the signal-selection evidence fails. A counterexample is that nonlinear crash recovery might escape fixed beta, and a pre-frozen high-vol model was still stressed -7.2856% in the holdout period. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars, and the research results are available in research/results.json, inference.json, and matched_train/test.csv.

A study of aluminum stocks found that after an outage, the stocks of different companies moved in different directions. The study used a method called SIP all-adjusted daily bars, which means it looked at the daily prices of the stocks, adjusted for things like dividends and stock splits, from January 2020 to September 10, 2026. The data came from a website. The study compared the daily returns of the stocks from January 2025 to March 27, 2026, with the returns from March 30 to September 10. The returns after the study period were: AA -17.01%, CENX -10.65%, CSTM +11.21%, and KALU +39.53%. The study also looked at something called SPY betas, which measure how much the stocks move compared to the overall market. The betas were: 1.444, 1.361, 1.969, and 1.842. The study found that none of the stocks had returns that were significantly higher than the market as a whole. The correlations between the stocks, which measure how much they move together, were: AA and CENX 0.673, and CSTM and KALU 0.655. This means that buying two aluminum stocks is partly like making one bet. The study found that the outage did not necessarily mean that the stocks of companies that make aluminum would go up. In fact, the stocks of companies that use aluminum went up while the stocks of companies that make aluminum went down. The study concludes that the direction of the outage should not be used to predict the direction of the stocks. The study had some limitations, including only looking at a few companies and a short period of time, which means that the results may not be repeatable.

A clinical trial, Registry 187, randomized 187 adults, with 145 receiving the active treatment and 42 receiving a vehicle, or placebo. The primary composite result showed 65.5% versus 4.8%, a difference of 60.7 percentage points. The approximate unadjusted Wald 95% confidence interval was 50.62 to 70.78 percentage points, based on rounded rates. This trial did not have a prespecified inference or approval odds, and the population excluded aggressive, recurrent, or high-risk H-zone lesions. Serious treatment-emergent adverse events occurred in 3 out of 145 patients versus 2 out of 42 patients, but this is not a causal comparison. A trading holdout of 74 fixed momentum trades had a mean return of -0.70697% per event, with a 95% mean confidence interval of -2.40017% to 1.15124% based on 20,000 independent and identically distributed bootstrap samples with a seed of 91295, resulting in a total return of -52.2123%. This sampling of trades ignores regime and time correlation, and the confidence interval crossing zero and negative point estimates does not prove that all FDA strategies fail. The sponsor's registry and press release have one clinical origin, while the ITC has an independent legal origin, and SIP market transactions have a separate origin. There is no active-comparator clinical claim, and no conversion of clinical effect into stock win rate. Sources include a clinical trial registry and a stock market data website.

A news scan found the headline about BlackBerry. A study of twelve quarterly releases from September 2023 to June 2026 was done. The median absolute change in the stock price after these releases was 9.987505%. The study looked at the prices of call and put options for BlackBerry on September 11 and October 2. The call option price was 0.40 and the put option price was 0.32. The study found that the call option would require a 11.405836% increase in the stock price to be profitable, and the put option would require a 11.405836% decrease. The study also found that the stock price did not move enough to make either option profitable in 8 out of 12 cases. The options expire on October 2, which is 8 days after the quarterly release on September 24. The study used data from the BlackBerry investor website and other sources. The study had some limitations, including a small sample size and no independent historical options data. The results of the study suggest that the current call and put options may not be profitable.

A news scan found the headline 'R91102 live track record cannot calibrate a scarcity strategy'. The R91102 fund has a record of trades, including two entries on September 2 and an operator liquidation on September 4. The return on these trades was 3.805163 percent and 0.752257 percent, with a mean of 2.27871 percent. The profit and loss was 4.977287 USD, with two wins and no losses, resulting in a 100 percent win rate. However, this record cannot be used to estimate the average loss, or the frequency of stop or target hits, due to a lack of data on losses and the fact that the original six to eight week theses cannot be confirmed. The Kelly criterion, a measure of bet size, also cannot be reliably estimated. The entry stop is null and there is no fresh entry quote, which prevents consistent estimation of risk and slippage. It is noted that a 100 percent win rate is not claimed to be an edge. All final open, pending, and closing queries are empty. There is no holding horizon or review of exit changes. Independent costed backtests, rather than this sample, guide the size and selection of the subject and period, due to limitations.

A test was done on TCOM's earnings coverage, which is a way to measure how well a stock's options reflect its potential price movement. The test looked at the cost of options for TCOM, specifically the October 16 call option at $45 and the September 10 put option at $35, which cost $32 and $48 respectively. The test required a 17.105943% increase or a 10.801034% decrease in the stock price to be considered successful. The results showed that 0 out of 8 call options and 2 out of 8 put options met this requirement, which is 25% of the total. The median absolute percentage change was 4.590973%. The test used the stock's closing price on September 15 as a reference point and did not consider any intraday lows. The test also used historical data from a website, but the data was incomplete and not representative of the stock's full potential. The test found that the current options contracts do not provide enough coverage, with less than 40% of the contracts meeting the required criteria. Additionally, there was no documented evidence of an expected payout of at least 3 times the cost of the option. The test also found that the expiry date of October 16 is more than 7 days after the event. While it is possible for residual option time value to pay out despite the lack of intrinsic coverage, there is no historical evidence to support this. Therefore, the test conditionally rejected the options contract. The test's soundness is limited by stale market data, sparse samples, and changes in market conditions. To reconsider the contract, new evidence is needed, such as a more affordable contract with at least 40% directional coverage or a documented expected payoff of at least 3 times the cost.