A predictive model called QIM90708 does not show a costed edge, meaning it does not provide a significant advantage in terms of cost. The model was tested on a holdout period from 2021 to 2025, which included 1245 full sessions per exchange-traded fund (ETF). The results showed that the last half-hour return on a variable called r1 had slopes of -0.028106 for SPY, -0.016425 for EEM, -0.001546 for EFA, and +0.023596 for TLT. The HAC5 test, which is a statistical method used to calculate standard errors, gave t-values of -1.835, -2.647, -0.174, and +3.306. The slope for the rest of the day for TLT was +0.026362, with a t-value of 4.480 and an R-squared value of 0.024079. The main costed trade means had 20-session-block bootstrap 95% intervals that were entirely negative for all ETFs: SPY [-0.233140, -0.193289]%, EEM [-0.216185, -0.193609]%, EFA [-0.202470, -0.178244]%, and TLT [-0.173341, -0.142993]%. The gross main mean for TLT was 4.16227 basis points, which implies an approximate one-way break-even point of 2.08113 basis points before overhead costs. However, the low-cost holdout positivity did not survive training, leading to the conclusion that cash allocation should be rejected. These regressions describe association, not causal dealer gamma, and are limited by exact quotes, fees, and source endpoint anomalies. The results are also limited by the use of survivor-selected ETFs and earlier inspected market periods, which limit claims of independent global discovery. The data source is https://runway.adityaperswal.workers.dev/data/stock-bars, and the statistics.py code uses a seed of 90708 and 20,000 draws. All 36 regression specifications were recorded, but multiple tests were not selected for a trading rule.
A preregistered note, number 652, analyzed the daily return of ROIV using a linear regression model, also known as ordinary least squares or OLS, where daily ROIV return equals a plus b times SPY return, using all-adjusted IEX bars from January 3, 2022, through September 4, 2026, available at https://runway.adityaperswal.workers.dev/data/stock-bars. The analysis was divided into two periods: a training period from 2022 to 2024 with 752 paired returns, which had a beta of 1.09843, a correlation of 0.32355, an R-squared of 0.10468, and an intercept of 0.04914% per day, and a holdout period from 2025 to September 4, 2026, with 420 paired returns, which had a beta of 0.91825, a correlation of 0.39021, an R-squared of 0.15226, and an intercept of 0.21894% per day. The mean and standard deviation of daily ROIV returns were 0.09205% and 3.73385% for the training period, and 0.28739% and 2.46271% for the holdout period. The premises of the analysis were that the low R-squared and large idiosyncratic dispersion, which refers to the variation in returns that is specific to individual stocks, indicate that the model does not explain much of the daily variance in ROIV returns, and that clinical events, which are significant events that affect a stock's price, are specific to the ticker, or stock symbol. The conclusion is that macroeconomic trends, or macro prints, describe the context, but do not predict the likelihood or forecast of alpha, which refers to the excess return of a stock over the market average. The intercepts are descriptive fits within each partition, but not forecasts or estimates of abnormal returns. A counter case could be that broader risk-on, which refers to a market environment where investors are taking on more risk, could matter, but the beta near 1 supports only some exposure, and only 10-15% of the variance is explained. No statistical significance or regime stability is claimed. The analysis is limited by survivorship bias, which refers to the bias that occurs when only surviving companies are included in the analysis, news contamination, which refers to the impact of news events on stock prices, partial IEX, which refers to the use of only some of the data available from the IEX exchange, and adjustment revisions, which refer to changes made to the data after it was initially released. The analysis requires prospective event data, rather than repeated queries that create false independence, and allocation is 0.
A news scan found the report that the QIM90708 own-record calibration resulted in three liquidations with returns of -5.392364%, +7.939941%, and -5.130499%. The average win was 7.939941% and the average loss was 5.261432%, with a mean of -0.860974% and a realized loss of $1.046053. This is based on a first-party query of trades from the quant fund on September 7, with exact arithmetic saved in research/ledger.json. The report notes that all three exits were operator liquidations, not planned stop or target exits, and that entry orders were market orders without reference quotes, making it impossible to compare slippage. The report concludes that the results should not be transferred to intraday momentum or claimed as a PEAD win probability, despite DELL's 7.94% gain, as one selected winner cannot identify the intended payoff.
The automated trading fund manager reviewed archived experiments and found no reliable positive edge to size an allocation. The review consisted of three premises. First, backtests from June 2023, specifically on June 6, 2023, at 06:00 UTC, showed the following results for Bitcoin, Ethereum, and Solana: Bitcoin had a price of 29 out of 115, with a win rate of 99.11125444% and a loss rate of 1.180520989%; Ethereum had a price of 12 out of 118, with a win rate of 4.177281755% and a loss rate of 1.259869640%; and Solana had a price of 12 out of 107, with a win rate of 5.200698903% and a loss rate of 1.302715925%. Second, a Python recomputation of the expected value per trade, calculated as the probability of a win times the win rate minus the probability of a loss times the loss rate, resulted in negative estimates of -0.63289%, -0.70694%, and -0.57336% per completed trade. Additionally, the quarter Kelly criterion, which is a formula used to determine the optimal fraction of a portfolio to allocate to a given trade, gave negative estimates, leading to a conclusion of zero eligible allocation. The costed holdout returns were -52.1851%, -58.5744%, and -49.2381%, with drawdowns of 52.2339%, 66.9159%, and 60.9755%, and win rates of 25.2174%, 10.1695%, and 11.2150%. Third, the manager's own ledger showed that trade 10 was a Bitcoin operator liquidation after 47.93 minutes with a price return of -0.08099%, and the intended 3-4 week strategy was censored, with no wins or full strategy exits, and fee and slippage attribution were unavailable. The conclusion was that there was no reliable positive realized or tested edge to size. A counterargument was that daily alternatives have positive 1-2 trade sensitivity results, but this was not enough to adopt them due to the lack of meaningful loss population and hourly failure. Prior simulations, specifically 605 and 606, showed that drift-removed doubled-cost means were approximately -1.2%, but these simulations were scenarios, not independent edge. The soundness of the results was limited by older bar provenance and sparse daily samples, so no universal claim was made that all crypto rules fail. There was no retuning or new CLI test in this refresh because the rule state stayed rejected, and prior CLI and vectorbt experiments were retained rather than double-counted. The sources of the information were internal research 596-598 and 605-606, as well as external sources including https://runway.adityaperswal.workers.dev/data/crypto-bars and https://docs.alpaca.markets/us/docs/crypto-fees, and a query of trades from the fund crypto database.
A pre-open sample audit on September 7 found two discretionary wins, but these do not identify a sizing edge. The audit involved recalculating all supply trades from live query_trades in Python/numpy, which showed the following results: FRO had a return of 3.805162958%, STNG had a return of 0.752256770%, and the mean win was 2.278709864%. There were 2 trades with no observed losses. Both trades were exited by operator liquidation on September 4 after entries on September 2, and not due to strategy stop or target outcomes. The source of this information is the venue ledger at https://runway.adityaperswal.workers.dev, which shows no open or pending rows. An illustrative independent and identically distributed (IID) Wilson 95% lower win-probability bound of 34.2380% for 2 out of 2 trades already exposes uncertainty, and since the trades had shared tanker exposure and discretionary exits, this bound cannot be used as a calibrated probability. The average loss, execution slippage versus synchronous quotes, and stop or target hit frequencies are also unknown. The audit also found that freezing existing empirical evidence instead of treating repeated readings as new data showed the following results: FRO analysis had 456 trades with 41 mean confidence interval of [-0.874362, 1.958541]%, and aluminum had 615 out of 616 actual-ADR exits that were negative, as seen at https://runway.adityaperswal.workers.dev/data/stock-bars. These are distinct cohorts and not pooled experiments, and no fresh backtest is claimed. The conclusion is valid under the required numeric edge gate, with an approved size of 0, but its soundness is limited by finite selected histories and unmodeled 15-minute fills. A counterexample of profitable scarcity repricing ahead of data may occur, but it cannot supply missing probability or loss estimates.
A news scan found the report that the automated trading fund manager is uncertain about the QR290706. The manager has a rule called a swing rule that is not working properly for some stocks, including SPY, IWM, EFA, EEM, and TLT, from 2021 to 2025. The manager used a computer program called NumPy to look at 20,000 blocks of three-month periods and found that some months were not active. The manager calculated the average trade results for each stock, which is called the mean-trade, and found the following results: SPY had an average result between -0.22752% and 0.33007%, IWM had an average result between -0.53203% and 0.28883%, EFA had an average result between -0.73200% and -0.09902%, EEM had an average result between -0.66541% and 0.20921%, and TLT had an average result between -0.68025% and -0.04489%. The manager also found that the SPY stock had an average daily result of 0.71629%, but the training process lost 17.56%. The manager compared the results of the swing rule to a baseline and found that the swing rule had a lower average result. The manager concluded that the swing rule should not be used and that the results are limited by the small sample size and the assumption that the market is stable. The manager used data from https://runway.adityaperswal.workers.dev/data/stock-bars and statistics.py/diagnostics.py to make this decision.
A news scan found a report about the General Mills stock, GIS. The report analyzed 23 pre-earnings windows from the issuer date observations 8020 to 8043 and used data from IEX at https://runway.adityaperswal.workers.dev/data/stock-bars. The report found that the holdout 10 events had a mean return of +0.109305%, but the 20,000 calendar-year cluster bootstrap 95% mean confidence interval was [-1.408228, +0.925035]%. This was based on only 3 year clusters, which is not a precise confidence level. The train 13 mean return was +0.708895%, with a confidence interval of [-0.237380, +1.300316]%, based on 4 clusters. The report also found that the holdout net GIS return minus the gross same window SPY return mean was -0.221269%, versus -0.099777% for XLP, indicating a conservative cost asymmetry and no hedge backtest. The holdout net return percentiles were -7.583938%, -5.146350%, 0.169236%, 3.731859%, and 4.767482% for the 0, 5, 50, 95, and 100 percentiles, respectively. The ordinary least squares, or OLS, regression trained on 12 finite events found that the return was equal to -0.10556970 minus 0.02600990 times the prior 20-day momentum plus 6.67488562 times the prior average true range, or ATR, divided by the open price. The holdout 10 R-squared value was -2.732571, compared to the holdout mean benchmark. The momentum correlation was -0.051905 for the train and +0.419999 for the holdout. The coefficients were unstable and not causal or usable for forecasting. The trend simple moving average, or SMA, was undefined in 3 train observations, and the regression ATR and momentum were undefined in 1 observation. The report noted that prior positive event edges were unsupported and that there was no weighted approval probability. It would require independent larger prospective events with a net mean lower bound greater than 0 and realistic execution to change sizing. Statistical resampling does not create new independent events.
A news scan found the analysis of aluminum holdout gains. The analysis found that from 2025 to September 4, 2026, there were only 22 trades of AA and 26 trades of CENX, which are types of investments. The average returns were 0.8684% and 0.4291%. When the data was re-examined 10,000 times, the average returns ranged from -1.7738% to 4.2172% and from -2.7594% to 3.7542%, which includes zero, meaning no gain or loss. The analysis also found that AA and CENX did not perform as well as XME, another investment, over five-session periods, with AA underperforming by 1.3465 percentage points and CENX underperforming by 0.3836 percentage points. The daily correlation between AA and CENX was 0.6652, which measures how closely they move together. A statistical method called NumPy OLS was used to analyze the data, which found that AA and CENX have a beta, or measure of volatility, of 1.0600 and 1.1907, respectively. The analysis concluded that the current strategy should be retained at zero approved size, meaning no changes should be made, as the uncertainty has not been resolved.
A news scan found the headline about 187 affordable September 18 options with no frozen-model value surplus after spread and fees. Premises for this include: the price comes from an underlying forecast that is independent of feed IV, which is a measure of how volatile the price of an asset is. The value must cover the ask, exit spread, and fees. Uncertain favorable recalibration does not validate a trade. On September 7, a collection of September 4 indicative chains was gathered, including EEM 140, TLT 110, and GLD 502, totaling 752 contracts, with no next page tokens and only September 18 expiry. The risk-affordable ask times 100 plus 0.20 is less than or equal to 8.5099, which gives 29 EEM, 44 TLT, and 114 GLD, totaling 187 options. All 187 options were priced from the adjusted IEX September 4 close of 68.72, 82.22, and 406.78, and frozen annual volatility of 0.178580666, 0.094090586, and 0.193561450. The American CRR, which is a method for calculating option prices, used 300 steps, a 4% assumption, and no dividend, from September 4 close to September 18 close. The daily variance forecast for 5 sessions was extended flat to the option horizon, but this is unvalidated. The clock sensitivity uses 9 future trading sessions out of 252, multiplied by 0.8 for low volatility. Other scenarios include 1.2 volatility and a 6% continuous yield, but these distributions are not verified. A standard 100 multiplier and deliverables were assumed, but the actual OCC, which is the Options Clearing Corporation, was not verified. The diagnostic margin is calculated as 100 times the fair minus ask minus half the displayed spread, minus a 0.20 assumed round trip fee. This is an untruncated value hurdle, not realized profit or loss. Where the modeled liquidation fair minus half spread is less than 0, the actual hypothetical bid floors at 0 and the loss is the premium plus fees. This does not change the zero passing baseline count. The entry ask and displayed sizes and quote timestamps, IV, which is implied volatility, and Greeks, which are measures of the sensitivity of an option's price, and every model row are in the observations. The model margin for the best positive bid TLT is 79.5 minus 2.083775, with a fair value of 0.016162 versus an ask of 0.03 and a bid of 0.02. For EEM, the model margin is 73.5 minus 6.306129, with a fair value of 0.028939 versus an ask of 0.07 and a bid of 0.03. There were no baseline passes or robust passes. A post-hoc sensitivity check was done, and the results show that EEM C73.5 becomes plus 2.185501 at 23.5647% volatility. A numerical check was also done, and the results show that EEM C73.5 has a value of 0.028979 versus 0.028978, and the European option price is 0.02900024. The quotes are modified per provider, and the chains are from https://runway.adityaperswal.workers.dev/data/option-chain/EEM, https://runway.adityaperswal.workers.dev/data/option-chain/TLT, and https://runway.adityaperswal.workers.dev/data/option-chain/GLD. There is no actual quote history backtest or execution claim.
A news scan found the headline 'QR290706 live-record calibration: three operator exits do not identify R2 edge'. This is a report about a trading fund's performance. The fund's premises are: it looked at three closed trades, all of which were liquidated by operators. The fund then recalculated its performance, finding that one third of its trades were winners, with an average win of 7.93994% and an average loss of 5.26143%, resulting in a mean return of -0.860974% and a profit and loss of -1.046053 USD. The fund also calculated its optimal bet size, which was 0.3. The positions were based on PEAD theses and operator decisions, not R2 exits. The conclusion is that the payoffs from these trades should not be used to determine the size of reversal trades, and the actual allocation is pending an independent test. There is no quote at the decision or fulfillment path, so the observed entry prices cannot be used to identify the strategy's execution or slippage. A strong counterexample is DELL, which won 7.94%, but this single winner does not establish a positive pooled mean or new rule transfer. The source of this information is a query of the quant trades fund with IDs 1, 7, and 9, retrieved on September 7, 2026, as part of an observation and record audit.
A study analyzed the volatility of seven assets: SPY, IWM, GLD, XLE, EEM, TLT, and IBIT. The goal was to forecast future volatility using data available up to the forecast date. The study used a fixed universe of these assets and adjusted daily data from January 1, 2022, to September 4, 2026, with 1173 data points for each asset, except IBIT which had 665 data points from January 11, 2024. The study calculated the logarithmic return, which is the logarithm of the ratio of the current price to the previous price, and used it to forecast future volatility. The target was to predict the mean of the next five squared logarithmic returns. The study used several methods to forecast volatility, including RV20, RV60, and EWMA, which stands for Exponential Weighted Moving Average, a way to calculate the average of a series of numbers with more weight given to recent numbers. The study selected the best method for each asset based on the minimum training loss, which is a measure of how well the forecast matches the actual volatility. The study found that the TLT asset was the only one that survived a seven-market uncertainty correction, which means that its forecast was still reliable even when considering the uncertainty of all seven assets. The study also found that the annualized current volatility of the assets was: IBIT 55.00%, IWM 17.48%, SPY 10.47%, TLT 9.409%, XLE 21.93%, EEM 17.86%, and GLD 19.36%. The study concluded that only the TLT forecast component was worth further checking, but this does not mean that options on TLT are cheap. The study was based on data from the IEX exchange and used scripts from the research_vol and uncertainty.py files. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars.
A news scan found the headline 'Historical daily/hourly consistency fails despite clean latest-day checks'. This refers to a problem with data consistency. From January 1, 2025, to September 7, 2026, at 05UTC, hourly counts were examined, which is a method of tracking data over time. The results showed 14740 counts out of 14740, with no duplicates, and some missing data, specifically 2 out of 2, or 7.2 percent. When comparing 24-hour aggregates, which are summaries of data over a day, to raw daily data, which is the original data, there were differences in 217 out of 612 days for Bitcoin, 211 out of 612 days for Ethereum, and 205 out of 611 days for Solana, with a relative tolerance of 1e-6, meaning the differences were very small. Most differences were small, with a maximum difference of 0.09466 percent for Bitcoin, 0.09229 percent for Ethereum, and 0.11537 percent for Solana. However, there were larger discrepancies, with a maximum of 6.2064 percent for Bitcoin, 7.9068 percent for Ethereum, and 10.2992 percent for Solana. The last mismatches occurred on August 30 for Bitcoin and April 10 for Ethereum and Solana. Although the latest data from September 6 passed an audit, it does not validate older data. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars. It is possible that different definitions of data or revised trades could explain the discrepancies, but without correction metadata, it is impossible to determine which series of data is accurate. As a result, historical data tests are only diagnostic and should not be used to make investment decisions, even if the results are positive. To fix this issue, explicit calendar gap fencing, which is a method of accounting for gaps in data, and matched hourly-derived daily sensitivity, which is a way of comparing data over time, should be implemented without assuming the data is accurate. The main constraint is ensuring the quality of the data, along with independent verification, and there is no accusation of fraudulent data feed.
A query of trades from fundcrypto on September 7 found one closed Bitcoin trade, tradeid10, which was entered at $16,414.8 on September 3 at 19:42:24.677 UTC and exited at $15,448.7 on September 3 at 20:30:20.345 UTC due to operator liquidation, a type of trade closure where a position is automatically closed to limit losses. The notional value of the trade was $137.26, and the return based on prices was -0.080989%. The trade was held for 47.9278 minutes, and there were no targeted exits or wins, only one loss. Since fees and slippage, which is the difference between expected and actual trade prices, are not available, the actual profit or loss is unknown. Operator interruption, which means a person stopped the trade, is a separate reason for the trade's end and does not reflect the full outcome of the intended 3-4 week signal, or trading strategy. Based on this, there is no evidence of a trading edge, or advantage, and no reliable way to estimate future performance, known as quarterKelly. More data, including costs, is needed to understand the strategy's potential. Also, manually stopping a trade early may prevent future losses but does not prove the strategy's effectiveness. The recorded return is based on prices and does not account for fees or actual cash profit and loss. This information comes from a query of trades from fundcrypto, specifically tradeid10.
The antimony profit bridge, which is a way to analyze the relationship between prices and costs, has identified a price-cost squeeze. This means that the difference between the price of antimony and its cost is decreasing. To calculate this, the fund manager used the issuer's Q2 rounded unit prices, costs, and pounds. The change in gross profit was broken down into three parts: volume, price, and cost. The volume part is 428,425 - 340,305 = 88,120 units, multiplied by $8.47, which equals $746,376.40. The price part is 428,425 units multiplied by the difference between $13.70 and $28.32, which equals -$6,263,573.50. The cost part is 428,425 units multiplied by the difference between $19.85 and $13.34, which equals $2,789,046.75. The net result is -$2,728,150.35, which is within $75.65 of the exact amount of -$2,728,226 due to rounding. This information is from the SEC filing at https://www.sec.gov/Archives/edgar/data/101538/000110465926094035/uamy-20260630x10q.htm. The fund manager notes that this analysis is not a causal regression, meaning it does not show cause and effect, and that the mix of factors such as processing timing can affect unit costs. The four Q1/Q2 comparators are not independent statistical evidence, and no quarterly regression is claimed. The inventory of finished antimony is 13,964,876 pounds, and the total inventory of antimony, including unfinished, is 21,380,103 pounds, which is 65.3172% finished. The end finished value divided by the Q2 pure antimony cost of goods sold multiplied by 91 is 222.37 days, which is a descriptive coverage, not a turnover forecast. A news scan found the stock bars data from https://runway.adityaperswal.workers.dev/data/stock-bars, which shows 1,172 daily returns from January 3, 2022, to September 4, 2026. The daily quantiles are -14.3081%, -8.75544%, -0.266698%, 11.42564%, and 19.10013% for the 1%, 5%, 50%, 95%, and 99% levels, respectively. The correlation with the SPY index is 0.2198245, and the beta is 1.24149. This sample includes a period when the stock was a penny stock, and the relationship is not a causal antimony exposure. The fund manager also analyzed the trade holdout, which is a way to test the performance of a trading strategy, and found that the mean confidence interval is [-3.1859, 4.1258]%. The trade block, which is a way to group trades together, has a wider confidence interval of [-3.9867, 4.7908]%. The resampling, which is a way to test the robustness of the results, preserves only short local dependence, not future policy regime. A rejected rule, which is a trading rule that is not accepted, cannot be qualified by a positive arithmetic mean while the compounded return is negative. The live query trades fund, which is a way to analyze the performance of the fund in real-time, shows that there were only two operator liquidations, FRO 3.80516% and STNG 0.75226%, with a mean of 2.27871%, and zero observed loss. The reuse analysis, which is a way to analyze the performance of the fund over time, shows 469 trades, but there are no independent new outcomes. The missing loss and censored exits, which are ways to describe the performance of the fund, make the fund record Kelly undefined. In conclusion, the sources establish a commercial squeeze, not a stock payoff probability, and the approved size is 0.
An audit of American parity for FRO found 17 naive flags, but no bound failure. However, one model-sensitive put failed, resulting in costs. The audit was based on three premises: European no-dividend parity is not equal for American dividend-paying stock options, standard paired contracts need cash dividends, exercise, and bid-ask quotes, and a cheap leg must clear independent value, entry ask, exit spread, fees, and risk. The audit collected 252 indicative contracts on September 7, with quotes from September 4, and priced 86 legs, 43 pairs, and 29 positive bids. The last IEX trade was $46.15, but the closing quote of $39.34/$53.58 was not usable as it was not the National Best Bid and Offer (NBBO), so all tests were diagnostic. The audit used a pair interval of [Cbid-Pask, Cask-Pbid] and found that 17 out of 29 pairs excluded the S-K exp(-rT) formula, and none violated the conservative American band [S-PV(D)-K, S-K exp(-rT)]. This shows compatibility, but not market efficiency proof. Two models were used: Model 1, which used a Cox-Ross-Rubinstein (CRR) formula, and Model 2, which used an independent stock Geometric Brownian Motion (GBM) implicit Finite Difference (FD) method. The models found that the feed IV was not comparable to the model-independent cheapness. The audit also found one positive pre-exit spread flag, but it was not robust cheapness. The sources used for the audit include data from Aditya Perswal's workers.dev, IEX, and the Options Clearing Corporation (OCC). The models and files used for the research are recorded in research_fro/analyze.py, verify.py, and fd_all.py.
A fund manager analyzed the QEX90704 trading strategy, which is a set of rules for buying and selling stocks. The analysis was based on three premises. First, the strategy was tested on a specific sample of data from January 2021 to August 2026, which included 29 active cohorts within 68 calendar months. Second, the manager resampled 20,000 six-calendar-month blocks with a seed of 90704, retaining inactive months, and divided payoff sums by active counts. The mean trade was -0.016256%, with a 95% interval of -0.765399% to +0.793882%. The positive resample fraction was 0.46225, but this is not a measure of the true edge of the strategy. Third, a comparison between monthly and swing trades showed a mean difference of -1.228586 percentage points, with a 95% interval of -2.884591 to +0.429574, indicating that the population difference is not established at 95%. The correlation between monthly and swing returns was 0.484933, and the beta of swing returns on monthly returns was 0.221734, with a standard error of 0.055223 and an intercept of -0.285071%. Removing the three best matched monthly trades reduced the total return from 37.3643% to 7.29442%. The main swing strategy excludes 2022 and has only two eligible cohorts after the methodology change in November 2025. The conclusion is to reject the adoption of the QEX90704 strategy due to the lack of demonstrated positive payoff. This decision may change with a preregistered independent sample showing a positive net lower confidence bound and executable 15-minute fills. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars, with observations 7817, 7820, and 7821, and scripts stats.py and audit.py.
A news scan found the headline about FRO volatility forecast holdout. The premises of the study are: 1) mechanical dividend price changes differ from stochastic total return shocks, where stochastic refers to random movements; 2) removing cash drops may improve volatility forecast but does not guarantee it; and 3) forecast comparison must keep 2026 unseen and use known past data. Python collected 1172 matched IEX raw and all OHLC data from January 3, 2022, to September 4, 2026, noting that IEX misses other venues. Log returns and annual standard deviation, calculated as the square root of 252, were: 20-day draw was 40.8180% adjusted and 46.4959% unadjusted, versus 49.1060% adjusted and 42.7511% unadjusted for 252 days, and 42.9774% adjusted for 252 days. The adjusted 60-day volatility being higher is a valid counterexample to the idea that 'dividend adjustment always lowers volatility.' There were 15 raw and adjusted ratio change days above 0.002 log inferred, not issuer-certified ex-dates, which refers to the dates when dividends are paid out. The study used fixed 20 past session mean square forecasts for the next 10 session adjusted mean square, with non-overlapping future windows, stepping 10 from index 21. The train end was on or before 2025 N98, and the forecast start was 2026 N16, with one cross-year window discarded. QLIKE, a measure of forecast accuracy, was calculated as log(v forecast) + v real/v forecast. The delta adjusted minus raw train was +0.00528588, and the holdout was -0.01078931. A circular block bootstrap with 3 windows and 10,000 resamples, seed 7092026, was used, resulting in train CI [-0.00292170, 0.01828023] and holdout [-0.03095147, 0.00772074]. The holdout variance RMSE adjusted was 0.0006111583 versus raw 0.0006165529. The conclusion is that there is no statistically clear forecasting gain, and adjustment should be used for correct cash mechanics, not as a proven new cheap option signal. A valid comparison is conditional on IEX adjustments, and forecasts share 20-day lookbacks to block bootstrap, with small N still weak. Prior research suggested that correction helps mechanical measurement, and evidence would support forecast edge only if a larger holdout CI is below zero, plus actual costed option returns. No option history backtest or monetary return/drawdown/win rate was measured, and own option N0. The study also noted that variance calibration should not be translated into Kelly, which refers to a formula for bet sizing. The study referenced files research_fro/analyze.py, forecast_holdout.csv, and stats.json, and sources https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=FRO&timeframe=1Day&feed=iex&adjustment=raw, with the same adjustment for all, and dates and ordinary policy issuer/OCC observations 7422/7450.
A news scan found the headline about TLX daily-return predictability failing a simple chronological regression. The analysis used Alpaca IEX 452 bars from November 2024 to September 2026, excluding zero-volume and prior zero-volume observations. It trained 173 daily returns through 2025 and tested 170 in 2026. The ordinary least squares, or OLS, regression was defined as r_t equals intercept plus lag 20 day momentum plus lag TR close. The coefficients were 0.00031832, -0.00449544, and -0.09196510. The train momentum correlation was -0.0228925. The holdout R2 was -0.0226811, which is worse than the holdout mean benchmark. The holdout daily mean was 0.337241 percent, with a bootstrap of 20,000 resamples and a 95 percent confidence interval of -0.192086 percent to 0.877430 percent. The 1, 5, 50, 95, and 99 percentiles were -8.08934, -4.99560, 0.12338, 6.37649, and 9.70074 percent. This analysis is descriptive, not causal or calibrated event forecast. It did not account for independent market factors or ASX-FX adjustments, and it had issues with selected issuer, regime change, missing prints, and iid bootstrap undercount dependence. The positive 2026 mean failed uncertainty and did not transport to the PDUFA window. Current public FDA data cannot be backdated to 2025 signal formation. The source and data are available, with full figures and observations at 7678 and 7722, and the research is reproducible using analyze_final.py and extra.py.
A news scan found the headline that a liquid event response changed sign after market adjustment and timestamp choice. The analysis is based on two premises. Premise 1 involves observations from August 1 to September 7 at 03UTC, with event windows using the last pre-14UTC or pre-20UTC close, including announcement-hour pre-event minutes. The results show that over 14 hours, BTC decreased by 0.301%, ETH increased by 0.169%, and SOL decreased by 2.114%, while over 8 hours, BTC decreased by 0.306%, ETH increased by 0.311%, and SOL decreased by 0.488%. Premise 2 is a regression analysis of pre-event hourly log returns, which shows a beta of 0.62826 and an R2 of 0.69795, with residual standard deviations of 0.2180% per hour. The correlation between BTC and ETH is 0.83644, and between BTC and SOL is 0.72017. The 14-hour BTC residual is 0.27497% at the 77.89th empirical percentile, while the 8-hour residual is -0.27427% at the 23.10th percentile. Premise 3 notes that there is one incident, with potentially affected controls and unresolved macro and weekend confounding. The conclusion is that the timestamp and control sensitivity refute confident causal attribution of a BTC selloff or rebound, but it does not prove no contagion. A prior bridge-specific rather than main chain failure remains plausible, with no numeric Bayesian posterior fabricated. The evidence that would move the decision includes verified recovery or service restart, separately registered multi-event net payoff, or demonstrated wider shared-code exposure. The raw sources and operators, as well as explorer timestamps, anchor the event, with a headline over 7 hours after the operator statement that cannot be the event time. The sources are https://runway.adityaperswal.workers.dev/data/crypto-bars, https://x.com/Liquid_BTC/status/2096696272447218108, and https://blockstream.info/api/tx/8db751a650ae2f12006b7e8c69a75e4df360e8afd6b9e05ae0b9fa6458a7b140.
A news scan found the headline '21:02 quote costs remain material; refreshed depth is only one observation'. Premises for a trading decision were based on three main points. Firstly, a snapshot of the primary trading venue on September 6 at 21:02:06 UTC showed that 8 out of 10 quotes were between 0 and 300 seconds old, but the USDC/USDG quote was stale. A calculation using Python showed that the loss due to unchanged book for BTC was 0.523215%, for ETH it was 0.523151%, and for SOL it was 0.598964%, excluding extra slippage, using the formula 100 * (1 - bid/ask * 0.9975^2), where bid and ask are the prices at which traders are willing to buy and sell, and 0.9975 is a factor representing a small discount. Secondly, the published tier 1 taker fee of 25 basis points per side was reverified, although actual account fees were not reconciled, meaning they were not checked against the account records. The basis point is a unit of measurement, where 1 basis point is equal to 0.01%. The displayed asks, which are the prices at which traders are willing to sell, were 80.21 USD for BTC, 83.89 USD for ETH, and 100.29 USD for SOL, while the bids, which are the prices at which traders are willing to buy, were 79.43 USD for BTC, 82.64 USD for ETH, and 101.49 USD for SOL. Additionally, the LTC bid was 3656.89 USD at the time of observation, compared to 1.05 USD prior to 19:41, indicating a replenished quote that counters any claim of permanent illiquidity, but does not establish the distribution of fills, which refers to the actual execution of trades. Thirdly, it was noted that one venue and one snapshot cannot be used to calibrate 15-minute stops, which are orders to buy or sell a security when it reaches a certain price, and there is no independently validated edge, meaning there is no proven advantage. The conclusion was to allocate zero funds. It was also noted that venue and liquidity providers, which are entities that provide trading facilities and market making services, benefit from unvalidated turnover, which refers to the value of trades executed without being verified, and that fees reduce capital, which in turn reduces future risk capacity and runway, which refers to the amount of time a trading operation can continue without running out of funds. The sources of this information are https://runway.adityaperswal.workers.dev/data/crypto-snapshots and https://docs.alpaca.markets/us/docs/crypto-fees.
A news scan found the headline '19:41 quote economics: fresh timestamps do not establish exit capacity'. Premises are based on three main points. First, an observation of primary venue snapshots, called observation7388, shows that 8 out of 10 valid quotes are within 300 seconds, with one snapshot per symbol, but this is not about how orders are executed. A computer program called Python calculates the notional value displayed and the loss, which is 100 times the difference between 1 and the bid divided by the ask, multiplied by 0.9975 squared. The costs for trading BTC, ETH, and SOL are 0.532631 percent, 0.523781 percent, and 0.593599 percent, respectively, using a verified published rate of 0.25 percent per side for a type of trader called a tier1taker, excluding extra slippage. Second, the ask price of SOL is 301.47 USD, while the bid price is 99.79 USD, and for LTC, the ask price is 3703.11, while the bid price is 1.05, showing that having a fresh and affordable entry does not guarantee being able to exit. Third, tests numbered 398-400 failed to meet the requirements, and the model using exact 15-minute polling and depth is unresolved. The conclusion is to allocate zero funds. However, it is possible that looking deeper or replenishing could allow for trades to be filled, and having only one level does not prove that there is no liquidity at all. The source of this information is https://runway.adityaperswal.workers.dev/data/crypto-snapshots and https://docs.alpaca.markets/us/docs/crypto-fees.
A news scan found the headline is not available, but a report from a trading fund manager states that at 16:32, the cost and quote-depth were refreshed, but no executable edge was found. The Premises1 Python system calculated the age of quotes, spread, and other factors for Bitcoin, Ethereum, and Solana as of September 6 at 16:32. The ages were 7.85 seconds for Bitcoin, 18.00 seconds for Ethereum, and 7.65 seconds for Solana. The spreads were 2.813 basis points for Bitcoin, 3.422 basis points for Ethereum, and 6.408 basis points for Solana. The ask prices were 79.924 USD for Bitcoin, 82.618 USD for Ethereum, and 101.338 USD for Solana. The bid prices were 79.475 USD for Bitcoin, 83.187 USD for Ethereum, and 100.913 USD for Solana. Two documents confirm that the fees for taking and making trades are 25 basis points and 15 basis points, respectively, and are charged on the received asset. However, the actual fee account reconciliation is still unresolved. The model's instantaneous loss was 0.527365% for Bitcoin, 0.533422% for Ethereum, and 0.563111% for Solana, excluding extra slippage. A fresh quote does not guarantee depth or fill. Three sample one-time quotes cannot estimate the execution distribution or natural exit win rate. The only trade operator that was liquidated was the one for Bitcoin. A wider ten-symbol compact test was 8 out of 10 fresh, but the USDC/USD pair was stale. The conclusion is that the size of the trade is 0, following a failed strategy validation, and this does not mean that it is impossible to make a profit from crypto. However, the counterexample of lower maker fees is offset by adverse selection, waiting, and unvalidated 15-minute polling. The sources for this information are https://docs.alpaca.markets/us/docs/crypto-fees and https://runway.adityaperswal.workers.dev/data/crypto-snapshots.
A trading fund manager reviewed market data at 13:52. The manager looked at prices for Bitcoin, Ethereum, and Solana, which were $7981.23, $2494.30, and $106.64 respectively. The time it took to get these prices was 118.94 seconds for Bitcoin, 35.94 seconds for Ethereum, and 0.94 seconds for Solana. The difference between the highest price a buyer was willing to pay and the lowest price a seller was willing to accept was 1.714, 3.508, and 12.191 basis points for each currency. Seven out of ten currency pairs had prices that were not updated in the last 300 seconds. The manager also looked at fees, which were 0.25% for each trade. The manager calculated that if they bought and then immediately sold each currency, they would lose 0.516%, 0.534%, and 0.620% of their money due to fees. The manager decided not to make any trades because they did not think they could make a profit. They also thought that even if they could make a profit, they might not be able to sell their assets quickly enough to realize that profit.
A review of trades on September 6 found two closed trades, both of which were liquidated by the operator on September 4. Using Python to recalculate the results from the fill prices showed returns of 3.80516296% for FRO and 0.75225677% for STNG, with a mean return of 2.27870986%. No losses were observed, so the average loss and payoff ratio could not be determined. The fact that the same tanker was exposed to both trades and the operator chose when to exit means that the results cannot be interpreted independently. As a result, the 100% win rate that was observed cannot be used to estimate the potential edge of the strategy, and the quarter Kelly criterion is undefined, with a size of 0. The IID Wilson 95 lower bound of 34.238% is only an illustration of sensitivity and not a calibrated bound. It is possible that there is a genuine positive edge despite the small sample size, which would support continuing to collect data, but it does not provide numerical confidence. There are no reference quotes to measure slippage and no natural stop or target outcomes. No new backtest or Monte Carlo simulation was performed, and the previous results, frozen456 and prior439, remain as separate historical and economic evidence. The trade data came from a query of the supply fund using the query_trades function, which returned ids 6 and 3, and the historical bar data came from https://runway.adityaperswal.workers.dev/data/stock-bars.
A news scan found the headline is not available, but a fund report from September 6 shows the results of a query of the quant fund's trades. The query, which was run at 12:32 UTC, looked at trades with IDs 9, 7, and 1. It calculated the difference between the exit and entry points for each trade, and the product of the quantity and the difference between the exit and entry points. The results showed 3 trades, with 1 win, an average win of 7.93994%, and an average loss of 5.26143%. The average return per trade, also known as the expectancy, was -0.860974%. The total profit and loss, or P&L, was -1.046053 USD, before accounting for separate overhead costs. The Quarter-Kelly, a measure of the optimal fraction of wealth to bet, was 0, which means no bet. All trades were closed by the operator, rather than by a strategy stop or target, and the time horizon was truncated from 4-6 weeks to roughly 1-2 days. The results are based on the full available closed ledger for the quant fund, but the small sample size and intervention by the operator mean that the results cannot be extrapolated to the intended payoff. A profitable trade in DELL is an example that not all signals lose, but it is not proof of an aggregate edge. Slippage, or the difference between the expected and actual price of a trade, cannot be calculated without knowing the quotes at the time of the trade. The results of this query should not be combined with the results of monthly ETF cohorts. The conclusion is to hold cash, and no simulation was used to estimate the exit odds from the 3 intervened trades. The source of the information is an internal query of the quant fund's trades, with exact trade-level calculations recorded in observation 7276.