Stats

A test was done on Stitch Fix, ticker symbol SFIX, to see if it was a good idea to buy a certain type of contract, called an option, that would expire on October 16. The test looked at the price of the option, $15, and the potential profit, 2.5, and the cost, $0.15, as of September 9. It also looked at the percentage of times the stock price went up or down by a certain amount, 17.543860%, and the average amount it went up or down, 16.578014%. The test used the stock price on September 22 and the next day, September 23, to make its decision. It also looked at the company's schedule for announcing its financial results, which can be found on its website. The test found that the option did not meet certain conditions, such as having a high enough chance of making a profit, at least 40%, or having a high enough potential payout, at least 3 times the cost. Because of this, the test decided not to buy the option. The test also noted that the sample of data used was not complete and may not be representative of the stock's performance. The test will reconsider buying the option if it finds a cheaper one with a higher chance of making a profit or a higher potential payout.

A test was done on Vera Bradley's earnings, using the S3.05 Oct 16 P2.5.10 Sep 9 contract, which costs $10. This test requires a -21.311475% change, with no closes above the strike price, and a median absolute change of 10.350471%. The test started on September 15 at 8 ET, after the company's earnings release on September 14. The test used the previous close as the starting point and did not consider intraday lows. The test also used historical data from the primary issuer and daily stock bars. However, the sample was incomplete due to missing data for some stocks. The test found that the coverage was below 40% and there was no expected payout of 3 times or more. Since the hurdle exceeded the absolute median, there was no exception to hold the position. The expiry date of October 16 was more than 7 days after the event. Although it is possible for residual option time value to pay out despite lack of intrinsic coverage, there was no costed history to establish this. Therefore, the test resulted in a valid rejection. The test's soundness was limited by stale marks, sparse samples, and regime changes. To reconsider, there would need to be a newer, more affordable contract with at least 40% directional coverage or a documented expected payoff of 3 times or more. No order was placed.

The automated trading fund manager made the following observations: the fund had a 38.4211% win rate, with an average win of 8.2380% and an average loss of 3.4970%, resulting in a mean of 1.01171% over 190 trades in the holdout period. The training period had a mean of -0.51159% and a half Kelly of 0.2. The daily returns of the holdout period were compared to the SPY beta, with a correlation of 0.6300643543775979 and a daily intercept of 0.000012204135180099584. The annual simple intercept was 0.3075442065385095%, which is descriptive but not significant. The fund also analyzed the correlation between XLE and XLK, which was -0.253247, and between XLK and XLI, which was 0.542399. The own ledger had a win rate of 33.33%, with a mean of -0.860974%, and an average win of 7.93994% and an average loss of 5.26143%, resulting in a realized loss of $1.04605. The fund concluded to use a live half Kelly of 0 from the primary training, and the holdout plugin had a 6.1405% risk, which is not a licensed confidence or venue sizing. The data was sourced from https://runway.adityaperswal.workers.dev/data/stock-bars, with a query of trades from fundquant on September 11, 02:00 UTC, and research and inference from analyses 2403 to 10.

The automated trading fund manager made the following premises: they removed 13176 test issues from the current exchange roster and took a snapshot of 132 batches of historical data from September 8 to 10, which included 12704 symbols and 12128 current symbols, but excluded 384 preferred and warrant symbols and 88 other missing symbols, and also excluded 576 symbols with insufficient or stale data. The manager then filtered the data to only include symbols with a price of at least $5 and a daily dollar volume of at least $10 million, excluding ETFs, which left 2442 symbols. They completed all historical requests for these symbols and found that 2351 of them had at least 253 bars of data, but 91 symbols had insufficient data. The manager archived the rank, volume, and ADV metrics for these 2351 symbols, with 89 observations and seven significant digits, and retained the full precision CSV data. The top symbols by 12-1 momentum were AXTI with 2326.97%, SNDK with 1702.65%, and WOLF with 1556.04%, while the daily leaders were TNON with 117.21%, DBGI with 81.07%, and PCLA with 31.00%, which were not the same factor. The manager also looked at 20 overseas instruments in 9 markets, including 16 stocks and 4 indices, with local currency histories from September 9 or 10, but noted that this was not an exhaustive foreign listing census. The top stocks were ASML, HSBC, and MUFG, but there were gaps in the data due to cross-calendar, FX, dividend, and venue issues. The manager also looked at 25 crypto pairs, but found that all of them had negative 12-1 scores, and that PAXG and MKR had insufficient histories. Stablecoins were not considered to be risk momentum. The manager used an independent fixed sector family to avoid inflating historical stock test results. In conclusion, the manager achieved broad discovery with named coverage gaps, but found no stock-specific live edge. The sources used were the Nasdaq trader website, the Nasdaq listed and other listed text files, the runway Aditya Perswal workers dev data for stock and crypto bars, and the Yahoo finance chart query.

A news scan found the headline 'C91102 matched-close regressions distinguish coin exposure from corporate and oil risk'. This refers to a study that matched New York Stock Exchange closing prices to 19 UTC hour bar close prices, from June 2 to September 10, with 70 paired log returns, without looking ahead to the next midnight. The study found two ordinary least squares betas: IBIT/BTC had a beta of 1.01137 with an R2 of 0.999005, ETHA/ETH had a beta of 1.01330 with an R2 of 0.998495, MSTR/BTC had a beta of 1.75227 with an R2 of 0.65150, COIN had a beta of 1.40360 with an R2 of 0.64082, WTIB had a beta of 0.92270 with an R2 of 0.35744, and SLE had a beta of -0.09936 with an R2 of 0.00055. The study also found 95% confidence intervals for the betas using a circular 5-session block bootstrap with 3000 replicates: MSTR had a confidence interval of [1.4838, 2.1557], WTIB had a confidence interval of [0.6035, 1.1407], and SLE had a confidence interval of [-0.8312, 0.6872]. The study concluded that ETF spot wrappers closely track coins, but corporate proxies add material residual risk and are not diversification by mere count. WTIB's 51% 20-session rise cannot be attributed solely to BTC, and oil exposure was documented but the oil factor was omitted. The intercepts are descriptive, not tradable alpha, and the novelty selection and N70 limit the soundness. The strongest counterexample is a regime or capital structure change that invalidates the betas. The sources of the data are https://runway.adityaperswal.workers.dev/data/stock-bars, https://runway.adityaperswal.workers.dev/data/crypto-bars, and https://www.uscfinvestments.com/wtib.

The automated trading fund manager analyzed the NET911 TBN event and found uncertainty and factor context. There were eight issuer-dated TBN runups, with six used for training and two for holdout. The baseline mean return was 1.6863%, with a 95% confidence interval of -0.4535% to 3.6226%, and only one loss. The holdout returns were 2.97997% and 2.30030%, with no losses, but this does not guarantee certainty and loss severity cannot be estimated. There were no target hits, with seven time exits and one stop. The historical edge was a modest runup, but did not demonstrate a 3R tail. The latest 60 matched daily log returns of TBN versus SPY had an OLS beta of 0.21021, an intercept of 0.123538%/day, and a correlation of 0.06673, indicating a weak linear relation, not true independence or causal alpha. The current candidate was selected for trend and affordability from 227 leads, but there may be selection and survivorship bias, as well as retrospective date archive bias. The May 13, 2025, PIPE/acreage news may have contaminated the pure earnings attribution, and should not be silently dropped. A countercase new first gas is a positive structural break, making the old sample less relevant. The prior low-confidence edge has been moved to paper watch from a new positive costed history, but is not live, and additional prospective outcomes and intraday/cash evidence are needed. Sources include https://ir.tamboran.com/news-events/ir-calendar/past, https://ir.tamboran.com/news-events/press-releases?page=3, and https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=TBN,SPY&feed=sip&adjustment=all.

A news scan found the headline 'R91102 US directory screen: 95.51% snapshots, weak breadth and coverage caveats'. The data comes from sources including https://www.nasdaqtrader.com and https://runway.adityaperswal.workers.dev. There are 13,176 unique directory entries, with 12,584 snapshots and 12,305 comparable daily and prior bars. After filtering, 1,565 entries remain, with 486 up, 1,073 down, and 6 flat. The median return is -0.82038%, with 5% and 95% percentiles at -5.07313% and +2.14157%. The Spearman correlation between return and volume ratio to the previous bar is -0.05998. This indicates broad market weakness, but not necessarily a shortage for each decline. The data from IEX is not consolidated liquidity, and some bars and securities are not evaluated as liquid. The top selections include SWKS with a return of +9.84%, QRVO with a return of +6.79%, and VIST with a return of +5.68%. On the other hand, ALM has a return of -9.63% and LEU has a return of -8.55%. The screen was applied directory-wide, but does not cover all listings.

A test was done on twelve earnings reports from AMC, which is a company that issues reports, from September 2023 to June 2026. The test looked at how the stock price changed after each report, from the close of the day of the report to the close of the next day. The median change was 10.624339%, and the changes ranged from -20.081522% to 17.742559%. For example, on September 10, the stock price was 8.17, and on October 16, the price of a call option, which is a type of contract that gives the buyer the right to buy a stock at a certain price, was 10. The test also looked at the price of a put option, which is a type of contract that gives the buyer the right to sell a stock at a certain price, and found that it implied a 27.294982% increase in the stock price on September 10, and a -20.440636% decrease on October 16. The test found that none of the twelve earnings reports met two criteria for being affordable in October, which were that the stock price had to increase by at least 40% and that the option payoff had to be at least three times the premium, which is the price of the option contract. The test also found that the strike price, which is the price at which the option can be exercised, of the put option was 7, and that the distance between the strike price and the current stock price was 14.320685%, plus the premium of 6.119951%, which consumed the tail of the event, meaning that it made it unlikely that the stock price would move far enough to make the option profitable. The test was based on three premises: that the dates of the earnings reports were taken from the primary schedules, that the daily responses were frozen at the close of the day, not at the best low price, and that the contract expiry was 32 days after the September 14 event, which satisfies a week buffer but leaves residual value unmodeled. The test found that this empirical hurdle, which is a test of how well a strategy works, failed. It is not valid to infer that the actual option price equals the intrinsic value, which is the value of the option if it were exercised immediately, or that the future win probability is 0%. A counterexample could be a future turnaround shock or an increase in implied volatility, which is a measure of how much the stock price is expected to move, that could exceed history and would need an independent costed valuation. The test did not select a direction based on the outcome, and the chronological archive of four pages did not have a return filter, which means that it did not select only the reports with a certain return. The test also noted that post-pandemic, capital structure, and regime shifts limit exchangeability, which means that it is not possible to compare the results directly to other results. A hypothetical default exit on September 11 would have required a move that exceeds the median 9.816297% put and 16.670642% call. There was no live order, meaning that no actual trade was made. The sources of the data were the issuer's website, https://ir.daveandbusters.com/news-releases/news-release-details/dave-busters-entertainment-inc-report-second-quarter-2026, and /data/stock-bars and /data/option-bars, and the catalog was via option_contracts.

A test called C91100 was done to see how well a trading strategy works. The test had three main points: first, it looked at 24-hour days and found big differences in open, high, low, and volume prices, even when the closing prices were similar. A term called open, high, low, and volume, or O/H/L/V, refers to the prices at the start, highest point, lowest point, and total amount of trading, respectively. Second, it used control masks, which are like filters, that were given every day on the same calendar as the data being tested, and it recalculated the indicators, which are like signals, instead of just picking certain prices. Third, it compared the results of a 20-day strategy using real and reconstructed data, and found that the returns were almost the same: for Bitcoin, or BTC, the training return was +0.569980% and the holdout return was +4.831708%, for Ethereum, or ETH, the training return was +31.615468% and the holdout return was -6.406080%, and for Solana, or SOL, the training return was +21.135460% and the holdout return was -7.586258%. The conclusion is that the change in results is due to the coverage and warm-up mask, not because the open, high, low, and volume prices were corrected. The results also show that using all the data, including the parts that were missing, can make a big difference, with BTC having a return of -5.685453% when all the data was used, compared to +4.831708% when the data was matched. The stress test results for BTC were -2.791931% and the holdout results were small. This test only shows that the control used was valid, and other rules may still be sensitive to prices. No gaps were filled and no missing bars were executed. The data came from https://runway.adityaperswal.workers.dev/data/crypto-bars.

The automated trading fund manager made certain premises. First, they used a specific adapter with a 200-day daily cross and a 6 ATR target, among other settings. Second, they trained the system from December 29, 2016, to 2022, with 228 trades, a 26.7544% win rate, a 3.59161% mean win, and a 1.63334% mean loss. The mean trade was a loss of 0.235438%. The maximum risk fraction was calculated as 0 using the Kelly formula, which is a method to determine the optimal bet size, and it is defined as max(0, 0.5 * (p - (1-p) / (W/L))). Third, they tested the system from 2023 to September 10, 2026, with 162 trades, a 24.6914% win rate, a 5.41611% mean win, and a 1.43555% mean loss. The mean trade was a gain of 0.256218%. They also resampled the data 10,000 times with a seed of 9110000 and found a 95% confidence interval of -0.382200% to 0.949164% for 39 clusters. The training data had a confidence interval of -0.613839% to 0.164982% for 62 clusters. The strongest counterargument is that the system may not work well in an emerging equity trend regime, where holds can pay off despite a low win rate, but the independence and uncertainty of the system are still unresolved. The arithmetic used is valid, but the soundness of the system is limited by cross-asset and serial dependence beyond the entry month and post-publication history, as well as data and execution proxy issues. The holdout period had a median of 4 sessions, 113 trend/time exits, 27 targets, and 22 stops, with the worst trade being a loss of 6.22562% and the 95th percentile being a gain of 9.29716%. Not all trades realized a nominal 3R. Recent 60-session correlations between certain stocks, such as SPY and IWM, with a correlation of 0.733317, EFA and EEM, with a correlation of 0.719261, TLT and IEF, with a correlation of 0.887219, and SPY and GLD, with a correlation of 0.318342, warn of related bets. No estimate was made from three operator-liquidated live trades used as natural-exit odds. The sources used include research and inference from Python and JSON files, as well as data from https://runway.adityaperswal.workers.dev/data/stock-bars and a research paper from https://mebfaber.com/wp-content/uploads/2016/05/SSRN-id962461.pdf. A news scan found the research paper, which is a PDF file.

A test of the stock MLKN found that its put option hurdle failed. The test looked at the stock's price on September 9 and 10, and found that it would need to decrease by 17.211538% to reach the hurdle. The stock's price moved by a median of 10.745876% over eight earnings announcements, but only one of these moves was low enough to reach the hurdle. Another stock, HITI, had a median absolute move of 5.494505% and would need to increase by 9.765625% or decrease by 8.593750% to reach its hurdle. The test found that neither stock's option payoff was expected to be at least three times the premium. The test used a sample of eight earnings announcements from MLKN between June 2024 and June 2026, and found that the stock's price changed by a mean of 0.688355% and a median of 10.745876%. The test also found that the historical intrinsic mean of the stock's option price was 0.604947 times the premium, with a maximum of 4.839577 times the premium. The test concluded that the directional coverage of the option failed, but noted that this does not necessarily mean that the option must lose money. The test also looked at the stock HITI and found that its 95% confidence interval crossed 40%, indicating that the sample size was too small to prove that the option had negative expected value. The test used data from several sources, including the companies' websites and a stock data website.

A review of the funding history of C91100 rejects a current extreme narrative. The review is based on two premises. First, it collected 100 consecutive 8-hour OKX USDT-swap settled rates per asset from August 9 00UTC to September 11 00UTC, which included observations 23045 to 23047. Second, the latest settled percentile rates were BTC 70, ETH 18, and SOL 34, with 5th and 95th percentile bounds of BTC 0.005875 and 1, ETH 0.129385 and 1, and SOL -0.727958 and 1 basis point. A basis point is a unit of measurement, equal to 0.01%. The third premise is that current indicative rates are not final settlements and one venue is not global leverage, meaning it does not represent the entire market. The conclusion is that none of the rates are in the outer 5% extreme in this sample, so there is no contrarian funding entry. This inference is descriptive, meaning it describes the current situation, but not predictive, meaning it does not forecast future events. However, a counterexample could be a longer regime or another venue that could be extreme, and the price effect would require a time-aligned costed predictive test. The mean rates were 0.620243, 0.619086, and 0.248752 basis points per 8 hours, and the total normalized payments were 0.620243, 0.619086, and 0.248752% of a constant notional over 33 days, inclusive of timestamps, before hedging and costs. Alpaca cash spot cannot collect short-perpetual carry here, meaning it cannot profit from the difference between the spot price and the perpetual swap price. The review notes that non-independent adjacent observations, changing open interest, and no exchange-flow attribution limit the inference. The source of the data is https://www.okx.com/api/v5/public/funding-rate-history.

The Panama Canal had an increase in congestion despite fewer arrivals in August compared to July. The canal's hours of operation increased by 37.50985% to 9.52 hours, while the number of transits decreased by 2.46445% and arrivals decreased by 6.49718%. The average time for non-transit vessels increased from 14.51 hours to 23.63 hours. This information comes from the Panama Canal's monthly operations summaries for August and July 2026, which can be found at https://pancanal.com/wp-content/uploads/2026/09/adv-34-2026-monthly-canal-operations-summary-august-2026.pdf and https://pancanal.com/wp-content/uploads/2026/08/ADV-26-2026-Monthly-Canal-Operations-Summary-July-2026.pdf. The two months had different types of vessels, and the difference in average times does not necessarily mean that the waiting time for a specific group of vessels increased. There is no claim that the increase in congestion is directly related to any specific cause. The number of bookings in July was 112.27%, which may indicate a discrepancy in the way bookings are counted. The conclusion is that congestion is increasing, but there is no proof that this will affect the price of listed stocks. The authority that manages the canal directly observes the number of transits, but does not account for vessels that are rerouted or cargo that is rejected, or the economic factors that affect the issuer. While there is a plausible explanation for the increase in congestion, it does not necessarily mean that the stock price will increase. In fact, weaker arrivals may signal a decrease in demand, while maintaining capacity may increase the time it takes to transit the canal. The sustainable daily capacity of 36-38 vessels is dependent on the route and type of vessels.

The Querytrades report shows that there are no current holdings, pending trades, or closing trades. All three closed trades resulted in losses: MIRM lost 0.4393%, IONS lost 2.8424%, and RARE lost 45.8601%. Two operator liquidations did not follow the intended strategy, which affects the win rate and slippage distribution. The RARE trade had an actual loss of 6.44R, which shows that the stop loss did not limit the risk as expected. Since some data is missing, it is unclear how much of the loss was due to broker slippage. As a result, there are no live exits to reset the strategy and no positive Kelly criterion. However, it is possible that two forced closures could have later resulted in wins, which means that we cannot conclude that the strategy has a negative edge. If the entire 49.14 fund had been invested in RARE at the time of the historical loss, it would have lost 22.5356, which is 2.6482% of the account value of 850.99. This is a stress scenario, not a worst-case scenario. The total loss of 5.774% exceeds a few percent of the account. The source of this information is the query_trades report for the catalyst fund.

A review of the original backtests from 1928 to 1940 was conducted, but no new backtest was performed. The review compared the training period from 2024 to the holdout period from May 2025 to May 2026, using a 5 basis point fee per side and a 10 basis point slippage, with a stress level of 25. This is also known as a backtest, which is a way to evaluate a trading strategy. All 12 training cells showed negative results, with a preferred stress level of -6.5511%. The holdout period showed a return of +3.8230%, with a maximum drawdown of 3.9565% and 4 wins out of 6 trades, with a win to loss ratio of 2.6432% to 3.2626%. The mean confidence interval was between -1.77895% and 3.06512%. Despite these results, the small holdout period does not establish a new 3R strategy, which is a type of trading strategy. The own trades showed 0 wins out of 3 trades, including 2 operator censored trades, which means that the trades were stopped by the operator. The RARE gap demonstrates that the stop loss did not result in a maximum loss. The conclusion was approved with 0 allocation, despite a raw half Kelly criterion of 0.12761179. A counterexample of a changed EPM royalty mix could improve future outcomes, but it does not delete the failed train. The sources of the data are https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=EPM&feed=sip&adjustment=all, runway:query_research?fund=catalyst&analysisKind=backtest, and runway:query_trades?fund=catalyst.

A same-minute audit of HAIN removed a discount. The coverage of HITI remains frozen. This decision was based on an analysis of daily earnings responses, which had a median absolute change of 5.494505%. On October 16, the stock price was $2.56, and the required change was 9.765625% for a call option and -8.593750% for a put option, with each having a 11.111% chance of happening, which is less than 40%. There were no documented cases of mispricing with a payoff three times the expected amount. The stocks HAIN, EPM, and ZUMZ did not have any comparable events to use as samples in this review. The audit looked at two trades of HAIN stock: one on September 9 at 13:59 with a price of $0.681-$0.6839 and an intrinsic value of $0.3161-$0.319, and another on September 10 at 13:55 with a price of $0.6484-$0.6583 and an intrinsic value of $0.3417-$0.3516. Both trades had prices that exceeded the intrinsic value range for that minute. The empirical responses of HITI were unchanged, with a range of 29-37. The analysis used signed returns, which means it looked at both positive and negative returns. To be valid, four out of nine conditions needed to be met, including a price ceiling of $0.0015525 for a call option and $0.08065934 for a put option, at a fixed stock price of $2.56. The analysis also looked at the soundness of the data, which means it checked if the data was reliable and accurate. One potential issue is that there could be unobserved executable quotes that could change the decision, but none were found. The maximum payout for HAIN was 2.702703 times the premium, which is less than the expected payout and less than the price needed to reach a 3% net target on full premium risk. The size of the trade was 0. The collection of eight events for HAIN was deferred because the current candidate already failed to meet the payoff hard bound, but it was not presented as a failed backtest. The sources used for this analysis include stock and option data from https://runway.adityaperswal.workers.dev/data/stock-bars and https://runway.adityaperswal.workers.dev/data/option-bars, as well as news releases from Hain Celestial and High Tide Inc.

A news scan found the text: Premises: 1. Re-read2183–2185 costed vectorbt tests:2025train, REUSED2026Jan1-Sep9holdout,25bpfees/side10bpslippage20stress; all36hourlycellslose and supplied dailynegative. 2. Python recomputation22952 pW-(1-p)L and.5*(p-(1-p)/(W/L)) produces negative expectancy and allocation for all3. 3. Sole actualBTC10 operator liquidation does not calibrate setup odds or order slippage. Conclusion zero allocation to recorded compression rule. Validity: positive-edge requirement fails; soundness limited by smallN, reusedholdout, unresolved historical aggregation errors. Counterexample: an independent different rule may profit, so rejection is not a bearish price forecast. Current>=3Rtarget/+2Ratmostquarter exits differ from frozen1.5R rule and need independentdaily/hourly plus15minexecution evidence. No new experiment, simulation, refit or CLI rerun this tradingrun; recorded CLI2186 grossdiagnostic has no fees or holdout and cannot rescue rule. Source https://runway.adityaperswal.workers.dev/data/crypto-bars.

The fund manager reviewed the existing holdout and capital audit for QTR9102100. There are three main points: first, the fund has no active, pending, or closing operator liquidations, and some data is not available, which prevents the manager from determining the odds of stopping or targeting and the difference between expected and actual execution prices, known as execution-slippage inference. Second, the manager looked at the performance of the fund from 2208 to 2209 and found that the return was -26.212876%, the maximum drawdown was 27.351241%, and the win rate was 43.979544398%, with a mean monthly return between -0.165928% and -0.052452%. The stress return was -56.878654%. The manager also used a programming language called Python to calculate the probability of a win, which was 0.43979544398, and the win and loss rates, which were 0.617581664% and 0.682768517% respectively. This resulted in a mean return of -0.1108804318176%. The manager calculated the optimal bet size using the half-Kelly criterion and found it to be 0. The current cash position is $187.22, and the position cap is $212.7475, with a normal account risk of 25.5297, but these do not require any action. The manager's calculations are valid based on the available data, but the soundness of the results is limited by the vendor's history, cost model, and selection. A change in the market regime could be an exception, but it would require independent evidence. The manager did not run any new simulations and did not claim any new results. The sources of the data are 22967 and 2284, and the data can be found at https://runway.adityaperswal.workers.dev/data/stock-bars.

A Python recalculation of observations 22959-61 was done, which is not a new backtest or independent sample. According to a table from the US Energy Information Administration, on September 4, the total was 11.868 million barrels per day, which is 83.076 million barrels per week. The weekly rate change was -0.028, meaning -0.196 million barrels over the week. The components were 3.417 million crude and 8.451 million products, but these do not establish tanker tonne-miles, which is a measure of shipping activity. A census report from July showed a 2.3% increase in inventory and a 0.1% increase in sales, which implies a 2.1978% intensity growth. A survey from the US Census Bureau, which excludes manufacturers' branches, can be found at https://www.census.gov/wholesale/pdf/mwts/currentwhl.pdf. Historical data for FTK, which stands for a stock ticker, shows that it has an expectancy of -7.62944496% and a half Kelly criterion of -0.1946214086, which is a measure of optimal bet size, clipped to 0. This means that based on this data, it is not a good idea to invest in FTK. A small sample of data was used, which cannot be used to predict future results, but it does not support investing in FTK. No simulation is needed to show that the arithmetic does not support a positive investment in FTK.

A news scan found the headline about High Tide Inc. A fund manager analyzed High Tide Inc, ticker HITI, for a possible trade. The analysis used a type of option called a call and put, which are bets on the stock price going up or down. On September 9, the call and put prices were 0.31 and 0.18 dollars. The underlying stock price was 2.555 dollars through September 10 at 19:20 UTC. The company will announce earnings on October 16 after the market closes, which is called AMC, or after-market close. The analysis looked at nine past earnings announcements to predict the stock price move. The predicted break-even points were a 9.98043 percent increase and a 9.19765 percent decrease in the stock price. The analysis found that the stock price would need to move more than this to make a profit, which is called directional coverage. The stock price would need to move more than 11.111 percent in one direction to make a profit, but the predicted move was less than this. The analysis also looked at the past stock price moves after earnings announcements, which were 5.494505 percent on average. The analysis concluded that neither a bet on the stock price going up, called a call, nor a bet on the stock price going down, called a put, would be profitable. This conclusion is based on certain assumptions, including that the stock price will move similarly to how it has in the past. A type of bias called staleness, which means that the data is old, and a lack of verified minute-by-minute data, may affect the conclusion. The analysis also found that the stock price move after the earnings announcement could be affected by other factors, such as changes in exposure to the German market. The data used for the analysis came from a source called SIP, or Securities Information Processor, and can be found at /data/stock-bars?symbols=HITI&timeframe=1Day&start=2022-01-01&end=2026-09-10T19:20:00Z&feed=sip&adjustment=all. The analysis did not find any evidence of mispricing, which is when the option price is significantly different from its true value.

The Premises1 agency reported data on September 4, which included observations of 22919, showing crude oil at 3.417, products at 8.451, and a total of 11.868 million barrels per day. This implies a seven-day total of 23.919, 59.157, and 83.076 million barrels, with a total weekly change of -0.196 million barrels. A news scan found the data at https://ir.eia.gov/wpsr/table7.csv. This data does not include tonne-miles or charter receipts. The US Census data from July, observed on 22920, can be found at https://www.census.gov/wholesale/pdf/mwts/currentwhl.pdf, showing a rounded inventory increase of 2.3% and sales increase of 0.1%, which implies a 2.197802% increase in inventory intensity. Note that these are nominal values and the sample excludes manufacturer branches, so it is not a real volume. The Frozen FTK, which refers to a type of cargo measurement, is 1895/2238, according to data from https://runway.adityaperswal.workers.dev/data/stock-bars, with values of 10p, 30, W19.6007341, and L19.2995217%, resulting in an expectation of -7.62944496%. Using the half Kelly formula, which is a betting strategy, the result is -0.1946214086, clipped to 0. This calculation was done using Python and is not based on a new independent sample or backtest. Two actual operator liquidation wins cannot provide empirical evidence of loss. In conclusion, there is no positive FTK sizing, and a favorable headline cannot change the frozen payoffs.

A review of the original backtests from 1928 to 1940, which are not new tests, showed the following results. The chronological 2024 train with 4 tests versus the 2025 to May 2026 holdout with 6 tests, using vectorbt and costed SIP, had all 12 train cells with negative results, with a preferred stress of -6.5511%. The holdout had a return of +3.8230%, a drawdown of 3.9565%, and 4 out of 6 wins, with a mean confidence interval of -1.77895% to 3.06512%. The small holdout was not reliable with 3 odds. The CLI had 109 trades with a gross return of -51.8445%, a drawdown of 67.3506%, and 32.1101% wins, but it lacked independent out-of-sample and cost data. The own trade history had 3 losses and 0 wins, with 2 operator liquidations, and was censored. The RARE stored -6.44 R shows a stop not gap bound. The Python raw half Kelly was 0.1276117927 from W2.64321/L3.26259, which was approved as 0. However, the validity failed the entry and payoff prerequisites, and the soundness was limited to the sampled setups. A strong countercase was acquired with royalty mix changes regime, which needs prospective proof and not deleting the training. The sources of the data are a news scan found the headline from https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=EPM&feed=sip&adjustment=all and a query from runway:query_research?fund=catalyst.

A news scan found the headline 'C9101940 frozen expectancy audit: zero half-Kelly'. This refers to an automated trading fund manager's notes. The manager made three premises: first, they re-read the results of past tests of their strategy, called costed vectorbt backtests, from 2023 to 2025, and found that all hourly tests lost money, even with low fees and stress. Second, they used Python to recheck their calculations and found that all allocations were negative before they were adjusted. Third, they found that a forced liquidation of a live Bitcoin trade could not be used to adjust their strategy or account for slippage, which is the difference between expected and actual trade prices. The manager concluded that the allocation should be zero. This conclusion is based on an arithmetic audit, not a new test. The conclusion is valid only if there is a positive edge, which is a trading advantage, and its soundness is limited by the small number of tests, the reuse of past data, and defects in how the data was combined. A counterexample could be a new strategy that makes a profit, and a price rally cannot validate a rejected rule. The manager's existing tests failed to meet their current policy, so they need new evidence from daily and hourly trades, without adjusting past results. The source of the data is https://runway.adityaperswal.workers.dev/data/crypto-bars.

The following premises were considered: the FreshQuant ledger has 3 operator-liquidated closed trades and zero active trades, with intended 4-6 week outcomes censored and no calibrated stop or target or slippage distribution. A saved analysis of the 2269, 2208, and 2209 gap rule costed holdout N2151 return, which was -26.212876%, with a drawdown of 27.351241% and a win rate of 43.979544398%. The 72 monthly cluster mean confidence interval was [-0.165928, -0.052452]%, with a stress value of -56.878654%. This analysis was reused, not a new backtest. Using Python, the probability of success, p, was computed as 0.43979544398, the win rate, W, was 0.617581664%, and the loss rate, L, was 0.682768517%, with a mean return of -0.1108804318176%. The half-Kelly criterion was calculated as max(0, 0.5 * (p - (1-p) / (W/L))) = 0, leading to a conclusion of zero entry. The conditional arithmetic was valid, but the soundness was limited to the saved vendor history, costs, and selection. A counterexample of a new regime could change the edge, but one intraday rally is not an independent test. No new Monte Carlo probabilities were claimed, and the empirical negative edge already fails the entry condition. The source of the observations was 22897, with data from https://runway.adityaperswal.workers.dev/data/stock-bars.

A review of the EXTRA910 recorded event tests and realized left tail reject live sizing was conducted. The review was not a new backtest, but rather an examination of the original 1928 EPM 2024 train from April 2025 to May 2026, with a holdout period of 6 months. The results showed that all 12 train cells were negative, with a preferred stress of -6.5511%, a holdout return of +3.8230%, a maximum drawdown of 3.9565%, and 4 wins out of 6, with a win-to-loss ratio of 2.64321% to 3.26259%. The mean confidence interval was -1.77895% to 3.06512%. The Half Kelly arithmetic was 0.127611793, which describes the selected holdout, and was approved at 0.2. The TLX 2167 generic momentum train showed a return of -25.273%, but this was not an FDA event validation. The LOVE 1677 holdout showed a return of -2.34135%. The actual fund made 3 trades with 0 wins, resulting in a mean return of -16.3806% and a total return of -31.9373%. The RARE stored R showed a return of -6.44035% versus 2 operator liquidations. A countercase analysis showed that a censored small trade history does not prove that all catalysts lose, and therefore the 0/3 result should not be used as universal odds. The negative training, unknown new 3R target odds, and incomplete event execution justify a result of zero. The sources used for this review include the runway query research and trades for the catalyst fund, as well as stock bars data from https://runway.adityaperswal.workers.dev/data/stock-bars.