Stats

A Python analysis used 2540 complete adjusted open, high, low, and close observations from September 19, 2016, to September 16, 2026, to examine Elmos' return distribution and market covariance. The analysis found that the correlation between Elmos and the DAX was 0.352652, between Elmos and LFUS was 0.418820, and between LFUS and the DAX was 0.290343, based on 2026 common-date log returns. An ordinary least squares regression of Elmos against the DAX showed a coefficient of 1.082706 and an intercept of 0.001876274 per day, with an R-squared of 0.124363. However, this analysis is descriptive only and does not provide predictive alpha inference or causal attribution. The analysis also noted that US and German close times differ, and the DAX performance index and adjusted shares give imperfect comparability. During the September 8-9 SK announcement window, Elmos' return was -0.418988%, the DAX return was -1.657904%, and the LFUS return was -0.229342%. The primary supplier release occurred at 7 ET during European trading, providing no independent event edge. The annualized RV20 was 42.28358%, and the historical worst adjusted opening gap was -12.621354%, which is not a loss bound. The primary input for this analysis was from the Yahoo Finance chart endpoint for ELG.DE, with the same endpoint used for GDAXI and LFUS. The analysis found 10 dividend events with no splits returned, but did not conduct an independent full corporate action audit. The budgeted research does not generalize one issuer to the entire 200 million euro industry.

A news scan found the following text: Premises: For the years 2023-2025, L20 held-apart net trades were 20 basis points per side, with no own live closed trades, so no measured live slippage or win rate. This means that the results are based on simulated trades, not actual trades. The results are based on 20,000 bootstrap means, which is a statistical method, and selection and autocorrelation were not corrected, which refers to the relationships between the data points. The win rate for ATI was 42.105%, the average win was 20.151%, and the average loss was 6.471%, with a 95% mean confidence interval of -1.249% to 11.189%. The win rate for SEI was 35%, the average win was 26.102%, and the average loss was 10.057%, with a confidence interval of -5.059% to 11.123%. The half Kelly risk, which is a formula to determine the optimal bet size, was calculated as 0.5 times the probability of winning minus the probability of losing divided by the average win divided by the average loss. The results were 0.029794 for LFUS, 0.117571 for ATI, and 0.049780 for SEI. The risk budget was then divided by the loss fraction to determine the notional value, and then fund, venue, and gap caps were applied. However, these values are not recommended equity allocations. For 2026, the SEI volatility was 80.119%, and the beta, which measures the volatility of an asset compared to the overall market, was 2.694 for SPY. The correlation between LFUS and VSH was 0.70093, which shows a shared factor, but the beta of PXS was 0.122, which does not remove the single issuer loss. The worst observed SEI gap was -16.949% on January 27, 2025, but this is not the maximum possible gap. All statistics observed are based on actual data. Positive confidence intervals cross zero, and business regime changes limit the soundness of the results. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars. No order was solely based on the Kelly criterion, which is a formula for determining the optimal bet size.

The R92102 census is a broad attempt to measure all listings, but it is not complete. It includes 12,416 US listings, 19 foreign equities, and eight crypto pairs. The census was created by trying to use all 13,221 non-test listings from the Nasdaq symbol directories from August 1 to September 18. Of these, 12,571 returned and 12,416 were measurable. The remaining 650 listings were missing, including 384 with unsupported symbol formats and 266 with valid formats that were missed after trying again. Additionally, 155 histories were too short. The census filtered out non-ETF listings with a close of less than $2 and a 20-day mean dollar volume of less than $1 million, resulting in 3,869 listings. Of these, 181 had breakouts in the prior 20 days. The 20-day distribution had a P10 of -15.3606%, a median of -3.4636%, and a P90 of +8.1108%. Some stocks, such as UMC, PXS, and SEI, are affordable breakout candidates. Others, such as ATI, LFUS, and VSH, are primary capacity candidates that are currently below their frozen breakout. The census also includes 19 foreign stocks and Brent, covering Japan, Korea, Taiwan, Europe, India, and Brazil, but not all foreign listings. DBHiTek had the strongest 20-day gain of +40.5797% and a 60-day loss of -17.2572%. The census also includes eight crypto pairs that completed on September 20, with AVAX having a 20-day gain of +56.3908% and an annual volatility of 86.9035%, and BTC having a 20-day gain of +3.3058%. However, there is no demonstrated token claim on paid industrial scarcity, so there is no mandate-based crypto entry. The sources used for the census include the Nasdaq trader website and other financial data websites. Every measured screen row was journaled in chunks, and missing coverage was named but not repaired beyond the discovery allowance.

A news scan found the source SIP/all daily at https://runway.adityaperswal.workers.dev/data/stock-bars. As of 2026, the analysis of stocks LPG, BWLP, and CSTM shows the following: LPG has a beta of 0.4719, correlation of 0.1391, and annual volatility of 44.97%; BWLP has a beta of 0.3225, correlation of 0.1040, and volatility of 41.12%; CSTM has a beta of 2.0005, correlation of 0.5441, and volatility of 48.73%. The correlation between LPG and BWLP is 0.7668, indicating that low market beta does not make the two shipowners independent. The analysis of LPG's performance shows a base holdout of 17, with 6 wins and 11 losses, an average win of 20.4379%, and an average loss of 7.9172%, resulting in a mean of 2.0905%. A 20,000 iteration bootstrap simulation shows a 95% confidence interval of -4.2318% to 8.8474%. Recently, the confidence interval is -0.1053% to 17.8982%. For BWLP, the mean is -2.2291%, with a confidence interval of -6.9539% to 4.1100%. The binary average payoff half Kelly risk is 0.0511419 for LPG and 0 for BWLP. The risk fraction must be divided by the loss fraction to convert to notional, then capped, not 51 USD notional. The sample and selection uncertainty are substantial, and there are no live closed trades to calibrate slippage and probabilities. The worst observed open gaps are -22.275% for LPG on March 16, 2020, -13.299% for BWLP on April 4, 2025, and -19.625% for CSTM on October 23, 2024. A corporate action audit is not complete. These are descriptive bounds of the sample, not future loss limits. The prior is neutral, and the current evidence is positive for LPG, but not enough to justify a new order.

The automated trading fund manager analyzed data from 2017 to 2026. They looked at daily returns, which are changes in stock prices, adjusted for a system called SIP. The manager found that the stock CSTM had a beta, or volatility measure, of 1.849, 1.488, and 2.001 for different periods. The stock CENX had a beta of 1.346 and a volatility of 64.02% for 2026. The manager also found that the correlation, or relationship, between CSTM and CENX was 0.7471, and between CSTM and KALU was 0.7195, meaning these aluminum positions are not independent bets. The manager used a statistical method called OLS, which is descriptive, but did not make any claims about causality or significance. The manager tested CSTM and found that it completed a pullback, with 6 wins out of 12, and an average win of 17.0718% and an average loss of 6.13565%. The manager also used a method called bootstrap to estimate the mean return, which was 5.4681%, and found a 95% confidence interval of [-1.3390, 12.5495]%. For CENX, the mean return was 7.7655% with a confidence interval of [-5.0361, 20.9290]%. The manager used a binary average-payoff half Kelly method to determine the risk fraction for CSTM, which was 0.160149. Dividing this by the average fractional loss of 0.0613565 gave a wealth allocation of 2.61014. The manager noted that this is not the same as a 16% stock notional. The manager trained CSTM with a risk of 11.2375% and a wealth allocation of 119.645%, and CENX with a risk of 11.7722% and a wealth allocation of 85.003%. However, these estimates ignore within-tail distribution and parameter uncertainty. The manager concluded that CSTM had a positive conditional point estimate, but the current entry does not allocate funds correctly. Any future trades will be capped by the fund at 120, the account position at 250, and gap risk, and will not be unbounded Kelly. The manager noted that selection and multiple lookbacks, as well as serial dependence, mean that the bootstrap confidence interval underestimates model uncertainty.

This analysis is based on certain premises. First, it uses adjusted daily returns from a specific data source, covering the years 2017-2022, 2023-2025, and 2026, with 1510, 752, and 179 data points, respectively. Second, it calculates ordinary least squares, or OLS, against the SPY, which is a type of stock market index, and finds equity beta values of 1.2436, 1.4622, and 0.2641 for CE, and 1.1838, 1.1138, and 0.9004 for EMN, with a correlation of 0.0657 in 2026 and an annual volatility of 53.24%. This analysis is descriptive and does not claim robust significance or causal relationships. Third, it looks at fixed-size trades, holding apart CE, and finds that out of 12 trades, 25% were winners, with an average win of 9.4077% and an average loss of 5.3917%, while for EMN, out of 15 trades, 13.333% were winners, with an average win of 15.1456% and an average loss of 4.5233%. The Half Kelly formula, which is used to calculate optimal bet size, is max(0, 0.5 * (p - (1-p) / (win/loss))), and in this case, it equals 0 for both. Fourth, it runs 20,000 independent and identically distributed, or IID, mean bootstraps, which are a type of statistical simulation, and finds 95% confidence intervals of -5.4125% to 3.1093% for CE and -4.8230% to 2.0034% for EMN, which represent uncertainty intervals, not downside bounds. The live supply ledger shows no measured slippage or historical fund edge, meaning no difference between the expected and actual trade prices, and no historical advantage of the fund. In conclusion, this analysis does not support a calibrated positive position size from this rule, and the changed equity beta is evidence against assuming an ordinary broad beta, but not proof of scarcity alpha, which refers to exceptional investment returns. The analysis is limited by the use of selected names and regimes, repeated historical windows, small samples, and serial dependence, which means that the results may not be applicable in other situations. Instead of claiming that the business cannot work, it is better to preserve a positive cash valuation countercase.

The premises of this analysis are based on two main points. First, the analysis uses a fixed long-only chronological holdout test from 1989 to 1992, which utilizes data from https://runway.adityaperswal.workers.dev/data/stock-bars. Second, the win rate, average win, and average loss percentages are as follows: CE 17.647%, 21.6997%, and 6.0195% with 17 trades, IMOS 29.1667%, 16.6557%, and 6.7616% with 24 trades, BTDR 33.3333%, 42.9153%, and 16.8298% with 24 trades, and AMTX 33.3333%, 41.2324%, and 15.0053% with 21 trades. The Half Kelly risk fractions, which represent the stake-at-risk fractions, not capital allocation, are calculated as 0, 0.002054, 0.035946, and 0.045360. Additionally, the 20000 IID bootstrap mean confidence intervals all cross zero: CE [-5.833, 4.665]%, IMOS [-4.275, 4.664]%, BTDR [-8.060, 15.665]%, and AMTX [-7.657, 16.539]%. It is noted that average positive trade returns can coexist with negative compounded returns, and the arithmetic mean cannot be used as a proxy for profit and loss. The conclusion is that uncertain regime-dependent distributions do not establish an actual execution edge. There are no live post-reset supply trades to estimate slippage, and costs are based on assumptions. Candidate selection uses current winners, and historical windows are not untouched prospective evidence. The actual size is zero because the cash value and current rule gates fail, not because uncertainty itself forbids a convex probe.

The Nasdaq directory at https://www.nasdaqtrader.com/dynamic/SymDir/nasdaqlisted.txt and https://www.nasdaqtrader.com/dynamic/SymDir/otherlisted.txt contains 13,220 unique non-test entries. A snapshot from https://runway.adityaperswal.workers.dev/data/stock-snapshots?feed=iex, which is a data feed, queried 12,836 entries after excluding 384 preferred formats, and returned 12,689 usable pairs, including 8,133 daily bars from September 18. A subset of 3,430 non-ETF entries with an IEX dollar volume greater than $100,000 had a median return of -0.55243%, with quartiles of -1.64402% and 0.40347%. The mean return was 0.59191% with a standard deviation of 50.059%, which was contaminated by a 29.17% corporate-action collision outlier, not economic market breadth. The initial whole batch failure from preferred syntax was repaired, and there were no final request errors. Broad discovery flagged FEAM, BTDR, SGML, IMOS, FPS, and AMTX, as well as a crypto-beta cohort, all of which require SIP actions and primary verification. Names without new bottleneck connections were rejected at screen only, not due to a fundamental verdict. The dataset is the current listing survivor universe, including ETFs, warrants, and preferreds, and is not a point-in-time all US security master or OTC. A full name keyword filter misses opaque company names, so CE, EMN, LYB, OLN, and MEOH were manually added to the chemical chain. Sources are close to the market, but IEX partial volume is not consolidated. The full raw files are located at /tmp/universe.csv, /tmp/snapshots.json, and /tmp/screen.csv. There is no order on 1-day returns.

The analysis is based on two premises. First, a backtest of 3555 trades was conducted, with 49 trades used for training, 16 for holdout, and 5 recent trades, using data from https://runway.adityaperswal.workers.dev/data/stock-bars. Second, 20000 random resamples were generated using NumPy with a seed of 92000, resulting in mean confidence intervals of -3.723% to 2.966% for training, -5.924% to 13.196% for holdout, and -4.230% to 32.607% for recent trades. The holdout average win was 27.949% and average loss was 11.193%, with a win rate of 37.5%. The half Kelly risk fraction, which is the fraction of capital to be put at risk, was calculated as 0.06236 for training and 0.22371 for recent trades. This fraction is based on a two-point payoff model and is not the same as a 6.24% portfolio allocation. The correlation between FRO and POWL daily returns was found to be 0.0122 over 2693 common dates. The analysis concludes that while positive regimes are meaningful, the small sample size means that the odds are uncertain, and a recent-only Kelly strategy should not be used blindly. The absence of a current POWL entry and an unsupported target remain binding, resulting in an actual size of 0. A confidence interval crossing zero alone is not a reason to reject the strategy.

This analysis is based on several premises. First, the cost of FRO was analyzed using a baseline of 3517 natural exits, with data from https://runway.adityaperswal.workers.dev/data/stock-bars. The results showed a mean return of -1.678% with a win rate of 22.92% in the training set of 48, a mean return of 0.0637% with a win rate of 36.36% in the holdout set of 11, and a mean return of 10.433% with a win rate of 66.67% in the 2026 set of 6, resulting in a compounded return of 72.863%. Second, using a NumPy seed of 91908, a bootstrap analysis of 20,000 independent and identically distributed trades showed a 95% mean confidence interval of -5.365% to 2.345% in the training set, -5.751% to 6.444% in the holdout set, and -0.630% to 21.096% in the recent set, but did not provide future loss bounds. The average win and loss were 12.564% and 7.079% in the holdout set, and 19.409% and 7.519% in the recent set, with a half Kelly risk of 0.002536, 0.002768, and 0.268768 in the training, holdout, and recent sets, respectively. Third, the daily return correlations between FRO and other stocks were 0.70715 for STNG, 0.60907 for LPG, and 0.74201 for LPG and BWLP, and 0.14275 for SPY and FRO. Several shipping names are partly one bet and not independent diversification. Fourth, the query trades supply was empty after the September 17 reset, and pre-reset operator liquidations cannot calibrate live edge or slippage. The recent regime improvement is real evidence, but was retrospectively selected. In conclusion, the analysis suggests retaining a watch on FRO and not claiming no edge just because the confidence interval crosses zero.

The analysis of PSIX is based on two premises. First, the daily adjusted returns of PSIX and other stocks, including ETN, POWL, NVT, and SPY, from December 27, 2024, to September 18, 2026, are available at https://runway.adityaperswal.workers.dev/data/stock-bars?symbols=PSIX,ETN,POWL,NVT,SPY&timeframe=1Day&feed=sip&adjustment=all. Second, the total return of PSIX is 55.5443%, and if the best and worst 5 days are removed, the returns become -40.3676% and 628.9994%, respectively. The best 5 days account for 217.025% of the net log gain. This symmetric deletion is a descriptive stress test, not a feasible trading strategy or a reason to avoid investing in PSIX due to its fat-tail nature, which refers to a distribution with extreme values. The mean daily return is 0.32275%, the median daily return is -0.34725%, and the 1st and 99th percentiles are -17.5354% and 19.0215%, respectively. An ordinary least squares regression, or OLS, which is a method to estimate the relationship between variables, was performed with the SPY index and an equal weight of ETN, POWL, and NVT. The results show an intercept of 0.065405% per day, a market beta of 1.37135, and a power beta of 0.83742, with an R-squared of 0.27974. However, the presence of correlated factors and selected survivors, which are stocks that have survived a certain criteria, precludes causal alpha, or the ability to generate excess returns. The natural exit in 2025 had a mean return of 23.6247%, with a 95% confidence interval of -2.3439 to 49.5933, and in 2026, the mean return was -1.9832%, with a confidence interval of -15.9065 to 24.5858. These wide intervals indicate that the bootstrap method, which is a statistical technique to estimate the distribution of a parameter, cannot recover missing crises. In conclusion, tail dependence is expected, but the regimes and few trades cannot calibrate this entry, meaning that the analysis cannot accurately predict the behavior of PSIX due to its complex nature.

A study of daily stock returns from 2026 through September 18 found the following: the daily adjusted simple returns for OCC, GLW, and SPY stocks were analyzed. The correlation between OCC and GLW was 0.481718, and between OCC and SPY was 0.199248. A statistical method called Ordinary Least Squares, or OLS, was used to analyze the relationship between OCC and SPY, which found an intercept of 0.869058% and a beta of 2.059435, with an R-squared value of 0.039700. The daily return percentiles for OCC were -13.2705% at the 1st percentile, -10.4709% at the 5th percentile, -0.28454% at the 50th percentile, 14.8018% at the 95th percentile, and 28.6742% at the 99th percentile. This means that OCC's returns are highly dispersed, with a positive mean and intercept, but a negative median and a fat positive tail. The study concludes that the relationship between OCC and its suppliers, such as GLW, is not a reliable indicator of future returns, and that the correlation between OCC and GLW is not strong enough to establish a causal link. The study also notes that GLW and OCC are partly correlated exposures, not independent positions, and that no significant or stationary relationship was found.

The automated trading fund manager made some observations about certain stocks. They looked at the daily returns of these stocks from February 6 to September 18, 2026, and found some connections between them. Specifically, they found that the correlation, which measures how much two things move together, between GLW and MU was 0.5976, between GLW and FPS was 0.5541, and between FPS and BTDR was 0.5052. They also used a statistical method called ordinary least squares, or OLS, to estimate the beta, which measures how much a stock moves compared to the overall market, of each stock. The betas were 3.254 for GLW, 3.387 for FPS, and 4.412 for BTDR. However, the manager notes that these findings are not conclusive and may be influenced by the fact that they were looking for connections after hearing about artificial intelligence, or AI, narratives. The manager concludes that stocks in the glass, switchgear, and computing industries may be related in terms of risk, despite having different supply chain issues. This means that if the manager decides to invest in these stocks in the future, they will need to consider their combined risk, rather than treating each investment as a separate bet. The manager also notes that unexpected events, such as funding or production shocks, can overwhelm the average correlation between these stocks. Currently, the manager does not have a position or plan to make a trade in these stocks.

A news scan found the text describing payoff distributions for a trading fund. The fund used a method called bootstrap draws, which is a way to estimate how well a trading strategy works. The results showed that for the stock GLW, the average payoff was 0.322% per trade, with a 95% confidence interval of -2.004% to 2.870%, and a win rate of 33.33%. The average win was 10.22% and the average loss was 4.63%. For the holdout set, which is a group of trades not used to train the model, the average payoff was 4.557%, with a confidence interval of 0.090% to 9.335%, and a win rate of 56.25%. The average win was 12.00% and the average loss was 5.01%. The fund also tested other stocks, including OCC, BTDR, FEAM, FCEL, and MU. The results for these stocks were mixed, with some showing positive payoffs and others showing negative payoffs. The fund used data from a source called stock-bars, which provides historical stock prices. The results are based on certain assumptions, including that the trades are independent and identically distributed, and that the companies chosen are still in business. The results are not proof of a current trading edge, but rather evidence that can be used to investigate further. A countercase is that new contracts or scales can improve future cash flows despite historic losses.

A fund manager analyzed the performance of two stocks, NVT and HUBB. The analysis used a method called bootstrap, which is a way to estimate how reliable the results are, and found that NVT had a higher average return per trade, 3.036%, compared to HUBB, which had a negative return of 21.13% in the holdout period. The holdout period is a time frame used to test the performance of a strategy, separate from the time frame used to train it. The results also showed that NVT had a higher win rate, 44.44%, and a higher average win, 13.558%, compared to HUBB. However, the analysis noted that the results are limited by the lack of precision and the reliance on a single ledger, which is a record of all transactions. The analysis also found that two trades, FRO and STNG, were winners, but there was no stop or target exit, which is a price level at which to sell a stock, and therefore, the strategy's stop probability, average losing trade, or execution slippage, which is the difference between the expected price and the actual price, could not be estimated. The data used for the analysis was from September 2-4, after a reset, which is a restart of the system. The analysis concluded that NVT merits a prospective signal and cash monitor, but statistical confidence alone is not enough to authorize entry into a trade, and other factors, such as a missing trigger and current target economics, control the decision.

A news scan found the report titled R91808 canal denominators, which states that there is no queue growth identifiable from monthly averages. The report includes the following points: 1. The DirectACP Sep10 AugustA34 report, available at https://pancanal.com/wp-content/uploads/2026/09/adv-34-2026-monthly-canal-operations-summary-august-2026.pdf, shows that the monthly total was 1029, with 33.19 transits per day and 33.1 arrivals per day. 2. A Python analysis found that the water-minus-transit time was 23.63 hours, which is 67.7077% of 34.90 hours, but this includes all non-transit time and not just congestion wait time. The rounded arrivals minus departures was -0.09 per day, which does not establish an actual queue change. 3. The report also found that 175 out of 178 Neo booked slots were used, which is 98.3146%, and 204 out of 227 auction slots were used, which is 89.8678%. However, Neo excludes auctions, so adding these percentages would double count and change the denominator. The physical capacity of 33.19 per day out of a sustainable combined capacity of 36-38 implies an utilization range of 87.34-92.19%. 4. The report concludes that prior constraints can raise freight, but the evidence weakens the inference of accelerating queue from headline booking fill. It also notes that concentrated Neo arrival waves can produce queues despite balanced monthly totals, and that daily class-specific arrivals, wait distribution, auction prices, and paid TCE are needed before calibrating stock edge.

The source https://runway.adityaperswal.workers.dev/data/stock-bars SIP/all Jan2016-Sep17 shows that TEN has 2691 daily gaps and none are less than or equal to -30%. The worst gap was -9.5963% on April 29, 2020, and the worst pre-2024 close return was -17.58995%. This does not mean there is no risk of a future disaster. Using a simple model, the daily upper limit is 11.1262%, which is 6.46126% over 60 sessions. A 5% probability per trade cannot be directly compared to a daily frequency, and crisis regimes can happen in groups, which undermines the simple model. So, the earlier 5%/-30% remains a stress case, not a calibrated probability, and the absence of such an event does not refute it. PSIX had an actual adjusted gap of -34.18735% on May 12, 2026, and -29.99754% on November 7, 2025. LPG had a worst pre-2024 gap of -22.27513%. The AA feed on November 1, 2016, showed an apparent -74.35646% gap, which was due to a corporate separation, as confirmed by https://investors.alcoa.com/resources/investor-faqs/default.aspx. If we start the current AA history from November 1, 2016, the clean worst gap is approximately -13.2465%. The correlation between TEN and other stocks is: TEN/FRO is 0.819413, TEN/STNG is 0.708182, and TEN/LPG is 0.600194. This means that the holdings would share a shipping factor, not independent bets. The conclusion is to retire the uncalibrated tail veto while keeping explicit gap sizing and regime uncertainty. The sample of closed trades cannot estimate the edge. The hard account risk and fund 101.99 cap remain, and there is no forced entry from removing an invalid inference.

A review of the R91806 trading strategy found that its premises were not accurate. The strategy assumed a 5% probability of a 30% loss, which gave a conditional mean of 1.4467783586% and a breakeven point of 4.60072044%. However, there was no data to support this 5% assumption. The trading history only had two winning trades with no independent samples of losses. Adding a potential large loss to the historical returns could double-count the risk. Therefore, the strategy's use of this scenario as a hard veto is being superseded. It will be retained as a sensitivity test, labeled as such, and used alongside other empirical data and stress tests. The correction is to the model, not a signal to chase trades. The current frozen prices of 52.81, 45.41, and 55.36 still yield a low return of 0.31453R. The data source is https://runway.adityaperswal.workers.dev/data/stock-bars, and new panel data from September 17, 2016, is being analyzed, with results pending.

The trade query for T9131350 contains only two completed market entries, both of which were liquidated by the operator on September 4. The Python gross returns for these trades were 3.805162958% and 0.752256770%, with a mean return of 2.278709864%. Both trades were winners, but since there were no losses, it is not possible to estimate the average loss or execution slippage. Additionally, the calibrated Kelly criterion cannot be determined from this limited data. The source of this information is a query of trades for the supply fund. A separate, independently recorded test, TEN3214, showed the following results: train 22 with a return of 19.190992%, a maximum drawdown of 49.046502%, and a win rate of 36.3636%. The known holdout sample of 11 closed trades had a return of 1.924385%, while the stress test result was -0.293468%, and the marked result was 16.576931%, which includes one open trade. The data used for this analysis can be found at https://runway.adityaperswal.workers.dev/data/stock-bars. No new backtest has been run, and no new rule has been trusted, so the selected name and previously viewed holdout remain limitations.

A news scan found the text describing the performance of a trading system, R91308. The system's training data showed a mean return of 1.649875% and a median return of -4.372739%. The returns ranged from -10.027048% to 30.464151%. The system's holdout data, which is a separate set of data used to evaluate the system's performance, showed a mean return of 0.487839% and a median return of -4.672473%. The system used a seed value of 91308 and was trained on 22 targets, with 4 of them being selected for further evaluation. The system's performance was also evaluated on 11 holdout targets, with 1 of them being selected. The system's stop-loss levels were set at 17 out of 22 training targets and 8 out of 11 holdout targets. The median number of sessions held was 14 out of 7. Two live trades were winners due to operator liquidations, but there were no strategy exits and no execution quote baseline. The system's positive arithmetic mean return supports further work on its policy, but the negative median return and closed stress loss are counterexamples. The data source is https://runway.adityaperswal.workers.dev/data/stock-bars, and the live query trades were for the supply fund with IDs 3 and 6.

A news scan found the headline 'R91306 tanker concentration: TEN and FRO remain one dominant exposure'. This refers to a report about the concentration of tanker stocks in a portfolio. The report analyzed daily log returns from 2020 to September 11, looking at five series: TEN, FRO, STNG, XLE, and SPY. The correlation between TEN and FRO was 0.672633, between TEN and STNG was 0.642277, and between TEN and XLE was 0.425455. The report also trained a model on 1257 days of data and found that the correlation between TEN and FRO was 0.672633, which is a measure of how much the two stocks move together. The report used a statistical method called Ordinary Least Squares, or OLS, to analyze the daily returns of TEN in relation to FRO and SPY. The results showed that for every 1% move in FRO, TEN moved 0.667135%, and for every 1% move in SPY, TEN moved 0.268046%. The R-squared value, which measures how well the model fits the data, was 0.460691. The intercept, or starting point, of the model was -4.788881 basis points per day. The report also found that the correlation between TEN and FRO can rise in times of stress, which means that the two stocks may not be as diversified as they seem. The report concluded that the current book, or portfolio, is flat, meaning that it has no net position. The data for this report came from https://runway.adityaperswal.workers.dev/data/stock-bars.

A primary cadence model, number 3183, was trained and the results were analyzed. The training data showed a mean of 2.112847%, with a range of -13.885557% to 30.214313% for the 5th percentile, median, and 95th percentile. The model was also tested with 20,000 independent and identically distributed, or IID, trades, which showed a mean confidence interval of -3.115923% to 7.696204%. Additionally, 10,000 circular block trades were tested, with a mean confidence interval of -2.617690% to 6.994972%. A known holdout set of 11 trades showed a mean of 1.906418%, with a range of -7.107902% to 15.266447% for the 5th percentile, median, and 95th percentile. The holdout set was also tested with IID and block trades, with mean confidence intervals of -3.505778% to 7.708049% and -1.634633% to 6.012300%, respectively. The model was trained on data up to 2025 and did not include any trades that spanned boundaries. The selection of trades was based on current surviving trades, not a random selection, and the holdout set had been examined in prior runs. The results showed 2 out of 2 live winners were liquidated by the operator, with no strategy exits, and no calibrated live Kelly or quote slippage. Despite a negative median and wide confidence interval, the positive arithmetic mean and payoff skew suggest that a small probe can still be investigated. The source of the data is https://runway.adityaperswal.workers.dev/data/stock-bars.

A managed raw paper train had a mean return of +0.66741% per trade. This was calculated using a method called bootstrapping, which is a way to estimate uncertainty, and the results showed a 95% confidence interval of -4.02761% to 5.78851%. Another method, called circular block, was also used and showed a 95% confidence interval of -3.52885% to 5.13261%. A holdout set, which is a portion of data set aside for testing, had a mean return of 6.01385%. The confidence intervals for the holdout set were -0.19308% to 11.98345% using one method and 1.15657% to 10.98076% using another method. However, the results from the holdout set with only 10 trades are not enough to outweigh the uncertainty in the model and data selection. The training data showed returns of -13.0172%, -6.7140%, and +28.1638% at different percentiles, while the holdout set showed returns of -8.0535%, +8.4027%, and +18.1984%. The historical sample used is the current TEN identity, which was selected after seeing strong recent performance, but the data is not independent and identically distributed, and some quotes are missing. The trade history for FRO and STNGN2 showed profits, but the operator was liquidated on September 4, which was not due to the strategy's exit rules. There is no measured order slippage because there is no contemporaneous quote available. The 100% observed win rate should not be used to calculate the Kelly criterion. More information can be found at https://runway.adityaperswal.workers.dev/data/stock-bars and https://runway.adityaperswal.workers.dev. What is needed is the exact implementation of the manager-time and raw quote, as well as prospective outcomes, rather than further looking back at past selections. A news scan found the headline, but in this case, the text is a raw execution uncertainty and limited live record report.

The automated trading fund manager analyzed the performance of a trading strategy. The results showed a daily return of 1.0437% with a 95% confidence interval of -3.7876% to 6.3721% based on 20,000 simulations. A separate test showed a return of 5.4980% with a confidence interval of -0.4618% to 11.4242%. The training data had 22 stops, 8 targets, and 3 time periods, while the holdout data had 5 stops, 4 targets, and 3 time periods. The correlation between TEN and FRO returns from 2025 to September 11 was 0.835826, and the correlation between TEN and CNQ was 0.276531. A statistical analysis using SPY and XLI showed an R-squared value of 0.0682, which is descriptive but not proof of idiosyncratic alpha. The existing FRO win and TEN setup are subject to one geopolitical and freight risk, not two independent bets. A news scan found the link https://runway.adityaperswal.workers.dev/data/stock-bars. The manager noted that positive point estimates require prospective validation and should not be treated as stable probabilities based on 12 trades. The bootstrap simulation assumes independent and identically distributed data, omitting regime clustering, and the currently selected ticker biases the screen.

The automated trading fund manager analyzed the KMT edge uncertainty and market exposure. The training data from L20 showed a mean return of -2.1782%, with a 95% confidence interval of -4.8701% to 0.8161%. The holdout data from N14 showed a mean return of 1.3471%, with a 95% confidence interval of -4.2810% to 7.4645%. The training data used 20 stop/gap stops and 2 targets, with 5 time exits, while the holdout data used 9 stop/gap stops and 5 target/gap targets. A paired daily analysis from 2025 to September 11, with 424 data points, found a correlation of 0.5665 between KMT and XLI. The ordinary least squares regression showed an intercept of 0.5074 basis points per day, a beta of -0.2713 for SPY, and a beta of 1.4806 for XLI, with an R-squared value of 0.3248. The analysis found no robust alpha significance and no causal tungsten coefficient. The own ledger still had two operator liquidations, with no natural stop or target exits, and a 100% win rate for 2 trades, which cannot be used to calibrate a new strategy or slippage. A news scan found the link https://runway.adityaperswal.workers.dev/data/stock-bars, with a query for trades fund supply. The analysis was based on two premises: a small conditional edge and a train negative premise. The conclusion was to not trade from the holdout data alone. The IID resampling understated serial and event clustering, and the current listings created survivor selection bias.