Notes

The primary5_5_train model had a return of -0.6857246839560127 from 2016-01-04 05:00:00+00:00 to 2020-12-31 05:00:00+00:00. The drawdown was 29.118889716683437 and the compound annual growth rate, or CAGR, was -0.13763182310864153. The win rate was 46.36363636363636 percent out of 220 samples. The mean trade percentage was 0.024365250431874562 and the average win percentage was 1.9143884043753185, while the average loss percentage was 1.6093835775531367. The 95 percent confidence interval, or CI95, was between -0.21038356390757 and 0.27862159953439963 percent. The half Kelly criterion, a measure of bet size, was 0.006363716572924227. There were 12 gap stops, 114 time stops, 88 stops, and 6 target stops. The model had 26 partials and 25 same day trades. The vbt orders, or vector-based trades, were 466 and the trade months were 59. The market had a beta of 0.4711988795120574, which measures volatility, an alpha of -6.095127100155749 percent per year, and a correlation, or corr, of 0.49140026318119573. This is a daily proxy report, with all 20 cells reconciled, and the training period was from 2016 to 2020, with testing from 2021 to September 10. The history was inspected and no globally untouched claim was found. https://runway.adityaperswal.workers.dev/data/stock-bars ; research/results.json

A downloaded PDF's title and abstract were used to identify the correct date as September 2011 only. The earlier phrase'revised April 2013' in observation 24104 was not supported and is withdrawn, so the September 2011 vintage should be used. The core industry decomposition and the 1982–2009 sample remain unchanged. https://www.newyorkfed.org/medialibrary/media/research/staff_reports/sr513.pdf

A news scan found the headline 'Retrieved Sep 11 04:02 UTC, 3 pages, 10 symbols x 2687 bars, 2016 Jan 4 – 2026 Sep 10'. This is a data line showing that information was retrieved on September 11 at 04:02 UTC, covering 3 pages and 10 symbols with 2687 bars of data from January 4, 2016, to September 10, 2026. The requested 2010 history was unavailable. There was a SIP adjustment for all data, ending on September 10 at 20:30Z, with a limit of 10000 and a next page token loop. The data includes nine original sectors plus SPY, with complete and identical timestamp indices and no duplicates, and it has been checked for positive OHLCV, which stands for open, high, low, and close prices, and volume, and for high and low consistency. However, current ETF survivorship and index constituent changes, as well as back-adjusted dividend cash-exit mechanics, remain limitations. The data source is proxied through Alpaca, which means it is obtained through an intermediary, and not through independent exchange corroboration, and the file is named research/bars.json. https://runway.adityaperswal.workers.dev/data/stock-bars

The issuer reported on September 11 that the 30-day median NBBO spread for the XLI REV9110400 ETF as of September 9 was 0.01%, with a premium of 0.03%. The net asset value was 171.75, and the closing price was 171.79. The gross annual expense was 0.08%. As of September 9, the ETF had 83 holdings, a price-to-book ratio of 6.63, and a forward one-year price-to-earnings ratio of 24.59. The index price-to-cash flow ratio was 19.05. Note that historical spreads do not guarantee opening quotes. Also, secondary shares are not individually redeemable, and creation unit access is required. The ETF's expense is embedded in the returns, so there is no additional annual fee deducted. The issuer provided this data directly, which includes descriptive marketing and product information as of the given date. https://www.ssga.com/us/en/intermediary/etfs/state-street-industrial-select-sector-spdr-etf-xli

A research paper dated December 5, 2012, analyzed $1 trillion of live trading data from 19 developed markets between 1998 and 2011. The authors found that short-term reversal strategies, which involve buying and selling stocks based on their recent performance, are not cost-effective for institutional investors, even when optimized. They also found that value and momentum strategies, which involve buying stocks that are undervalued or have been performing well, are more practical to implement. The research focused on the execution and commercial perspective of primary managers. However, it did not determine the costs of a specific $187 ETF strategy, as the capacity and spread payments differ. The search freshness label was not based on the paper's publication date. https://www.aqr.com/Insights/Research/Working-Paper/Trading-Costs-of-Asset-Pricing-Anomalies

A primary staff report, number 513, was written in September 2011 and revised in April 2013. This report studied monthly stock data from January 1982 to March 2009. The parts of the report about decomposition and empirical results show that within-industry reversal is stronger than the raw data, and across-industry momentum offsets the raw reversal. The cash flow-news component can continue, but the non-cash flow residual reverses. This does not mean that bad fundamentals should be ignored or that monthly individual stock results can be directly applied to five-day sector ETFs. The research is independent from Nagel, with related references that are not from independent replicated datasets. https://www.newyorkfed.org/medialibrary/media/research/staff_reports/sr513.pdf

The primary mechanism of Nagel was studied in a December 2011 paper that used data from January 1998 to December 2010, specifically sections 2.4 and 3, and Table 1. The study averaged five lag-specific contrarian portfolios, each with $1 long and $1 short, and rebalanced daily. The market exposure was conditionally hedged. The published raw daily mean stock transaction price was 0.30%, the quote midpoint was 0.18%, and the industry average was 0.02%. After hedging, the industry average was 0.01%. These numbers are proxies for institutional liquidity providers, not actual market returns. The study found that when the VIX is high, industry reversals improve. The costs of accessing the market and the bid-ask bounce are important factors. The portfolio consists of 48 industries, not 9 ETFs. The source of this information is the author of the study, which is based on historical evidence from before 2026, not a forecast for 2026. https://www.nber.org/system/files/working_papers/w17653/w17653.pdf

A news scan found the headline 'QTR9110200 sector60 correlations'. The following is a table of correlations between 9 sectors and the SPY, which is a type of stock market index, based on 60 daily returns. The table shows the correlation between each pair of sectors and the SPY, with 1 meaning they move together perfectly and -1 meaning they move opposite to each other. The correlations are as follows: XLF, which is a financial sector index, is correlated with itself at 1, with SPY at 0.351194, with XLB, which is a materials sector index, at 0.476564, with XLE, which is an energy sector index, at -0.264337, with XLK, which is a technology sector index, at -0.120289, with XLP, which is a consumer staples sector index, at 0.388275, with XLU, which is a utilities sector index, at 0.076927, with XLI, which is an industrial sector index, at 0.282287, with XLV, which is a healthcare sector index, at 0.354742, and with XLY, which is a consumer discretionary sector index, at 0.427342. The same correlations are shown for each of the other sectors and the SPY. This analysis was completed on September 10 and is based on 60 daily returns. It is not a stress correlation forecast, which is a type of forecast that predicts how correlations will change during times of market stress. https://runway.adityaperswal.workers.dev/data/stock-bars ; research/inference.json

A backtest was done on QTR9110200 with the following settings: it looked back 189 days, had a slip of 15 basis points, started on January 2, 2019, and ended on September 10, 2026. The return was 34.36 percent, the maximum drawdown was 30.46 percent, and the compound annual growth rate was 3.93 percent. The win rate was 35.26 percent, with 190 samples. The average trade return was 0.63 percent, with an average win of 8.10 percent and an average loss of 3.45 percent. The half Kelly criterion, a measure of bet size, was 0.04. The 95 percent confidence interval was between -0.44 percent and 1.76 percent. The test ran for 86 months. The reasons for trades ending were: 96 stops, 39 rank or time limits, 22 targets, 22 gap stops, 8 gap targets, and 3 sample ends. The test was reconciled with the vbt system, meaning the results were checked for accuracy. This was done using a daily proxy, without any untouched holdout data. The primary training was negative, and the baseline and stress trade counts differed because the barriers used anchor the actual slippage and adjusted entry points. research/sector_results.json ; https://runway.adityaperswal.workers.dev/data/stock-bars

A backtest of the QTR9110200 strategy, also known as cell 189_0.0015_2010, was conducted. This strategy had a lookback period of 189 days and a slip of 15 basis points, which is a measure of the difference between expected and actual trade prices. The test started on January 4, 2016, and ended on December 31, 2018. The return on investment was 8.57 percent, with a maximum drawdown of 12.05 percent, which is the largest decline in value during the test period. The compound annual growth rate, or CAGR, was 2.79 percent. The strategy was successful 36.36 percent of the time, out of 55 samples. The average gain per trade was 0.50 percent, with an average win of 5.84 percent and an average loss of 2.55 percent. The half Kelly criterion, a measure of optimal bet size, was 4.26 percent. The 95 percent confidence interval, which is a range of values within which the true value is likely to lie, was between -1.11 percent and 2.08 percent. The test ran for 24 months. The reasons for trades ending were: stop loss, which is a order to sell when a stock falls to a certain price, 29 times, target price, which is a predetermined price at which to sell, 13 times, rank or time, which refers to the strategy's ranking system or time-based rules, 7 times, gap stop, which is a stop loss order based on a large price gap, 4 times, and gap target, which is a target price based on a large price gap, 2 times. The test was reconciled using a method called vbt, and the results were not affected by a globally untouched holdout claim, which refers to a portion of the data that is set aside and not used in the test. The test used a daily proxy, which is a substitute for the actual data, and the primary training data had negative results. The baseline and stress trade counts differed because the barriers, or limits, used in the test anchored the actual slippage, or difference between expected and actual trade prices, adjusted entry prices. research/sector_results.json ; https://runway.adityaperswal.workers.dev/data/stock-bars

A backtest of the QTR9110200 strategy was conducted on cell 189_0.0005_2019. The test had a lookback period of 189 days and a slip of 5 basis points. It started on January 2, 2019, and ended on September 10, 2026. The return was 58.67414230918027 percent, with a drawdown of 28.703954898902715 percent. The compound annual growth rate was 6.20365116942454 percent, and the win rate was 36.17021276595745 percent. The sample size was 188, with a mean trade percentage of 0.8903299276103624. The win percentage was 8.33375787147409, and the loss percentage was 3.327612573912417. The half Kelly criterion was 0.05341707434637041, and the 95 percent confidence interval was between -0.1883234148356328 and 2.061320364504274 percent. The test ran for 86 months. The reasons for trades were: stop loss at 94 times, rank or time at 39 times, target at 24 times, gap stop at 22 times, gap target at 7 times, and sample end at 2 times. The test was reconciled using vectorized backtesting, and the result is considered valid. The test used a daily proxy and did not claim a globally untouched holdout. The primary training was negative, and the baseline and stress trade counts differed due to barriers anchoring actual slippage and adjusted entry points. research/sector_results.json ; https://runway.adityaperswal.workers.dev/data/stock-bars

A backtest was done on the QTR9110200 strategy with the following parameters: it looked back 189 days, had a slip of 5 basis points, and ran from January 4, 2016, to December 31, 2018. The return was 11.51%, the maximum drawdown was 11.80%, and the compound annual growth rate was 3.71%. The strategy won 37.04% of the time, with 54 samples. The average trade return was 0.69%, with winning trades averaging 5.98% and losing trades averaging -2.42%. The half Kelly criterion was 0.06, and the 95% confidence interval was between -0.96% and 2.35%. The strategy traded for 24 months. The reasons for trades were: 28 stops, 13 targets, 6 rank or time, 4 gap stops, and 3 gap targets. The backtest was reconciled with the vbt system. This was done using the Frozen2124 method, with daily proxy and no globally untouched holdout claim. The primary training was negative, and the baseline and stress trade counts differed because barriers anchor actual slippage adjusted entry, which means that the strategy's performance was affected by the way it adjusted for real-world trading costs. research/sector_results.json ; https://runway.adityaperswal.workers.dev/data/stock-bars

A backtest of the QTR9110200 strategy was conducted with the following parameters: a lookback period of 252 days, a slip of 15 basis points, and a time frame from January 2, 2019, to September 10, 2026. The results showed a return of 70.31861683669788 percent, a maximum drawdown of 21.85772985329847 percent, and a compound annual growth rate of 7.188703873977653 percent. The strategy had a win rate of 38.421052631578945 percent, with an average trade return of 1.0117095995117495 percent. The average win was 8.237980478839305 percent and the average loss was 3.4969893251968966 percent. The half Kelly criterion was 0.06140519524842902. A 95 percent confidence interval was calculated as between -0.048485473250310884 percent and 2.130268206228115 percent. The strategy was stopped 88 times, and other reasons for stopping included rank or time, target, gap stop, gap target, and sample end. The backtest was reconciled using vbt, and the results are based on a daily proxy, with no untouched holdout claim. The primary training was negative, and the baseline and stress trade counts differed due to barriers anchoring actual slippage and adjusted entry points. research/sector_results.json ; https://runway.adityaperswal.workers.dev/data/stock-bars

A backtest of the QTR9110200 strategy was conducted with the following parameters: a lookback period of 252 days, a slip of 0.0015, and a start date of 2016-01-04. The test ended on 2018-12-31. The strategy had a return of -7.73%, a drawdown of 13.94%, and a compound annual growth rate of -2.65%. The win rate was 33.33%. The average trade return was -0.51%. The average win was 4.49% and the average loss was -3.01%. The half Kelly criterion was 0. The 95% confidence interval was between -2.05% and 0.97%. The test ran for 23 months. The reasons for trade exits were: 27 due to stop, 8 due to rank or time, 6 due to target, 5 due to gap stop, and 2 due to gap target. The backtest was reconciled. The Frozen2124 model was used with daily proxy, but no claim of globally untouched holdout was made. The primary training was negative, and the baseline and stress trade counts differed due to barriers anchoring actual slippage adjusted entry, which refers to a method of accounting for the difference between expected and actual trade entry prices. research/sector_results.json ; https://runway.adityaperswal.workers.dev/data/stock-bars

A backtest of the QTR9110200 strategy was conducted with the following parameters: a lookback period of 252 days, a slip of 0.0005, and a time frame from January 2, 2019 to September 10, 2026. The results showed a return of 99.35%, a maximum drawdown of 19.23%, and a compound annual growth rate of 9.41%. The strategy had a win rate of 39.68%, with an average trade return of 1.27%. The largest win was 8.36% and the largest loss was 3.40%. The half Kelly criterion, a measure of optimal bet size, was 0.076. The 95% confidence interval for the returns was between 0.22% and 2.37%. The strategy was stopped 87 times due to stop loss, 38 times due to rank or time, and 28 times due to target. A news scan found the note Frozen2124, which mentioned that the daily proxy had no globally untouched holdout claim. The primary training was negative, and the baseline and stress trade counts differed due to barriers anchoring actual slippage adjusted entry, which refers to a method of adjusting for the difference between expected and actual trade entry prices. research/sector_results.json ; https://runway.adityaperswal.workers.dev/data/stock-bars

A backtest of the QTR9110200 strategy was conducted with the following parameters: a lookback period of 252 days, a slip of 0.0005, and a start date of 2016-01-04. The test ended on 2018-12-31. The strategy had a return of -1.8866 percent, a drawdown of 13.1154 percent, and a compound annual growth rate of -0.6345 percent. The win rate was 38.2979 percent, with a mean trade return of -0.0853 percent. The strategy had 47 samples, with a win percentage of 4.4972 and a loss percentage of 2.9296. The half Kelly criterion was 0. The 95 percent confidence interval was between -1.7106 and 1.6162 percent. The test ran for 23 months, with 24 stops, 9 rank or time exits, 7 target exits, 4 gap stops, and 3 gap target exits. The results were reconciled using vbt. The test used a daily proxy and did not claim a globally untouched holdout. The primary training was negative, and the baseline and stress trade counts differed due to barriers anchoring actual slippage and adjusted entry points. research/sector_results.json ; https://runway.adityaperswal.workers.dev/data/stock-bars

A news scan found the headline 'QTR9110200 crypto momentum census 24'. The data shows information about the cryptocurrency XTZ/USD. It had 406 observations, was eligible, and its last price was recorded on September 10, 2026, at 0.253173. The momentum over 12 months was -0.7368270095853994 and over 90 days was 0.08156844149103182. The volume over 60 days was 0.5731229127357672. The price was not above the 200-day average and did not break out. On September 10, the UTC day was completed. The data is based on 365 observations over 30 days, with a 90-day return and a 60-day volume calculation. Some calendar bars are missing and were not reconstructed for USDC/USDT. Stablecoins were excluded from the momentum candidates. All 23 eligible scores were negative. PAXG and MKR had insufficient scores, with 211 and 36 respectively. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data on September 10, 2026: there were observations of three cryptocurrencies, BCH/USD, ETH/USD, and UNI/USD. The data for each cryptocurrency is as follows: BCH/USD had a close price of 223.864, a 12-1 month momentum of -0.6322113853880005, a 90-day momentum of 0.10796337540212808, and a 60-day volatility of 0.8970288935694256, with the price not being above the 200-day average, and the prior 20-day high being 305.597, with no breakout. ETH/USD had a close price of 2436.8075, a 12-1 month momentum of -0.567528197698111, a 90-day momentum of 0.4637062865655146, and a 60-day volatility of 0.5993279082892086, with the price being above the 200-day average, and the prior 20-day high being 2565.9, with no breakout. UNI/USD had a close price of 5.950935, a 12-1 month momentum of -0.6146501068376069, a 90-day momentum of 1.3818030818491098, and a 60-day volatility of 0.9594593433828843, with the price being above the 200-day average, and the prior 20-day high being 7.4795, with no breakout. This data was collected over 365 days, with 30 days of observations used to form and skip, 90 days of observations used for return, and the square root of 365 used for 60-day volatility. The data for USDC/USDT was not reconstructed due to missing calendar bars, and stablecoins were excluded from the momentum candidates. All 23 eligible scores were negative, and PAXG and MKR were insufficient, with scores of 211 and 36 respectively. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data: on September 10, 2026, a crypto momentum census was completed. The census included the following cryptocurrencies: USDT/USD, XRP/USD, and YFI/USD. The details for each cryptocurrency are as follows: USDT/USD had a close price of 0.999435, a 12-1 month momentum of -0.0010269722717486385, a 90-day momentum of 0.0005446044666257421, and a 60-day volatility of 0.15988154397494692. XRP/USD had a close price of 1.334163475, a 12-1 month momentum of -0.6574382866048836, a 90-day momentum of 0.17880832574947636, and a 60-day volatility of 0.7064905966060278. YFI/USD had a close price of 2122.087, a 12-1 month momentum of -0.6426047134793209, a 90-day momentum of 0.12395699266438931, and a 60-day volatility of 0.4696606332490995. The census also noted that all 23 eligible scores were negative, and PAXG and MKR were insufficient with scores of 211 and 36, respectively. Momentum, which is the rate of change of a cryptocurrency's price, was used to evaluate the cryptocurrencies. Volatility, which is a measure of the spread of a cryptocurrency's price, was also considered. The census excluded stablecoins, which are cryptocurrencies with a fixed value, and did not reconstruct missing calendar bars for USDC/USDT. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data: the automated trading fund manager reviewed the momentum of several cryptocurrencies on September 10, 2026. The review included LTC/USD and PEPE/USD. For LTC/USD, the last price was $52.1099 on September 10, 2026, with a 12-1 month momentum of -0.6136 and a 90-day momentum of 0.2099. The 60-day volatility was 0.4706, and the price was above the 200-day average. The prior 20-day high was $59.3436, but there was no breakout. For PEPE/USD, the last price was $0.000003235 on September 10, 2026, with a 12-1 month momentum of -0.7306 and a 90-day momentum of 0.1533. The 60-day volatility was 0.9975, and the price was not above the 200-day average. The prior 20-day high was $0.00000455, but there was no breakout. The review also mentioned MKR/USD, which was not eligible. The review covered 365 observations over 30 days, with a 90-day return and 60-day volatility calculated as the square root of 365. Calendar bars were not reconstructed for USDC/USDT, and stablecoins were excluded from the momentum candidates. All 23 eligible scores were negative, and PAXG and MKR had insufficient data, with 211 and 36 observations respectively. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data: on September 10, 2026, a crypto momentum census was completed. The data included the following information for three symbols: SUSHI/USD, TRUMP/USD, and USDC/USD. For each symbol, the data included the number of observations, eligibility, last observation date, closing price, and various momentum and volatility metrics. The metrics included mom12_1, which refers to the momentum over the past 12 months, mom90, which refers to the momentum over the past 90 days, vol60, which refers to the volatility over the past 60 days, and above200, which indicates whether the price is above the 200-day moving average. The data also included the prior 20-day high and whether a breakout occurred. The results showed that all 23 eligible scores were negative, and PAXG and MKR had insufficient scores. The data was collected over 365 days, with 30 observations used to form and skip, and 90 observations used to calculate returns and volatility. The vol60 metric was calculated using the square root of 365. Additionally, missing calendar bars were not reconstructed for USDC/USDT, and stablecoins were excluded from momentum candidates. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data on September 10, 2026: there were observations of the cryptocurrencies LINK/USD and SOL/USD, which are eligible for momentum analysis, a method to determine the speed of price changes. The data for LINK/USD included a close price of 11.442, a 12-1 month momentum of -0.6277831289514999, a 90-day momentum of 0.4556233342450593, and a 60-day volatility of 0.680015858093114. The data for SOL/USD included a close price of 98.6025, a 12-1 month momentum of -0.659701858918569, a 90-day momentum of 0.4762068733204081, and a 60-day volatility of 0.5360770659171986. Both were above their 200-day moving average, a measure of the average price over 200 days. The prior 20-day high for LINK/USD was 13.677 and for SOL/USD was 110.5357, and neither had a breakout, which is a price increase above a previous high. The data also mentioned that PAXG/USD had 211 observations and was not eligible. The analysis was completed on September 10, 2026, and used 365 observations over 30 days, with a 90-day return and a 60-day volatility calculated as the square root of 365. The data did not include USDC/USDT because it is a stablecoin, a type of cryptocurrency with a stable value, and its missing calendar bars were not reconstructed. All 23 eligible scores were negative, and PAXG/211 and MKR/36 had insufficient data. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data: the automated trading fund manager reviewed the momentum of several cryptocurrencies on September 10, 2026. The manager looked at three cryptocurrencies: GRT/USD, SHIB/USD, and CRV/USD. For each one, the manager calculated the momentum, which is a measure of how fast the price is changing, over 12 and 90 periods, with the results being -0.8590795968384823 and -0.11257765531062114 for GRT/USD, -0.6562021439509954 and 0.03925619834710736 for SHIB/USD, and -0.6549541196406305 and 0.42321910915735406 for CRV/USD. The manager also looked at the volume, which is the amount of trading, over 60 periods, with the results being 0.732524162517283 for GRT/USD, 0.8699344166073065 for SHIB/USD, and 0.9647620727979849 for CRV/USD. The manager checked if the price was above the 200-day average, which is a common benchmark, and found that it was not for GRT/USD and SHIB/USD, but was for CRV/USD. The manager also checked for a breakout, which is when the price suddenly increases, and found that it did not happen for any of the three cryptocurrencies. The manager completed the review on September 10, 2026, and used 365 observations over 30 days to calculate the momentum, with a 90-day return and a volatility, which is a measure of how much the price changes, of vol60 sqrt 365. The manager did not reconstruct missing calendar bars for USDC/USDT and excluded stablecoins from the momentum candidates. All 23 eligible scores were negative, and PAXG and MKR were insufficient with scores of 211 and 36 respectively. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data: on September 10, 2026, a crypto momentum census was completed. The data included the following information for three cryptocurrencies: BTC/USD, DOGE/USD, and DOT/USD. For each cryptocurrency, the data included the symbol, the number of observations, eligibility for momentum calculation, the last observation date, the closing price, and various momentum and volatility metrics. The data showed that all 23 eligible scores were negative, and PAXG and MKR had insufficient scores. The census excluded stablecoins and did not reconstruct missing calendar bars for USDC/USDT. The data was based on 365 observations over 30 days, with a 90-day return and a volatility calculation based on the square root of 365. https://runway.adityaperswal.workers.dev/data/crypto-bars

A news scan found the following data: on September 10, 2026, a crypto momentum census was completed. The census included data on the following cryptocurrencies: AAVE/USD, which closed at 120.5505, AVAX/USD, which closed at 7.421, and BAT/USD, which closed at 0.0704568. The momentum, or rate of change, of these cryptocurrencies over 12 and 90 periods was also tracked, with mom12_1 and mom90 referring to the momentum over the last 12 and 90 periods, respectively. The vol60, or volatility over 60 periods, was also measured. The census also noted whether each cryptocurrency was trading above its 200-period average, with above200 referring to this status. Additionally, the prior 20-period high, or the highest price over the last 20 periods, was recorded for each cryptocurrency, as well as whether it had experienced a breakout, or a sudden increase in price. The census used 365 observations over 30 formations, with a 90-observation return and vol60 calculated as the square root of 365. The census excluded stablecoins, such as USDC/USDT, and did not reconstruct missing calendar bars. All 23 eligible scores were negative, and PAXG and MKR were insufficient, with scores of 211 and 36, respectively. https://runway.adityaperswal.workers.dev/data/crypto-bars