Darren 🥚🐣🕊️
@ReformedTrader
Followers
15K
Following
32K
Media
7K
Statuses
39K
Trend, value, quality, carry, and volatility... Strength is made perfect in weakness. See the pinned tweet for compound return generators.
Joined July 2011
1/ How do you maximize compound returns?. A. Don't use too much risk. B. Don't rely on bounce-backs. C. Appropriate risk depends on expected Sharpe ratios. D. Diversification increases Sharpes. E. Hedge the (relevant) tails. "Rule #1: Never lose money.".
1/ Fortune's Formula (William Poundstone). "Even unlikely events must eventually come to pass. Therefore, anyone who accepts small risks of losing everything eventually _will_ lose everything. The compound return rate is acutely sensitive to fat tails.".
13
61
315
The Stimulus Deception | Rob Arnott on the Unexpected Impact of Deficit Spending. "Rob walks us through why government stimulus often fails to deliver real growth and how decades of rising spending have shaped today’s economic environment.".
0
0
1
14/ Stimulus Does Not Stimulate.
Stimulus Does Not Stimulate (Arnott, Pickard). "Beyond 15% to 30% of GDP, government spending imposes a drag on economic growth. "Government debt issuance is correlated with reduced economic growth; short-term stimulus leads to long-term headwinds.".
0
0
0
Stimulus Does Not Stimulate (Arnott, Pickard). "Beyond 15% to 30% of GDP, government spending imposes a drag on economic growth. "Government debt issuance is correlated with reduced economic growth; short-term stimulus leads to long-term headwinds.".
1
0
4
It would be interesting to see if, . instead of using theory, . a researcher grouped anomalies using trading principles (these are value-like factors, these hedge the same kind of risk, etc.). and then looked for performance consistency across specifications within those groups.
0
0
1
14/
It would be interesting to see if, . instead of using theory, . a researcher grouped anomalies using trading principles (these are value-like factors, these hedge the same kind of risk, etc.). and then looked for performance consistency across specifications within those groups.
0
0
1
It would be interesting to see if, . instead of using theory, . a researcher grouped anomalies using trading principles (these are value-like factors, these hedge the same kind of risk, etc.). and then looked for performance consistency across specifications within those groups.
1/ Does Peer-Reviewed Research Help Predict Stock Returns? (Chen, Lopez-Lira, Zimmermann). "Mining accounting ratios led to cross-sectional predictability similar to peer-reviewed research. "Peer review may also systematically mislabel mispricing as risk.".
0
0
6
13/ Related reading:. Most Claimed Statistical Findings in Cross-Sectional Return Predictability Are Likely True. The Limits of p-Hacking: Some Thought Experiments. Andrew Chen (Rational Reminder podcast).
Andrew Chen: "Is Everything I was Taught About Cross-Sectional Asset Pricing Wrong?" (Rational Reminder podcast). "Andrew Chen delves into out-of-sample performance, replicating financial studies, publication bias, data mining, and false discovery rates.".
1
0
1
12/ "Though the samples are small, Tables 6 and 7 suggest that focusing on the most renowned predictors does not affect our results. B/M and momentum, the two predictors that academics deem to be most worthy of attention, perform similarly to data mining post-sample."
1
0
0
11/ "Peer review seems to uncover data-mined predictors with the very largest t-stats (Fig 2, panel b). "Excluding correlated predictors leads to lower post-sample returns, but the low-correlation benchmarks with higher t-stats still perform similarly to published predictors."
1
0
0
10/ "Data mining closely mimics risk-based and mispricing-based predictors but somewhat underperforms agnostic predictors. "A potential explanation is that many agnostic predictors are based on past returns, which are missing from the accounting ratios we use for data mining."
1
0
0
9/ "While the sample is small, point estimates imply the opposite of what is commonly believed. The mean normalized post-sample return is monotonically decreasing in modeling rigor. "One interpretation is that a talented theorist can justify nearly any empirical pattern."
1
0
0
8/ "Risk-based predictors decay more than other predictors, but the difference is only marginally significant. "(The figure gives the illusion of outperformance in the first few years post-sample, but the trailing 5-year return is not fully post-sample until month 60.)"
1
0
0
7/ "To study risk-based research, we categorize predictors as risk-based, mispricing-based, or agnostic. "We manually read each paper, identify a passage (typically taken from abstract, introduction, or conclusion) that summarizes the main argument, and categorize the passage."
1
0
0
6/ "Mining accounting ratios generates out of-sample returns, even if only using the sign of past returns. Tickers, in contrast, are uninformative. "More data mining does not necessarily mean worse OOS performance (the accounting dataset is 10x as large as the ticker dataset)."
1
0
0
5/ "Research and data mining lead to eerily similar event-time returns. "Results are similar if we limit the published predictors to those that focus on accounting data (Figure A.2). "It is difficult to reject the null that the research process is built off of data mining."
1
0
0
4/ "Data mining works, in part, by uncovering the same ideas found by peer review. "One may have thought that a deep understanding of financial economics is required to uncover these themes, but data mining could have uncovered these themes decades before they were published."
1
0
1
3/ "The predictors are, to a significant extent, distinct. More than 80% of pairwise correlations are below 0.25 in absolute value. "Many dozens of principal components are required to span 80% of total variance—though there is a non-trivial factor structure."
1
0
1
2/ "Out-of-sample returns monotonically increase in the in-sample return. "The decay is in the ballpark of the post-sample decay for published predictors (McLean and Pontiff (2016)). "Similarly, out-of-sample predictability is much weaker post-2004, though it still exists."
1
0
1
1/ Does Peer-Reviewed Research Help Predict Stock Returns? (Chen, Lopez-Lira, Zimmermann). "Mining accounting ratios led to cross-sectional predictability similar to peer-reviewed research. "Peer review may also systematically mislabel mispricing as risk.".
1
0
8
