Portfolio managers at Bridgeway are encouraged to have capability in data analysis, financial research, and implementation, with junior staff focusing more on data understanding and senior staff focusing more on idea generation and literature review.

Price-to-Book has shifted from being an alpha factor to functioning as a risk factor in portfolio construction, particularly after 2007, and this shift is consistent with the growth in intangibles observed in their research.

Because sentiment changes faster than fundamental changes, high-intangible stock portfolios have higher expected turnover compared to low-intangible portfolios.

Classical measurements for valuation and quality have degraded significantly in their ability to be used for stock picking because intangible assets now comprise a much larger portion of company value that traditional accounting frameworks fail to capture.

Bridgeway uses machine learning with singular value decomposition (SVD) or similar techniques to identify latent systematic risks in higher-frequency price data across 35 countries, enabling detection of geopolitical risks and trade-related exposures that classical risk models miss.

Bridgeway does not hedge out currency exposures in their long-only international strategies because hedging is expensive and if investors want to diversify, they should diversify into both currencies and international equities.

Examples of intangible assets include intellectual property (Google's PageRank algorithm), proprietary algorithms (TikTok's recommendation engine), brand equity (Coca-Cola, Nike, Disney), and customer relationships (Amazon, Costco), all of which are material to business value but very difficult to value precisely.

Capitalized intangible assets as a proportion of total assets doubled from 1994 to 2018, while R&D expenses increased 50% as a proportion of total revenue, but SG&A expenses remained relatively stable over the period.

Intangible intensity has grown more for international and emerging market companies compared to the US, despite the US having a higher starting point, which was a surprising finding.

When model assumptions fail or externalities occur, Bridgeway mutes stock selection signals by downgrading stocks: for example, meme stocks, M&A activity, and government regulation all cause them to downgrade selections because fundamental analysis becomes unreliable in these contexts.

For low-intangible industries, fundamentals (ROE, book value, earnings) continue to explain price action quite well and have stable explanatory power, but for high-intangible industries, fundamentals have much lower explanatory power in predicting future price action.

Bridgeway uses completely different gearing and stock selection metrics for high-intangible versus low-intangible stocks across global and emerging market strategies, rather than using the same metrics with industry adjustments.

When using sentiment-based analysis for high-intangible stocks, it is critical to identify which sentiment is already priced into the stock and which is not, because investment opportunities exist only in unprice sentiment.

Bridgeway is not currently using machine learning or AI for stock selection, but is using machine learning for risk modeling, with natural language processing on quarterly earnings calls as an active area of research for 2024-2025.

A major challenge with using large language models (LLMs) like ChatGPT for historical analysis is look-ahead bias: an LLM trained in 2024 knows what happened in 2020-2022, making it impossible to trust what it tells you about what it would have done in 2020 without contamination from future knowledge.

A solution to the look-ahead bias problem with LLMs is to develop a specific accounting taxonomy and apply NLP techniques carefully to maintain temporal integrity in backtesting.

An absolute return strategy differs from a market-neutral strategy in that market-neutral aims for zero beta backward-looking, but beta is not predictive forward-looking, so absolute return strategies are better characterized as agnostic to market direction.

Bridgeway's absolute return strategy maintains equal amounts of longs and shorts, typically net-zero at rebalancing with 100% longs and 100% shorts (gross 200%), and gross exposure is proportional to stock selection efficacy, varying by geography and market cap.

Human beings can be better than computers in systematic quantitative investing by understanding and questioning the assumptions underlying their models, and by recognizing when those assumptions fail in real-world conditions.

Bridgeway is very careful evaluating model assumption failures on the short side of portfolios because shorts have asymmetric loss potential compared to longs, with potential losses that can be quite substantial.