Video summary

the skill university never taught you

Main summary

Key takeaways

Educational

Main ideas / concepts / lessons

  • Textbook knowledge is lagged: by the time a textbook is published, it’s typically at least ~1 year behind the field (arguably more). In fast-moving areas (e.g., sciences and machine learning), the real “current” knowledge often lives in the research literature, not textbooks.
  • University often underemphasizes how to read papers: many students may read a paper once for a class, but rarely learn heuristics for reading papers systematically.
  • Start by choosing the right paper type:
    • For learning “new theory,” focus on method development papers (a.k.a. method papers, method development, method dev).
    • Other useful types exist, but they serve different purposes:
      • Application papers (use methods in practice)
      • Reviews, including:
        • Theoretical reviews
        • Literature reviews
  • Research paper “anatomy” (common structure):
    • Introduction
    • Methods / methodology (sometimes explicit)
    • Results and Discussion
    • Conclusion (sometimes absent)

How to approach reading research papers (method, step-by-step)

Overall strategy

Build a framework to:

  • understand the paper’s structure,
  • judge quality,
  • extract what you personally want (e.g., ideas for projects),
  • and work up to deeper math when appropriate.

Step-by-step reading process

  1. Skim the paper first

    • Goal: identify “quality signals” to decide whether it’s worth time.
    • What to look for:
      • Figures
      • Paper structure (so you know what to expect)
      • Where to pause to develop ideas yourself
    • Extra skimming objective: identify mathematical “machinery” early.
      • Distinguish between basic math and complex math to avoid wasting time if you’re out of your depth.

  2. Read the introduction and conclusion

    • Purpose: check consistency and quality signals.
    • What to look for:
      • The intro should clarify the paper’s objective (“what are we going to do?”).
      • The conclusion should be consistent with the intro’s stated objective (“what was done?”).
      • Strong conclusions often mention shortcomings and/or future work.
    • Project ideation:
      • Search for phrases like “future work could…”—these often describe presented shortcomings and next steps as “low-hanging fruit.”

  3. Read the results + discussion more deeply

    • Often emphasized as where you engage with the core mathematical results (especially for method development papers).
    • What to do:
      • Build familiarity with the notation used by the authors.
      • Understand key mathematical results intuitively/qualitatively before fully deriving.
      • Track the flow of the paper as the authors’ thought process.
    • Outcome: understand “what the paper is doing” beyond surface reading.

  4. Go through the math derivations carefully

    • What to do:
      • Walk through derivations, key results, and assumptions.
      • Stress test assumptions.
      • If applicable, implement/code to validate.

Abstract handling (specific advice)

  • The speaker generally skips the abstract initially because it’s often jargon-heavy and reads like a “sales pitch”, assumes subfield familiarity, and may be useless if you’re new.
  • Exception: if you’re already deep in the subfield, the abstract can be quicker than the introduction.

Example walkthrough (used to demonstrate the framework)

  • Paper used as an example (described in the video):

    • “High frequency trading in a limit order book”
    • Presented as foundational (older; ~2007), with many citations (~700–800).

  • What the speaker did with it (mapped to the steps):

    1. Skim: identified sections like intro/model; noted core elements:

      • mid price modeled as Brownian motion
      • finite to infinite time horizon
      • limit order / trading intensity / Poisson process assumptions
      • derivation of optimal bid/ask quotes

    2. Read intro + conclusion: extracted the paper’s modeling goals:

      • dealers quote bid/ask to provide liquidity
      • risks include inventory risk and information/asymmetric information risk
      • the model derives optimal quotes accounting for inventory
      • mentions combining utility from established approaches with limit order book structure and econophysics-inspired order-arrival modeling
      • models arrival/execution likelihood as a function of distance from mid price

    3. Read deeper results/math: stepped through core concepts like:

      • reservation (indifference) bid/ask prices
      • how inventory shifts willingness to quote
      • dynamic programming / Hamilton–Jacobi–Bellman (HJB) framing
      • mapping value function ideas to optimal control intuition

    4. (Planned) deeper math: acknowledged full derivations can be time-consuming and would continue later (e.g., another video or livestream) if desired.

Time/effort expectation conveyed

  • Reading with math and derivations is time-consuming.
  • However, the framework is presented as manageable: early pages may require only:
    • basic algebra
    • intuition-building
    • taking notes step-by-step
  • The speaker suggests continuing later via another video or a live stream, depending on interest.

Speaker / sources featured

  • Speaker/Source: an unnamed narrator/content creator (the video has no explicit named host in the subtitles).
  • Authors / referenced researchers (mentioned, not interviewed):
    • Hoen Stole (referenced as having studied a “dealer” problem; cited in the example explanation)
    • “these same two people / these authors” (referenced but names are not clearly given in the subtitles)
    • Econophysics literature (referenced as a source category for modeling order arrival intensities)
    • “the microstructure literature” (referenced as a source category for dealer pricing and risk)

Original video