Howard Marks
Oaktree Capital founder who redefined risk management through market cycles and second-level thinking
Howard Marks co-founded Oaktree Capital Management, which manages over $170 billion in assets and specializes in alternative investments. His investment philosophy centers on market cycles, risk perception, and second-level thinking. His Oaktree memos, written since 1990, are considered must-reads by top investors including Warren Buffett. His 2011 book 'The Most Important Thing' became a classic in value investing literature.
Describe risk in terms that can change a decision
Calling for caution does not finish a review. The following separates Marks’s discussion of risk from our service-capacity exercise. Name the unwanted outcome, then identify choices that depend on an unverified forecast.
This review covers the linked memo passages, analysis, FAQ, capacity exercise and replacement peer quotation. It does not verify every biographical, book, model or investment-performance claim, or assess current markets.
Fluctuation is not a complete definition of risk
Source record: The 2015 memo distinguishes fluctuation from the possibility of permanent loss.
Our interpretation: In our capacity exercise, a calm monitoring chart cannot establish that requests are preserved. Specify the unwanted outcome, such as acknowledging success without saving the request. Monitoring remains useful; calmness alone does not establish reliability.
One outcome cannot reconstruct the prior risk
Source record: The same memo argues that a profitable outcome alone does not establish the prior probability of loss.
Our interpretation: Our exercise records assumptions and stop conditions before release. After a quiet first launch, ask whether the decisive load and failure conditions occurred at all. An unencountered scenario remains unknown; one successful release has not ruled it out.
Distinguish a view from treating it as fact
Source record: The 2021 memo distinguishes having a view from assuming it is right and betting heavily on it.
Our interpretation: In our exercise, “we expect few calls” is a forecast, not a capacity test. Compare behavior under lower and higher load assumptions. Scenarios expose dependencies; listing them does not mean their probabilities have been measured.
Hypothetical exercise: can a new service expand access?
For this teaching example only, assume a service processes at most 100 requests per second and consider arrivals of 60, 120 and 240 per second. These numbers are assumptions, not measured capacity or traffic forecasts. No scenario probabilities are assigned.
Accepted requests need a clear outcome; unsaved requests must not be reported as successful. Requests beyond capacity need an understandable waiting, rejection or retry path.
Under the fixed-capacity assumption, 60 is within 100, while 120 and 240 exceed it by 20 and 140 each second. A queue alone cannot indefinitely absorb a sustained excess. Compare narrower access, explicit rejection of excess requests or tested capacity expansion. Do not expand commitments without evidence.
Use synthetic requests to check measured throughput, delays, errors and saved outcomes. Distinguish explicit rejection from silent loss and check the agreed handling of a replayed request. Record duration and limitations for each scenario. This is a proposed check list, not an existing test report.
Stop expansion if an unsaved request is acknowledged as successful or excess load produces no clear result. Have the service owner repair acceptance and rejection behavior and recheck it. A successful low-load run cannot substitute for an untested high-load case.
This is our engineering application, not a workflow personally authored by Marks or investment advice. Scenario counts and numbers do not form a probability distribution. Measure real capacity; this checklist cannot enumerate every risk.
Methodologies
- Market Cycle Positioning Checklist - Systematically observe credit conditions, valuation levels, investor sentiment, and capital flows to determine where the market sits in its cycle.
- Second-Level Thinking Question Framework - For any investment judgment, add one more layer of questioning: what does the market already know? Is the consensus already priced in?
Key decisions and timeline
- 1969 Joined Citibank, Beginning Credit Research Career - Seeking out inefficiently priced niches outside mainstream markets is an effective path to building differentiated competitive advantages.
- 1978 Joined TCW and Launched High-Yield Bond Fund - Building specialized capabilities early in emerging market niches can create competitive barriers lasting decades.
- 1990 Published First Investment Memo 'The Randomness of Investment' - Systematically recording and sharing ideas not only helps others but deepens and refines one's own thinking through repeated articulation.
Beliefs and mental models
- Risk Control Before Return Pursuit - Superior investing is not about maximizing returns but about achieving reasonable returns for a given level of risk; the ability to perceive risk is what separates ordinary from exceptional investors.
- Markets Are Always in a Cycle - Markets do not random-walk; they pendulum between fear and greed in cycles. Identifying where we are in the cycle is the most critical context for investment decisions.
- Second-Level Thinking Is the Source of Alpha - First-level thinking says 'this is a good company, buy it'; second-level thinking says 'this is a good company, but the market already knows it, so the price isn't cheap.' Alpha comes from thinking one level deeper than the consensus.
- Pendulum Thinking - Market sentiment swings like a pendulum between optimism and pessimism; extreme positions signal the best buying or selling opportunities.
- Second-Level Thinking Framework - For every investment judgment, ask one level deeper: what does the market already know? Is the consensus already priced in?
- Risk Perception Model - True risk is not volatility but the possibility of permanent capital loss; risk accumulates during market optimism and is released during pessimism.
Questions and answers
Does a calm monitoring chart establish low service risk?
Not from the chart alone. Low load in our exercise may never trigger the overload path. Check that requests are actually saved and failures accurately reported. Define the unwanted outcome before choosing evidence that could reveal it. Applying an investment discussion to software reliability is our interpretation, not a statement by Marks about software.
Why retain risk assumptions after a successful first launch?
One observed outcome does not reveal every path that was possible. Our recommendation is to retain the original capacity assumption, observation scope and stop conditions, then record which conditions actually occurred. A low-traffic launch cannot be described as a passed high-load test. Revisit commitments when new load or failure evidence appears instead of citing the first success.
Do three scenarios provide probabilities of risk?
No. Considering 60, 120 or 240 requests per second helps compare choices under different conditions; it does not assign each a one-third chance. Probabilities require stated data, a model and a scope of applicability. Leave them unknown without evidence. In this example, identify overload conditions first, then test rejection, waiting or capacity-expansion arrangements.
Sources
- The Most Important Thing: Uncommon Sense for the Thoughtful Investor, Howard Marks, 2011
- Mastering the Market Cycle: Getting the Odds on Your Side, Howard Marks, 2018
- Oaktree Capital Management Memos, Howard Marks, 1990-2024
- Risk Revisited Again — Howard Marks, June 8, 2015 — PDF pages 1–2: What Risk Really Means; paragraph beginning Forget about a priori
- Thinking About Macro — Howard Marks, July 29, 2021 — Opening discussion: macro forecasts; paragraph beginning Finally, I want to point out; paragraph beginning I’ll try below
- The Most Important Thing — Columbia University Press, 2011 — Reviews: Warren Buffett endorsement, first sentence