Probabilistic Preparation
Prepare for a range of futures without pretending to predict one.
- Difficulty
- Advanced
- Time to result
- ~ongoing to results
- Steps
- 6
- Confidence
- 97%
Probabilistic Preparation treats the future as a distribution of possible outcomes rather than a single event that can be confidently forecast. The decision-maker identifies plausible scenarios, assigns considered probabilities, estimates their consequences, and determines which outcomes would be unacceptable. Preparation concentrates on likely cases while selectively protecting against low-probability disasters whose damage would be unbearable. Because insuring against every possibility would consume all available resources, the method compares protection costs with expected value and survivability. The resulting decision may favor the most likely outcome, but position size, liquidity, diversification, or avoidance leaves room for alternatives. This distinction also improves evaluation: a good process can encounter an unlucky result, while a reckless decision can temporarily appear successful.
Origin
Marks connects this approach to his maxim, “You can't predict, you can prepare,” his reading of Peter Bernstein and Nassim Nicholas Taleb, and Oaktree's decisions during the 2008 crisis.
Core principles
- 01The future is a range of possibilities, not one knowable outcome.
- 02A forecast should express likelihood, not certainty.
- 03A sound decision can produce a bad result because outcomes contain randomness.
- 04Preparation must account for consequences as well as probabilities.
- 05Avoid exposures whose plausible downside is unbearable.
How to run it
- 1
Map Possible Outcomes
Describe the materially different ways the situation could unfold, including adverse alternatives to the base case. Keep the set broad enough to capture uncertainty but focused enough to guide action.
Pro tip Include at least one outcome that challenges your favored thesis.
Watch out Listing only minor variations of the same forecast creates false breadth.
- 2
Assign Considered Probabilities
Estimate the relative likelihood of each outcome using available evidence and explicitly acknowledge uncertainty in those estimates. Do not present the highest-probability case as certain.
Pro tip Use ranges when precise percentages imply more knowledge than you possess.
Watch out An unlikely event occurring does not by itself prove the probability estimate was unreasonable.
- 3
Measure Consequences
Estimate gains, losses, liquidity needs, and second-order effects for each scenario. Give special attention to outcomes with consequences that cannot be tolerated.
Pro tip Evaluate both probability and severity rather than ranking scenarios on likelihood alone.
Watch out Attractive average returns can conceal a small chance of ruin.
- 4
Set the Survival Constraint
Identify exposures that would make an adverse but plausible outcome unbearable. Avoid them, reduce their size, or add protection.
Pro tip Ask whether you could continue operating after the downside scenario occurs.
Watch out Do not accept ruin risk merely because the expected value appears positive.
- 5
Choose Selective Protection
Prepare for likely scenarios and insure against tail risks when the protection is worth its cost. Accept some risks whose likelihood and consequences are tolerable.
Pro tip Prioritize protection against outcomes that combine severe damage with limited recovery options.
Watch out Trying to insure against every imaginable event can consume all potential return.
- 6
Act Without Requiring Certainty
Make the best risk-adjusted decision while retaining allowance for alternative histories. Review the process separately from the eventual result.
Pro tip Document probabilities and reasoning before the outcome is known.
Watch out Do not assume that what should happen will happen promptly.
In the wild
After Lehman's collapse, investors feared that the financial world might end. Oaktree treated systemic failure as a possible disaster but concluded that distressed assets still offered sufficient value to invest. The firm committed capital while retaining cash and avoiding dependence on one perfectly timed recovery path.
→ Oaktree acted amid uncertainty without requiring assurance that the most favorable scenario would occur.
A portfolio manager finds a trade with favorable average returns but a small chance of losses large enough to force liquidation. Rather than relying on the base case, the manager reduces the position and purchases limited protection against the severe scenario.
→ The portfolio preserves upside while remaining able to survive an adverse tail event.
Common mistakes
Replacing Certainty with Vagueness
Saying that anything can happen without judging relative likelihoods avoids the responsibility to make a useful decision.
Insuring Against Everything
Protection has a cost, and attempting to cover every theoretical outcome can eliminate the resources or returns needed to succeed.
Judging Process Only by Outcome
Randomness means a sound probabilistic decision can lose and a reckless one can win temporarily.
Is it for you?
Best for
It is best for consequential decisions where several outcomes are plausible and randomness can separate decision quality from results.
Not ideal for
It is not ideal for deterministic problems whose outcomes can be reliably calculated from stable physical or contractual relationships.
From the transcript
“You can't predict, you can prepare.”
“We don't know what's going to happen in the future, but we have to prepare for the things that may happen.”
From the episode
Oaktree Howard Marks: Masters in Business (Audio)