Palace/Workshop

Decision Quality vs Outcome

decision-makingprobabilityinvestinguncertaintyhoward-marks

Decision Quality vs Outcome

A good decision can produce a bad outcome, and a bad decision can produce a good outcome. Howard Marks returns to this idea through Nassim Taleb's "alternative histories," C. Jackson Grayson's Decisions Under Uncertainty, and Annie Duke's Thinking in Bets: the things that reasonably could have happened but did not.

The quality of a decision should be judged by the process and probabilities available before the outcome, not by hindsight alone. Outcomes are single draws from a distribution; they reveal what happened, not necessarily what was most likely or what was wise.

Evaluation Checklist

When judging a decision, ask:

  1. What information was available at the time?
  2. What alternatives existed?
  3. What probabilities could reasonably be assigned?
  4. What were the possible payoffs and losses?
  5. Was the process sound even if the result was bad?
  6. Was the result lucky even if the process was weak?

Why It Matters

Outcome bias makes people over-credit winners and over-blame losers. In investing, this can cause managers to abandon good processes after temporary losses, imitate lucky winners, or mistake a single success for skill.

This concept overlaps with active-management-as-error-detection because active investors must distinguish sound reasoning from lucky outcomes, and with ergodicity because survival depends on paths, not just average results.

In "You Bet!" (2020), Marks connects this to games: blackjack, backgammon, gin, bridge, poker, and betting all force decisions without complete information. The skilled player does not seek certainty. They know probabilities, update as evidence changes, size bets appropriately, and accept that even good process can lose on a single hand.

In "Nobody Knows II" (2020), he applies the same logic to COVID. The correct response was not to pretend to know the path of the virus or the market, but to separate facts from inferences and guesses, then act in stages as price improved relative to value.

Taleb's Alternative Histories

Taleb makes the same idea more severe in fooled-by-randomness. A decision must be evaluated against alternative-histories: the plausible paths that could have happened but did not. A trader who made money may have been exposed to a ruinous path that simply did not arrive during the observed sample.

This means outcome review should include the invisible distribution:

  • What could have gone wrong?
  • How bad was the tail?
  • Was the successful result robust or lucky?
  • Would the same process survive many repetitions?

art-of-trading-with-light-su-zhu-and-hasu restates the same lesson in the language of poker and trading. Pocket aces losing to kings does not make the all-in bad, and one profitable trade does not prove the reasoning was good. The point is to optimize the process and the odds, not to worship one realized path.

Timeframe Conflation

A closely related failure mode: a position starts as a trade with a defined thesis, but after moving against you, gets reclassified as a "long-term investment" to avoid realizing the loss. The reclassification feels seamless psychologically — you can always find a longer-horizon framing — but it is lethal in markets where returns are auto-correlated. Light in art-of-trading-with-light-su-zhu-and-hasu: "You start with a trade and then suddenly you become a bag holder. There's nothing more dangerous than having that happen to you because the human mind is pretty amazing in what it can rationalize."

The fix is to set the timeframe before entering, not after losing.

Nullification as the Right Approach

Most traders build a case for why their trade is right. The correct process runs in reverse: try to find all the reasons the trade is wrong first. If you cannot nullify the hypothesis, then trade with more confidence. This connects to inversion — the best test of a conviction is whether it survives serious pressure from the other side.

Light: "Whenever I have a trade idea that feels good, the way I actually spend time with it is by trying to figure out all the reasons why this is a terrible idea. And then if I can't come up with something conclusive, if I can't nullify the hypothesis, then I feel more confident." The counterpart exercise: put yourself in the shoes of the counterparty and ask why they are right and you are wrong. If you have not done that, you have no real basis for the trade.

Trade Journal as the Feedback Loop

Cognitive biases from volatility — loss aversion, anchoring to high-water marks, the emotional memory of wins and losses — make unassisted memory unreliable as a learning tool. A trade journal (entry thesis, invalidation level, size rationale, exit) is how a discretionary trader creates a record that can actually be reviewed.

Without it, you cannot do leak-finding: identifying which parts of your process are structurally costing you. Light's comparison: trading without a journal is like playing poker without hand histories. You keep making the same structural mistakes because you never see the pattern. Find the big leaks first — the spots where you are losing in obvious ways — before worrying about fine-grained edge cases.

Sources