By Grant Mercer
An educational guide for futures students.
If trading outcomes are as noisy as the companion article ("Noisy Feedback in Trading: Why Outcomes Mislead") describes, a serious question follows: how does anyone actually get better at this? Skill in commodity trading is not absorbed from experience automatically — it is built deliberately, through learning systems designed to work despite noisy feedback. This article expands on the learning material in Chapter 12 of Futures Under Fire.
Why experience alone doesn't create expertise
Research on expert performance — most associated with psychologist K. Anders Ericsson — shows what separates experts from experienced non-experts is deliberate practice: structured, feedback-rich effort aimed at specific weaknesses, sustained over years. Trading is hostile to casual "learning from experience" for three reasons: feedback is noisy, feedback is slow (commodity cycles run years while performance is judged annually), and the environment is non-stationary. Deliberate practice in trading has to be engineered.
The decision journal: making thinking visible
The single highest-leverage learning tool is also the simplest: the decision journal. Before entering a position, write down the thesis, key assumptions, invalidation conditions (what would prove the thesis wrong), and expected holding period. After exit, record what happened — and whether the outcome resulted from the thesis being right or from unrelated events. The journal defeats hindsight bias and creates the dataset that makes learning possible: without written theses, there is nothing to review except P&L — and P&L misleads.
Post-mortems: the after-action review
Borrowed from the U.S. Army's after-action review process, the structured post-mortem asks:
What was the original thesis, in the journal's own words?
Was the thesis correct? (Separate from: was the trade profitable?)
If profitable, was it the thesis or unrelated events?
If unprofitable, where exactly did the analysis fail — thesis, timing, sizing, execution?
What does this imply for the process, stated as a concrete change?
The key discipline: conduct post-mortems on winners as well as losers — with equal skepticism. The win that is never examined teaches nothing.
The pre-mortem: learning from failure before it happens
Decision researcher Gary Klein's pre-mortem: before committing, imagine it is a year later and the decision has failed spectacularly — then work backward to explain why. People are far better at explaining failures retrospectively than predicting them prospectively, and the imagined-failure frame licenses candor that optimism suppresses.
Bayesian updating: changing your mind by the numbers
Start with a prior: your best estimate of a strategy's expected performance before seeing new data, stated as a distribution. As live performance data arrives, update the prior according to how surprising the data is given the prior, producing a posterior. This answers the two questions traders get wrong most often: when is underperformance just noise? and when is it evidence the strategy is broken?
Walk-forward testing: backtesting that respects time
Walk-forward optimization: the strategy is trained on an expanding window of past data and tested on the next out-of-sample period, repeatedly — simulating what deployment would actually have felt like through time. A strategy that only works when optimized on the full dataset but fails walk-forward is overfitted to the past.
Regime-aware review: judging strategies against their design
A strategy designed for trending markets should not be judged during a range-bound year. Regime-aware review means classifying market conditions explicitly and evaluating each strategy against the regimes it was built for. Much of what looks like "strategy decay" is regime mismatch; much of what looks like "hot performance" is regime luck.
Institutional memory: learning that survives turnover
Individual traders learn; then they leave, and the learning leaves with them. The most mature trading organizations invest in knowledge management: structured, searchable records of post-mortems, market-event analyses, and regime studies, maintained over decades. In a high-turnover industry, the organization that documents learns faster than the organization that merely experiences.
Learning from others' blowups: the case-study library
Metallgesellschaft (1993) — roughly $1.3 billion lost on oil derivatives; stack-and-roll hedge mechanics and funding risk.
Amaranth Advisors (2006) — roughly $6 billion lost in natural gas; concentration, leverage, and seasonal patterns vs. laws of nature.
LME nickel (2022) — execution risk, clearing risk, and exchange intervention in one episode.
Negative WTI (2020) — storage, expiry mechanics, and the day "prices can't go below zero" broke.
Study each twice: once for what happened (mechanics), once for what participants believed beforehand (assumptions). The second reading is where the learning lives.
Putting it together: a personal learning system
A decision journal, maintained daily, with theses and invalidation criteria written before positions.
Post-mortems on significant trades, winners included, reviewed quarterly for patterns.
Pre-mortems before large commitments or new strategies.
Regime-tagged performance review, so strategies are judged against their design conditions.
A personal case-study library, built trade by trade and crisis by crisis.
Periodic Bayesian review of the strategies in use.
None of this is glamorous. All of it compounds.
Common misunderstandings
"Ten years of experience means ten years of learning." Only with structured feedback. Ten years of outcome-watching without journals or post-mortems is one year of experience repeated ten times.
"Studying theory is enough." Theory without a feedback system doesn't stick — the journal turns theory into self-knowledge.
"Post-mortems are for failures." The most dangerous unexamined trades are the lucky winners: confidence without competence.
Risk disclosure: There is a substantial risk of loss in trading commodity futures and options products. Losses in excess of your initial investment may occur. Past performance is not necessarily indicative of future results.