Allan MacDonald - Decision Making in Sport

Episode 220: In this episode of Accelerate Podcast, host Nicola Graham is joined by Allan MacDonald — a decision-making specialist who has spent 15 years in elite sport before transitioning into government, and who holds a professional doctorate in judgmental forecasting and decision-making.

Allan's career has taken him across high-performance sport, public sector strategy, and academic research — giving him a rare cross-domain perspective on how decisions actually get made under pressure. His doctoral work focuses on the science of forecasting and human judgement, exploring why even experienced professionals fall short when it matters most.

At the centre of the conversation is a question every high-performance environment faces: how do we make better decisions — not just with better data, but with better thinking? Allan unpacks the gap between having information and knowing how to weigh it, and why expertise alone is no defence against the biases that quietly shape our choices.

The discussion also explores how the principles of judgmental forecasting and calibrated confidence translate beyond sport — into business, strategy, and any environment where decisions are made in the face of uncertainty.

Topics Discussed

• Decision making as a deliberate skill rather than an instinctive ability
• Why experience can negatively affect decision quality in professional sport
• Common decision-making failures observed in elite performance environments
• Contamination and groupthink in multidisciplinary team meetings
• Probability-based thinking as an alternative to consensus seeking
• Calibration: matching subjective confidence to actual accuracy
• Outside-to-inside thinking and the use of base rates in sport decisions
• Why more data does not lead to better decisions
• The role of AI and large language models in supporting decision making
• Pre-mortem analysis and structured tools for navigating uncertainty

Key Points

• Decision making is a skill that can be improved through deliberate practice, yet most practitioners in professional sport receive no formal training in it. Research confirms that the majority of high-performance staff rely solely on experience to develop their decision-making ability. However, experience is insufficient for most of the complex, novel decisions encountered in elite environments. Engaging with the decision-making literature, attending structured education, and applying specific tools are the approaches that produce measurable improvements — analogous to how technical skills are developed through targeted training rather than passive accumulation of hours on the job.

• Experience can actively impair decision quality because it encourages selective recall. Practitioners tend to remember situations that resolved well and apply the same decisions in future scenarios, despite materially different conditions such as a different athlete, different team dynamics, different external pressures, or different stakes. This retrospective bias means that prior success becomes a flawed template rather than a genuinely transferable lesson. The assumption that experience leads to better decisions is deeply embedded in performance sport culture and is contradicted by research showing that experienced experts frequently perform no better than chance when their forecasts are independently tracked and scored.

• Observation of performance meetings across Premier League, NFL, and NBA environments consistently reveals a small number of recurring decision-making failures. A dominant individual or pair of individuals leads the discussion, suppressing input from others whose views remain unspoken. Contamination occurs when a more experienced colleague shares opinions prior to or early in the meeting, causing junior staff to shift their independent assessments toward agreement. Consensus seeking — framing questions as whether everyone agrees — produces artificial unanimity and bypasses genuine deliberation. Additionally, data overload results in inconsistent processes across athletes, as different information is raised in different sequences for each discussion.

• Contamination and groupthink are related but distinct phenomena. Contamination occurs when pre-meeting communication or early disclosure from an experienced practitioner shifts the independently formed views of colleagues. A junior practitioner may hold relevant information but withhold it to avoid appearing to contradict a more knowledgeable senior. Groupthink, by contrast, arises from structural homogeneity within the team — staff selected to share similar philosophies, beliefs, and information access who lack the diversity of perspective required to identify alternative solutions. Both phenomena lead to premature convergence on suboptimal decisions, but they require different mitigation strategies to address effectively.

• Thinking in probabilities rather than absolute positions offers a more effective decision-making framework than consensus. When individuals independently assign numerical confidence levels to competing options and those estimates are averaged, the result is consistently more accurate than discussions that seek a single agreed answer. This approach reveals meaningful differences in certainty that consensus conceals — two people who mildly disagree at 45% versus 55% are very different from two people firmly at 90% versus 10%, yet both pairings appear as disagreements in a binary consensus model. Averaging independent probability estimates addresses this, particularly across a series of decisions such as a full competitive season.

• Calibration refers to the degree to which an individual's stated confidence matches their actual accuracy rate. A well-calibrated practitioner who expresses 80% confidence in a decision should be correct approximately 80% of the time. Research across approximately 150 practitioners in sport indicates that calibration is reasonably accurate at lower confidence levels but degrades substantially at higher ones. Practitioners expressing 90% or 100% confidence are frequently correct only around 54% of the time. This overconfidence at high certainty levels is a key decision risk, and awareness of it serves as a practical trigger to seek additional perspectives or actively challenge one's own reasoning before acting.

• Outside-to-inside thinking is a structured approach in which a decision is initially anchored to base rate data — what typically happens in situations of this type — before being adjusted for case-specific nuance. This sequence is important: commencing with the broader normative reference class prevents the inside view from dominating too early and introduces an evidence-grounded starting point. The adjustment then incorporates contextual factors such as the specific athlete, environmental conditions, and emerging information that the base rate cannot capture. Practitioners who apply this approach consistently produce more accurate forecasts than those who rely exclusively on contextual judgment without first establishing what the data show for comparable situations.

• Research from multiple domains demonstrates that increasing the quantity of data available does not produce better decisions beyond a relatively small threshold. Studies in which forecasters were given progressively more data points — five, ten, twenty, and forty — show that accuracy plateaus at approximately five data points. Beyond this threshold, additional data increases confidence without increasing accuracy, a divergence that introduces systematic overconfidence into the decision process. In professional sport, this manifests when meetings introduce extraneous metrics without pre-agreed criteria for what data will be consulted. Establishing decision criteria in advance, and only introducing disconfirming information after a tentative conclusion is reached, produces more consistent and reliable decision processes.

• Artificial intelligence and large language models offer practical support for decision making in performance sport, particularly for retrieving base rate information rapidly, facilitating pre-mortem analyses, and acting as a structured devil's advocate. AI can process large datasets to identify patterns and trends more efficiently than human analysts. However, the human practitioner retains a critical role in incorporating information that is not captured in collected data — athlete conversations, subtle movement changes, interpersonal dynamics, tactical context, and environmental factors. Research conducted across performance sport practitioners suggests that, as AI availability increases, the need for stronger human decision-making skills grows rather than diminishes, since the volume of data requiring interpretation also increases.

• Pre-mortem analysis exploits a cognitive principle called prospective hindsight, in which participants are placed in a realistic future scenario where the decision has already failed and asked to explain what went wrong. This approach generates between 20 and 40% more granular and specific issues than forward-looking risk identification. For simple, well-researched decisions with established guidelines — such as concussion protocols — pre-mortem adds little value, as clear policies already exist. For complex, novel, or one-of-a-kind decisions with no applicable research base, pre-mortem is substantially more effective than conventional risk assessment. Mapping the decision environment by complexity and available time is therefore a prerequisite for selecting the appropriate decision-making tools for each situation.

Sponsors

Gameplan Performance, Gameplan Performance is a rehab Project Management & Data Analytics Platform that improves operational & communication efficiency during rehab. Gameplan provides a centralised tool for MDT's to work collaboratively inside a data rich environment.

Hytro: The world's leading Blood Flow Restriction (BFR) wearable, designed to accelerate recovery and maximise athletic potential using Hytro BFR for Professional Sport.

VALD, makers of the Nordbord, Forceframe, ForeDecks and HumanTrak. VALD Performance systems are built with the high-performance practitioner in mind, translating traditionally lab-based technologies into engaging, quick, easy-to-use tools for daily testing, monitoring and training.