Position sizing and sell discipline rarely get the same airtime as asset allocation or manager selection, yet they influence long-term portfolio outcomes just as much. In many investment processes, these decisions sit in a grey area—left to ad-hoc judgment or “PM discretion”—without a formal framework. The result is often inconsistent risk-taking: allocations that grow too large relative to conviction, or positions that linger well past their expiry date. I’ve found that treating sizing and exit rules as core policy decisions, on par with allocation targets and liquidity management, produces more consistent results and reduces the influence of short-term emotion. This piece lays out how I think about turning conviction into risk-aware position sizes, limiting correlated exposures, and pre-defining exit triggers—so that sizing and selling are deliberate, not reactive.

Position sizing and sell discipline: core but under-structured

 In practice, these decisions often make or break outcomes, but many investment policies barely mention them. Allocators pour over asset allocation targets and manager selection, while position size and sell criteria live in the gray zone of “PM discretion” or ad-hoc judgment. This gap exists even though the lack of a formal sizing and exit framework can lead to systematic mistakes, like chasing recent winners and ditching underperformers at the worst times. In fact, the common habit of firing managers after a few bad years and hiring new “hot hands” has been shown to hurt returns. Too often, investment committees end up reacting emotionally – doubling down on crowd-favorite trades, or hanging on to a thesis well past its expiration – precisely because they never pre-defined how conviction translates to size or what signals will tell them to exit. In this piece, I’ll share how we’ve sought to bring strategic clarity and pragmatic rigor to these decisions, so that sizing and selling are as systematic as buying.

Convert conviction into risk‑based size

One of the first challenges allocators face is how to translate an abstract conviction level into a concrete position size. It’s not enough to say “we really believe in this strategy” and allocate by gut feel. Conviction must be balanced with risk: volatility, drawdown tolerance, and each position’s marginal contribution to portfolio risk. In our framework, a high-conviction idea earns a larger allocation only if it can be sized within the volatility and drawdown limits. This is where tools like volatility-adjusted sizing and marginal contribution to risk (MCR) come in. Rather than allocating equal capital to each idea or manager, we allocate based on how much risk each position adds or diversifies. For example, instead of spreading assets evenly, we might give more capital to a lower-volatility, uncorrelated strategy and less to a higher-volatility, correlated one. In risk budgeting terms, we “allocate based on marginal contribution to risk and reward” – maximizing return per unit of risk, not just return - as was highlighted in The Rising Relevance of Risk-Adjusted Returns in Portfolio Construction. The goal is that each position’s size reflects both conviction and its risk profile.

Correlation-aware sizing: avoiding overlap and hidden concentrations

Another trap for allocators is sizing positions in isolation and ending up with a portfolio that’s secretly concentrated because of correlated exposures. It’s easy to think you’re diversified when you have 10 different managers or asset classes – until a market event hits and you discover many of them were riding the same underlying risk factors. We’ve learned to size and cap positions with an eye on shared exposures, factor overlap, and even behavioral clustering among managers. For example, you might have an equity hedge fund, a credit fund, and a “diversified” income fund – and not realize all three are essentially short volatility or long credit risk. If left unchecked, you’d size each at, say, 10% of the portfolio, blissfully unaware that in a stress scenario they could all sink together.

To combat this, we incorporate correlation matrices and factor analysis into sizing decisions. If two allocations have a high correlation or load on similar risk factors, we treat them collectively when setting limits. Perhaps each on its own could merit 10%, but together we cap them at, say, 15% total because they represent one compound bet. The goal is to limit “diversification in name only.” We explicitly ask: “What could cause these positions to all falter at once?” and size accordingly. During portfolio construction, we often employ a shared exposure limit – for instance, ensure that all equity-like risk (across public equities, equity hedge funds, equity L/S credit, etc.) doesn’t exceed a certain fraction of the risk budget. This way, even if different sleeves have different labels, we’re not over-exposed to one outcome such as a broad equity downturn or a credit crunch.

Heuristics that work in practice

While quantitative risk models are essential, we also embrace simple heuristics as position sizing guardrails. These are rules of thumb that have stood the test of time in portfolio management. One such heuristic is volatility targeting within bands. For many strategies, we set an expected volatility range (say 8–12% annualized); if the strategy’s actual volatility drifts above the band, we trim the position to bring risk back in line, and if it drifts too low (perhaps indicating underutilized risk budget), we might add – provided the opportunity still warrants it. This prevents a position from quietly becoming much riskier (or tamer) than initially intended. It’s essentially a way to “position size on the fly” as risk fluctuates, keeping the portfolio’s risk allocation tuned.

Another enduring principle is the Kelly criterion – a formula from betting that tells you the theoretically optimal fraction of capital to wager given an edge. Pure Kelly sizing maximizes long-run growth but comes with high volatility and ruin risk if you’re wrong. In institutional practice, we rarely, if ever, size up to full Kelly bets. Instead, we use fractional Kelly as a guide. For example, if a Kelly calculation suggests a 20% position (a huge bet by institutional standards), we might take a fraction of that – say one-half or one-quarter Kelly – to stay on the conservative side. This aligns with the wisdom that even if the math says 20%, you probably shouldn’t put more than about 20–25% of capital into a single position because diversification and unknown unknowns trump theoretical optimization. Using a fractional Kelly approach cushions against estimation error and bad luck. It acknowledges that probabilities and forecasts are never certain; better to size a bit smaller and survive to play again tomorrow.

We also maintain practical boundaries that serve as circuit-breakers on enthusiasm. For instance, no single external manager gets more than X% of the total portfolio, no matter how much we like them – both to cap exposure and to maintain humility that anyone can stumble. Similarly, if an internally run strategy has, say, a 10% volatility target, we might stipulate it cannot exceed Y% of the portfolio’s risk budget. These guardrails echo the point above: even the strongest convictions live inside a risk-aware box. Such rules might seem blunt, but they endure because they protect us from ourselves. I’ve found that having preset position limits (e.g. “we never go above 15% in any one equity strategy” or “we keep at least 25 positions in the portfolio to avoid concentration”) forces discipline in good times, before hubris or greed can tempt us to over-allocate. They are the quiet heroes of portfolio management – simple, a bit boring, but very effective in the long run.

Pre‑commit your exit rules

If sizing is about building a position, sell discipline is about dismantling or reducing it—ideally in a pre-planned, unemotional way. We set exit rules at the point of entry because decisions made in euphoria or panic are rarely optimal. The first category is thesis-driven breakpoints: we document why we expect an allocation to make money and what would invalidate that thesis. If those breakpoints occur, we exit or reassess rather than invent new reasons to stay. The second is drawdown triggers: if a position falls beyond a set limit (e.g., 20% from peak or 10% from cost) and is underperforming, it prompts an immediate review. The aim isn’t a blind stop-loss, but to avoid the paralysis of hope and limit damage from sharply wrong calls.

We also employ time-boxed evaluations to avoid perpetual extensions for underperformers, forcing a decision after a defined period (say three years). On the positive side, we use “stop-in” scale-ups—the opposite of a stop-loss—adding to positions that are performing well and meeting pre-set milestones, such as a 10% gain while still far from our value target. These rules help counteract the reluctance to add to winners. All scale-ups are pre-planned and size-constrained to preserve balance, ensuring both exits and increases are grounded in a process rather than driven by emotion.

Upgrade vs. exit: diagnose edge decay versus noise

When a position underperforms or circumstances shift, we ask a simple but telling question: If we didn’t already own this, would we buy it today? A “no” points toward exit; a “yes” suggests re-underwriting the thesis – perhaps because the asset is cheaper or sentiment overly negative – rather than abandoning it. This can mean upgrading to a similar strategy with a stronger edge, or sticking with the existing one but adjusting expectations and sizing after a thorough review. The key is distinguishing between short-term noise and genuine edge decay. Every active strategy will have losing periods; firing managers solely for underperformance is often counterproductive, as studies show it tends to lock in losses and chase past winners. We look at forward indicators: Is the original edge intact? Have market or structural changes impaired the strategy, or is the setback cyclical? Factor and attribution analysis help clarify – e.g., a value-style slump may be noise, but arbitraged-away process or key team departures point to decay.

The upgrade vs. exit framework is essentially a decision tree. If the thesis is broken or risk exceeds tolerance, we exit. If not, and there’s a clearly superior use for the capital – higher expected return or lower risk for similar return – we rotate into that “upgrade.” If no better alternative exists and conviction remains high, we hold or even add, recognising the cycle. These decisions are debated in committee and documented: “Maintaining allocation to Fund X despite slump due to mean reversion potential” or “Redeeming Fund Y to fund Fund Z with superior alpha.” By framing choices explicitly as upgrade or exit, we avoid inertia; each underperformer is either sold, replaced, or affirmatively retained with rationale, bringing rigor to an area prone to emotion and bias.

Governance: make discipline durable

Even elegant frameworks fail without governance. Embed sizing and selling rules into the Investment Policy Statement or a companion risk policy so they outlast individuals and committee turnover. Codify risk budgeting principles, maximum risk or capital per position, and the requirement that every new investment include documented exit criteria. Put risk and sizing reviews on the standing agenda to catch drift early and to enforce triggers that have fired. Record decisions and rationales in minutes to create accountability and institutional memory.

Build a culture of post‑mortems after significant exits—especially painful ones—to refine triggers, thresholds, and team behaviors. Assign a persistent “risk champion” to challenge concentrations and test assumptions. Celebrate the wins that come from discipline (a capped position that didn’t sink the portfolio; a stop that prevented deeper loss; a timely rotation into a higher‑edge idea) so the organization internalizes that good process, not just good outcomes, is the benchmark of professionalism. Over time, these practices weave discipline into daily operations, turning sizing and selling from episodic debates into reliable strengths.

Conclusion

An allocator’s sizing and selling playbook ties conviction to controlled risk, monitors and limits shared exposures, relies on a few robust heuristics rather than over‑optimization, and commits in advance to what will trigger scale‑ups, trims, or exits. Portfolio‑level guardrails catch what slips through and force measured responses to stress. Governance cements all of this into policy, process, and culture so the approach survives market cycles and leadership changes. The result is a portfolio that avoids catastrophic overbets, gives great ideas enough weight to matter, and redeploys capital decisively as edges evolve—turning implementation details into a durable source of advantage.

 

Subscribe to our Newsletter