McKinsey Sea Wolf Scoring Explained: The 5 Conditions
McKinsey is unusually transparent about how the Sea Wolf game is scored in its final phase. Unlike most of the Solve assessment — where scoring criteria are undisclosed — the Phase 4 Treatment scoring formula is publicly documented and consistent across all scenarios.
Understanding exactly how it works, and what each condition actually requires, changes how you approach every phase of the game. This article covers the five conditions, how each one is evaluated, and what the scoring logic means in practice.
The basic structure
In Phase 4 — Treatment — you select 3 microbes from a pool of 10. Your selection is evaluated against five binary conditions. Each condition is worth 20% of the phase score. Satisfy all five and you score 100%. Miss one and you score 80%. Miss two and you score 60%. And so on down to 0%.
There are no partial marks. Each condition is either fully met or not met at all.
| Conditions met | Score |
|---|---|
| 5 of 5 | 100% |
| 4 of 5 | 80% |
| 3 of 5 | 60% |
| 2 of 5 | 40% |
| 1 of 5 | 20% |
| 0 of 5 | 0% |
The five conditions
Condition 1 — Mean of attribute 1 in range
The average value of the first attribute across your three selected microbes must fall inside the site's target range. Each attribute range is exactly 2 points wide (e.g. 2–4, 5–7, 8–10) on a 1–10 scale.
This is evaluated on the mean, not on individual values. A microbe with an attribute value well outside the range is not disqualified — it contributes a number to an average. Whether that average lands in range depends on what all three microbes contribute together.
Condition 2 — Mean of attribute 2 in range
Identical logic to Condition 1, applied to the second attribute. The mean of your three microbes' values for attribute 2 must fall inside its target range.
Condition 3 — Mean of attribute 3 in range
Identical logic, applied to the third attribute.
Three separate conditions, one for each attribute — three separate opportunities to score or lose 20%.
Condition 4 — At least one microbe has the desired trait
At least one of your three selected microbes must carry the site's desired trait. Just one is enough — you don't get extra credit for two or three. Having multiple desired-trait microbes in your selection satisfies this condition exactly as well as having one.
This is where a significant portion of candidate errors occur: treating the desired trait as the primary selection criterion, stacking desired-trait microbes in the pool, and discovering in Phase 4 that those microbes don't combine well on the attribute conditions.
Condition 5 — No microbe has the undesired trait
None of your three selected microbes may carry the site's undesired trait. One microbe with the undesired trait fails this condition entirely — regardless of how well it performs on the attribute conditions.
This is the only elimination-style condition in the scoring formula. The others require you to achieve something. This one requires you to avoid something.
The arithmetic: working with means
Because conditions 1–3 evaluate averages, the useful unit of analysis is the sum of the three values, not the individual values. If the site range for an attribute is 4–6, the sum of your three microbes' values for that attribute must land between 12 and 18 (3×4 and 3×6).
This makes partial-selection reasoning fast. If you've chosen two microbes with attribute values of 3 and 5, their partial sum is 8. The third microbe needs to bring the total between 12 and 18, meaning its value for that attribute must be between 4 and 10. A quick subtraction tells you what you need from the third microbe — you don't need to recompute averages from scratch each time.
The same logic applies when a microbe's value seems too extreme. A value of 9 in a 4–6 range (target sum 12–18) is not automatically a bad pick. If your other two microbes have values of 2 and 4, the total is 15 — right in range. The extreme value is only a problem if it pushes the sum outside the target window.
The most important implication: all five conditions are equal
Each condition is worth exactly 20%. The desired trait is worth the same as each attribute condition. The undesired trait condition is worth the same as each attribute condition.
This has a direct consequence for how to make trade-offs:
- Sacrificing an attribute condition to include an extra desired-trait microbe is a breakeven trade at best — you lose 20% on the attribute, gain nothing extra on the trait (you only need one)
- Including a microbe with the undesired trait because its attributes are strong loses you 20% on condition 5, guaranteed, regardless of how much it helps on conditions 1–3
- Giving up one condition deliberately to secure two others is sometimes correct — a combination that satisfies 4 of 5 conditions (80%) beats a flawed attempt at all 5 that falls short on two (60%)
Not all pools can score 100%
McKinsey's pool generation occasionally produces scenarios where no combination of three microbes from the pool satisfies all five conditions simultaneously. In these cases the maximum achievable score is 80% or 60%.
This is intentional. It tests whether candidates can recognise that a perfect answer doesn't exist, identify the best available compromise, and commit to it — rather than continuing to search for a 100% combination that isn't there.
If you've evaluated the most promising combinations and can't find a five-condition solution, that's the signal to stop searching for perfection and find the combination that fails the fewest conditions instead.
How the phase score feeds the overall score
Phase 4 is one of five scored phases per site, across three sites. Your overall Sea Wolf score is a composite across all phases and all sites — not Phase 4 alone. Strong Phase 4 scores matter, but phases 1, 2, 0, and 3 all contribute to the total.
Phase 3 in particular directly affects Phase 4: the pool you build in Phase 3 determines what you have to work with in Phase 4. A poorly built pool caps what's achievable in Phase 4 before the phase even begins.
For a complete breakdown of how to build the optimal Phase 3 pool and approach Phase 4 selection strategically, see the full Sea Wolf strategy guide.
Quick reference
| Condition | What's required | Common mistake |
|---|---|---|
| Attribute 1 mean in range | Average of 3 values falls in 2-point range | Filtering by individual values instead of means |
| Attribute 2 mean in range | Same | Same |
| Attribute 3 mean in range | Same | Same |
| Desired trait present | At least 1 of 3 microbes has desired trait | Stacking multiple desired-trait microbes at cost of attribute fit |
| No undesired trait | None of 3 microbes has undesired trait | Including a strong-attribute microbe with undesired trait |