What the Ecosystem game actually is
The Ecosystem game replaced the original Imbellus ecosystem case and is now one of three mini-games in the 2026 McKinsey Solve assessment. Candidates almost universally call it "Sea Wolf" after the marine setting most variants use, but McKinsey's official framing is the Ecosystem game.
You're given a site (often a reef, kelp forest, or deep-sea variant), a set of desired traits the ecosystem should reward, and a pool of species each carrying their own traits, calorie production, calorie need, and predator-prey relationships. Your job is to assemble an 8-species ecosystem that fits the site's trait profile AND forms a valid food chain.
The game runs roughly 30 minutes inside the broader Solve session, which also contains Red Rock and Sustainable Future Lab. Of the three, Sea Wolf is the most stable across attempts and the most "drillable" — the underlying mechanic doesn't change, only the species and site combinations.
The 2026 format at a glance
The site
A marine environment with specific conditions (depth, temperature, current) and a list of 4–6 desired traits. The trait list is the scoring spec — memorise it before looking at species.
The species pool
Around 16–20 candidate species. Each has 3–5 traits, a calorie produced number, a calorie needed number, and what it eats / what eats it.
Calorie balance
Every predator must have at least one prey present whose calorie production meets the predator's calorie need. Producers (plants, plankton) anchor the chain.
The 8-species cap
You submit exactly 8 species — no more, no fewer. This forces hard trade-offs between trait coverage, food-chain validity, and apex diversity.
Variants differ in surface detail (which species, which site) but never in mechanic. If you understand one site well, you understand them all.
How scoring really works
McKinsey doesn't publish the scoring rubric, but the model the community has converged on — and that holds up across thousands of replays — has three layers:
- Trait fit. Your 8 species collectively must carry the site's desired traits. Coverage of each desired trait contributes positively.
- The 20% deduction per undesired trait. This is the part candidates miss. Every undesired trait a species brings to the ecosystem deducts roughly 20% from that species' contribution. A species that carries 2 desired traits AND 1 undesired trait often nets negative versus a cleaner alternative.
- Food-chain validity. If the predator-prey graph has dead ends or orphaned predators (no prey to eat), the ecosystem fails on viability regardless of how well the traits match.
The implication: a "perfect-looking" ecosystem with eight beautiful, capable species routinely scores worse than a leaner, plainer ecosystem with cleaner trait alignment. Most candidates lose 30–40 points to undesired-trait deductions they didn't notice.
The 5 mistakes that kill Sea Wolf scores
Over-stacking apex predators
Three large predators on one site looks ambitious and fails the calorie check almost every time. Each apex needs a viable prey base — adding a second or third apex without doubling the prey collapses the food chain.
Ignoring calorie balance
Each species has a calories-produced and calories-needed value. The chain has to net out. Most failed ecosystems aren't 'wrong' species — they're species with mismatched calorie math no one checked.
Picking on aesthetics
Cool-looking species — sharks, octopi, the dramatic ones — pull candidates off-trait. The site brief specifies which traits the ecosystem rewards. If a species doesn't carry one of those traits, it costs you 20% per undesired trait at scoring time.
Submitting late or in panic
Last-minute swaps without re-checking the food chain are how a 75% ecosystem becomes a 45% ecosystem. Lock your selection with 3–4 minutes left, then use the remaining time to verify, not to add one more 'interesting' species.
Brute-forcing without reading traits
Some candidates try every combination by trial. The clock punishes that approach. Read each species' trait list once, mentally bucket species as 'trait fit' or 'wrong site', and only then start assembling.
A worked example
Suppose the site is a temperate reef with four desired traits: cold-tolerant, schooling, bottom-dweller, filter-feeder. You scan the species pool and find a tempting candidate — call it Species X — that carries cold-tolerant and schooling (both desired), but also nocturnal-hunter (not in the desired list).
Naive read: 2 of 4 desired traits, looks like a strong pick. Real scoring: the 2 desired traits give you roughly +2 trait credits, but the 1 undesired trait deducts roughly 20% of that species' contribution. Net contribution is closer to +1.6 than +2.
Now compare Species Y, which carries only cold-tolerant and filter-feeder — fewer desired traits than X (2 vs 2), but zero undesired traits. Net contribution is a clean +2 with no deduction, AND it covers a different desired trait (filter-feeder), expanding your trait coverage across the 8-species ecosystem.
Pick Y. This single discipline — preferring clean trait coverage over flashy partial-fit species — is the most reliable score lift on Sea Wolf, and most candidates never internalise it because the game's UI rewards visual completeness, not arithmetic.
Pattern fluency: what to actually drill
Sea Wolf is a pattern game. There are three pattern muscles worth building, and they're the same three regardless of which variant you draw on test day:
- Trait reading speed. You should be able to scan a species card in under 10 seconds and bucket it as "site fit" or "site miss." This is pure reps — there's no shortcut.
- The predator-prey graph in your head. By minute 15 of the game you should hold the food chain as a small mental graph. Drill by sketching the chain on paper during practice attempts; on test day you'll do it in your head.
- Calorie arithmetic. Quick check: does the prey base produce enough calories for the predators? You don't need exact sums, you need a "feels right / feels short" intuition you can trigger in a few seconds.
These three together are what the 4-phase prep methodology calls "pattern fluency" — and they're the only thing that actually transfers from practice to the real assessment.
Runway: how to drill on 3, 7, or 14+ days
| Days | Focus | What to do |
|---|---|---|
| 3 days | Format orientation + 2 full attempts | Day 1: read this page and play 1 free attempt. Day 2: review scoring, play 1 paid attempt with full debrief. Day 3: light reps, sleep. |
| 7 days | Pattern fluency on traits and calorie math | 4–5 full Sea Wolf attempts spaced over the week. Between attempts, drill trait recognition flashcards and one 10-minute calorie balance exercise per day. |
| 14+ days | Full mastery alongside Red Rock + SFL | 1–2 Sea Wolf attempts per week, with the bulk of time going to Red Rock (the score-killer for most candidates). Use Sea Wolf to maintain pattern fluency, not as your main drill. |
The two most common allocation mistakes: spending all your time on Sea Wolf (because it's the most fun) and ignoring Red Rock (which is where most candidates actually lose points). See the tips page for full allocation guidance.
The Sea Wolf solver — when it helps, when it hurts
SolvePrep includes a Sea Wolf solver: paste in the site traits and species pool, and it brute-forces the highest-scoring 8-species ecosystem. It exists because reviewing your own attempts is far more valuable when you can compare what you submitted against what was actually optimal.
Use it for: post-attempt debriefs. Play a full attempt, submit, then run the solver on the same inputs and compare. The deltas tell you exactly which patterns you missed — usually the same 2–3 patterns repeating across attempts.
Don't use it for: live practice. If you solve every practice attempt with the tool, you skip the trait-reading and graph-building reps that actually transfer to the real assessment. The proctored Solve environment doesn't let you run a solver — only your own pattern fluency.
Frequently asked questions
Is the Sea Wolf / Ecosystem game still in McKinsey Solve in 2026?
Yes. The Ecosystem game (still nicknamed 'Sea Wolf' by candidates) remains one of the three core mini-games in the 2026 McKinsey Solve format, alongside Red Rock and Sustainable Future Lab. Format details have been refreshed but the underlying mechanic — pick 8 species that form a viable food chain — is the same.
How long is the Ecosystem game?
About 30 minutes within the larger Solve session. You won't get extra time for studying the map or species traits, so reading speed and trait recognition are themselves scored indirectly via the clock.
Can you retry or undo choices inside the Ecosystem game?
You can swap species in and out as much as you like inside the 30-minute window. There is no penalty for changes. There is no retry once submitted, and there is no retake of the game inside the same Solve session.
Does the order in which you pick species matter?
No. Only the final 8-species ecosystem you submit is scored. What matters is the combination — trait fit to the site, predator-prey validity, and calorie balance — not the order you added them.
Is using a Sea Wolf solver tool allowed?
Solvers are practice and review tools — you use them on your laptop while drilling, not during the real assessment. The McKinsey assessment is proctored and runs in its own environment. Solvers are valuable for understanding why a given ecosystem scores well, but treating them as a crutch during practice will hurt you on test day when you have to reason live.
Drill Sea Wolf on the 2026 format
The free tier gives you one full Sea Wolf attempt with the 2026 mechanics — enough for a clean diagnostic. Elite unlocks unlimited reps, the solver, and the same treatment for Red Rock and Sustainable Future Lab.