You are the project developer for Mozcon Power, awarded a five-year contract to supply a fast-growing island that is decarbonising while it courts new industry. The regional authority has issued its official demand forecast. Your investors expect a profitable plan.
Budget: £180m in Year 1, then £60m a year; unspent budget carries forward. At the end of Year 5, assets return their unused life as residual value. Score = cash + residual.
The loop: plan while the clock is paused, press ▶, and live the year month by month — weather, demand, cash, blackouts. Capacity placed while the year runs breaks ground as scaffolding and commissions at New Year. Once a year starts, its capital is committed. No takebacks.
Your build schedule, unchanged, replayed across 20 equally likely futures. The set is unbiased by construction: on average, demand equals the forecast and the weather equals its long-run mean.
◆ Swan futures are the tail draws from this known set of twenty. A true black swan sits outside any list of futures — the defence is the same: plans that keep options open.
| Future | Final profit | Shortage | Curtailed | Story | Verdict |
|---|
The five forecast errors sum to exactly zero, and the weather averaged exactly its long-run mean. Nothing was wrong with the forecast — what failed was using its average as if it were the future. A short MWh costs the £660 penalty plus the £220 sale you didn’t make; a surplus MWh earns nothing. When payoffs are asymmetric like that, the profit of the average future is not the average of profits. Your gap: £—.
Before Round 2, worth a minute:
Round 2 — Managing the Flaw of Averages. Same island. Same five years. One new instrument: the Scenario Lab. Before you commit each year, test candidate moves against all twenty futures — and choose by the distribution, not the average. Your Round 1 score of £— is the benchmark to beat.
Each candidate is only what you add this year. For later years, every candidate assumes the same brochure rule — build just enough solar to tile the official forecast, wind held flat — so you are comparing this year’s move in isolation. You’ll re-decide each later year for real. Green underlines mark the best affordable value in each column.
| +Solar | +Wind | Capex | Promise @avg | Expected | P10 | Worst | Severe futures | Distribution |
|---|
Look at the whole distribution, not the average case. The candidate that looks best at averages is usually not the one with the best downside.
Both of your plans, replayed across the same twenty unbiased futures. The winning strategy isn’t the one that looks best under the average forecast — it’s the one that protects profit when the future changes.