IRR Explained: Getting the "Annualized Return" Right

The left-hand side of the Return Spread is the "probability-weighted IRR." IRR (internal rate of return) sounds intimidating, but it simply condenses three things—how much I pay now, how much I get back later, and how many years pass in between—into a single, comparable annualized return.

The Three Components of IRR

Put these three together and you get "if everything unfolds as assumed, roughly how much this money can compound each year."

• Entry: the current share price, corresponding to the current Forward P/E.
• Terminal Value: earnings in some future year x the reasonable multiple at that time (see "Terminal Value and Terminal Multiple").
• Effective horizon (H_eff): the number of years from today to that endpoint, to a decimal (see "Effective Horizon H_eff").

The Key: Probability-Weighted, Not Just Optimistic

A single-scenario IRR is easy to fool yourself with. The engine requires four scenarios—Bull/Base/Bear/Stress—each with its own IRR, then weights them by their probabilities of occurring to get the "probability-weighted IRR." This step is crucial: it forces you to count the downside in, rather than fixating on the optimistic scenario to spin a grand story. A stock with a beautiful Base case but a non-trivial probability of Bear/Stress will see its probability-weighted IRR pulled down substantially.

“Think in terms of expected value, not in terms of the best case.”— Probabilistic Thinking

IRR Is Bound to the Holding Period

IRR must be consistent with the effective horizon: you cannot fit a "3-year return" to a "5-year terminal value." The engine locks IRR, terminal value, and the dynamic P/E range all onto the same H_eff, avoiding time-basis conflicts—one of the most common yet most hidden errors in amateur calculations.

How It Shows Up in the Report

The report gives the IRR for each scenario and the probability-weighted IRR, along with an "IRR bridge" that lays out the terminal EPS, terminal multiple, and other assumptions behind each number, so you can check them item by item rather than accepting a black-box figure.