Actionable insights — Update Buy-Hold-Sell List: February 2026

1. Rate every name on three independent valuation methods, then read where they disagree

1. Run the same three checks on every watchlist name, every month: (a) current forward PE against its own five-year average; (b) an Earnings Growth Model — EPS growth + dividend yield, plus or minus the move from today's multiple to a "fair exit PE" — giving an expected annual return; (c) a reverse DCF solving for the growth rate the current price already assumes.
2. Record all three outputs side by side rather than collapsing them into a single score. The value of the sheet is in the spread between them.
3. Rate Buy / Hold / Sell on the composite, and accept that a Buy will often fail one method.
4. Treat a disagreement as the research prompt: a name cheap on the multiple but failing the DCF has a growth problem; a name expensive on the multiple but passing the DCF usually has an accounting problem worth understanding.

• A "fair exit PE" that is set to the same round number (25) for most names — it standardises the model but imports one big assumption into every row.

2. Track intrinsic value and price as two separate series, and measure the gap

1. Once a year, estimate the change in what each holding is worth, independent of what the shares did.
2. Put the two percentages side by side. If value rose and price fell, compute how much cheaper the stock became — that number, not the drawdown, is the improvement in expected return.
3. Aggregate it across the portfolio to get a single sentence about the opportunity set.
4. Use it to reframe a losing year: a portfolio whose value compounded while prices fell has a better forward return than it did twelve months ago.

• An intrinsic-value estimate that is itself a function of the multiple you assume — if fair value moves with sentiment, the "gap" is circular.

3. Use the count of names passing all three methods as a market-breadth gauge

1. Each month, count how many watchlist names clear every valuation method, not just one.
2. Track that count as a time series and compare it to its own history rather than to an index level.
3. Read a record high as evidence about your own opportunity set, not as a market forecast — the claim is "there is more to buy," not "the bottom is in."
4. Let the count, rather than a macro view, set how much capital you deploy this month.

• Whether the count is rising because prices fell or because your growth assumptions rose — only the first is genuinely new information.

4. Read the top of a valuation screen as a map of the market's current fear

1. Sort the watchlist by discount to its own historical multiple and look at the top fifteen as a group, not as fifteen separate ideas.
2. Name the common factor. If one industry dominates the list, you are looking at a single narrative repriced across many tickers.
3. Decide whether you are underwriting the narrative once or fifteen times — buying the whole top of the screen is one concentrated bet wearing a diversified costume.
4. Then look for the names on the list that do not fit the theme; those are the idiosyncratic opportunities.

• A historical average computed over a window that contains a bubble — a 2021-inflated five-year mean makes everything look cheap today.

5. Decompose an "expected return" before trusting the ranking

1. For each high-ranked name, split the expected return into its parts: growth, dividend yield, and multiple change.
2. Flag rows where the dividend does most of the work — the growth assumption may be low precisely because the business is not growing.
3. Flag rows where multiple expansion does most of the work — those depend on the market changing its mind, which is not something you control.
4. Prefer the rows where growth carries the return, and treat the others as income or as re-rating trades with a different risk profile.

• A dividend yield that is high because the price collapsed — the model treats it as return, the business may treat it as unaffordable.

6. Keep the mechanical rating separate from the thesis, and expect it to whipsaw

1. Let the monthly rating move purely on price versus modelled fair value, without rewriting the business case each time.
2. When a rating flips, ask which input changed. If only the price moved, the flip is information about valuation, not about the company.
3. Publish the flip anyway. A rating that never downgrades a favourite is not a rating.
4. Keep the thesis in a separate document, so a Hold does not silently become a sell and a Buy does not silently become a conviction.

• Ratings changed without a published reason — two downgrades here carry no figures, so the reader cannot check whether price or model moved.

7. Audit the Holds — that is where the discipline is actually tested

1. Run the same three methods on what you already own and publish the output, including the rows that embarrass the portfolio.
2. Identify holdings the model marks as overvalued, and state explicitly why they are being kept rather than trimmed.
3. Reconcile the sheet against the actual book each month — a stale row is a control failure, not a rounding error.

• A Hold that is really "I don't want to pay the tax" — the model's verdict and the decision should be reconciled in words, not left implicit.