How Delphi recommendations work

A good company can be a bad stock if the price already assumes everything goes right. So every recommendation answers one question: at today's price, is this stock likely to beat the S&P 500 over the next 12 months, counting dividends?

The labels

Conviction is high when the gap is 20 points or more, medium at 15, and low otherwise. For most readers, Sell means “don't buy it now, and consider trimming if you own it,” not “bet against it.”

1. The same arithmetic for every company

A year from now, investors will be pricing the following year's results. So each recommendation starts with Wall Street's consensus estimate for that year: earnings per share for most companies, sales for companies without meaningful profits, and average-across-the-cycle earnings for boom-and-bust businesses like memory chips.

That estimate is multiplied by a valuation. The base case starts from what investors pay for the stock today and moves 25% of the way toward its own five-year middle, but never more than 15% from today. The bull and bear cases use the high and low ends of analyst estimates and of the stock's valuation history.

Fast growth rarely keeps its price. When a company's earnings are forecast to grow more than 20% in a year, every scenario's valuation is trimmed, by up to 30% for the fastest growers.

2. Judgment goes in two places only

Research sets the odds of the bull, base, and bear cases, and explains in plain English what has to happen in each. The odds come from evidence: whether earnings estimates have been rising or falling, the company's record of beating or missing them, and how its own guidance compares with Wall Street's. Every recommendation shows that evidence.

3. Compared with the market, the same way

The S&P 500 goes through the identical arithmetic, using its own consensus earnings and valuation history. Comparing each stock with the index cancels the biases they share. Wall Street's estimates tend to run optimistic, for example, and that lifts the index's expected return just as much as any single stock's.

Dividends count

Owners earn the dividend as well as any price change, so every comparison uses total return: the expected price change plus the stock's current dividend yield, against the S&P 500's expected price change plus its own yield. A stock paying 6% starts about five points ahead of the index.

4. Recommendations move with the price

The inputs stay fixed until we update them, but prices change every day, and the recommendation is recalculated after every close. If a Buy-rated stock rallies toward its target, the expected return shrinks and it can fall to Hold on its own.

5. When there is no recommendation

• The company reported earnings after the inputs were gathered. New results can change everything.
• The inputs are more than 120 days old.
• The model produced an extreme result (more than 50 points ahead of or 40 points behind the market), which usually means an input needs re-checking.
• The stock has traded publicly for less than two years, so there is no valuation history to anchor on.
• Fewer than three analysts publish estimates for the company, so there is no real consensus to build on.

6. Every recommendation is on the record

Each change is logged with the date, the stock price, and the S&P 500, and the log is never edited. The track record shows whether Buys beat the market and Sells trailed it.

What this is not

Delphi recommendations are generated by a model from public information and published automatically. They are general, not personalized investment advice, and they don't know your goals, timeline, taxes, or what else you own. Estimates can be wrong, and past results don't predict future ones.