How Guru Grades Are Graded

Where calls come from

Calls are collected from gurus’ own sites and letters and from news articles that quote them. Each one is kept with the exact words, the date it was published and a link to the source. Only calls about the direction of the U.S. stock market are kept.

Which calls count

• One call per guru, direction, timeframe and week. The same forecast is often reported by several outlets. Within one week only a guru’s earliest bullish call and earliest bearish call for each timeframe count; the rest are listed as repeats (412 so far). A call that week with a different timeframe (three months rather than year end, say) is a different forecast and counts on its own.
• Neutral calls (“could go either way”) are listed but not counted.
• Calls too vague to grade are listed but not counted.
• Gurus quoted far more often than others are sampled: a set number of randomly chosen calls a month is collected for them, so they don’t outweigh everyone else. Their pages say how many.

When a call comes due

Each call gets its due date when it is collected, before the outcome is known. A stated timeframe sets it: “by year end” or “this year” is December 31, a named month or quarter is its last day, “six months” is 126 trading days and “next year” or “12 months” is one year. A call with no timeframe comes due three months (63 trading days) after it was made.

Right or wrong

At the due date the call is checked against the S&P 500 from its close on the day the call was published. The index move is taken as it is, not adjusted for the market’s normal rise. The words matter: a bullish call needs a gain of about 2% or more, a “correction” needs a fall of about 5%, and a “crash” or “bear market” a much larger one. A stated index target must be reached. A call with two equal parts that went different ways counts as one right and one wrong.

Claude, Anthropic’s AI model, proposes each grade with a one-line reason, using rules drawn from the editor’s own 2005–2012 grades. Grades are published as Claude proposes them, without editorial review beforehand. The editor can correct any grade.

Score

Each graded call also has a score from 0 to 100. It measures only how far the S&P 500 moved over the call’s timeframe compared with a normal market for that length of time (a gain of about 8% a year). 50 means the index did what it normally does. A bullish call scores above 50 when the index did better than normal and below 50 when it did worse: about 84 means it beat its normal gain by one typical swing, and about 16 means it fell short by one. A bearish call is the mirror image, and a neutral call scores highest when the index did what it normally does. The score ignores the call’s wording, so a call can be wrong with a high score (a target missed in a rising market). Scores do not affect percent right or the letter.

Percent right and letter grades

Percent right is right calls divided by counted graded calls. A letter needs at least 20 counted graded calls, with about two years or more between the first and the last. Every counted call since 2022 is used, not only the last two years. A grade resting on few calls is less reliable than the total for all gurus.

A

65% or more

B

55–64%

C

45–54%

D

35–44%

F

below 35%

Today 20 gurus have a letter: A 1, B 4, C 2, D 6, F 7.

The always-bullish benchmark

The S&P 500 rises more often than it falls, so a forecaster who always said “stocks will go higher” would score well. The benchmark is that forecaster, graded by the same rules: for every counted call, a pretend call “stocks will go higher” made the same day with the same timeframe. Across all counted calls it was right 62% of the time. Each guru’s page compares the guru with it on that guru’s own dates.

The figure above covers every guru’s call dates. A guru’s own dates can be harder or easier: someone who made many calls just before falls is compared with always bullish on those same dates, on the guru’s page. A bullish call that names a target, a size or a deadline must also meet it, so a mostly bullish guru can score below always bullish.

Like any bullish call, the pretend call is right only if the S&P 500 is higher at the due date, and:

• for a calendar date (year end, a named month), that is all;
• for six months or a year, the index must not fall more than 10% in the first three months;
• for three months or less, or no stated timeframe, the index must also rise at least 2% at some point, must not fall 5% in the first month without rising 2% that month, must never fall more than 10% and, if its best rise was under 5%, must never fall 4% or more.

These are the grader’s rules for a plain bullish call; checked against Claude’s own grades of the pretend calls, they give the same answer on 95% of counted calls. The S&P 500 itself was higher at the due date after 73% of counted calls; the benchmark is lower because, like a guru, it also needs a real rise without a deep early fall. Gurus’ bullish calls were 63% right and their bearish calls 22%.