What is wrong with a backtest?

Nothing, as research — but rules tuned on past data almost always look better than they perform live. Only a record logged on the day of each pick shows real-world results.

Why compare picks with the S&P 500?

Because buying an index fund is the free alternative. A pick that rose less than the market over the same dates did worse than doing nothing clever.

How many graded picks make a record meaningful?

There is no magic number, but ten is far too few: a 6-for-10 record fits a true hit rate anywhere from roughly 30% to over 80%. Dozens start to mean something.

How to read an AI stock picks track record (and spot a bad one)

Every stock-picking service shows a number. Most of those numbers are built to impress rather than inform. Six questions separate a record you can trust from a sales pitch.

1. Is it live, or a backtest?

A backtest runs today's rules over past data. It is easy to tune until it looks brilliant, because the rules were written knowing what happened. A live record logs each pick on the day it is made, before anyone knows the outcome. Only a live record tells you how the picks do in the real world.

2. Is every pick in it?

Watch for survivorship: picks that went badly quietly disappear, or the record starts after a rough patch. A trustworthy record lists losers exactly like winners and does not reset.

3. Compared with what?

"Up 12%" means nothing if the whole market rose 15% over the same weeks. A good record measures every pick against a benchmark — usually the S&P 500 (SPY) — over identical dates, and reports the average edge, not just the raw return.

4. Over which window?

A fixed, stated grading window — say one week and one month after each pick — stops anyone from choosing the flattering moment to measure. Be suspicious of records that report each pick's best price.

5. How many picks?

Small samples are noisy. A 6-for-10 record is consistent with a true hit rate anywhere from roughly 30% to over 80%. Dozens of graded picks start to mean something; hundreds mean a lot more. A young record is a reason to keep watching, not a verdict.

6. Does the win rate come with the size of the wins?

A high win rate can hide small wins and large losses. Look for the average return next to the win rate, and for both next to the benchmark.

How WhatToTrade grades its own picks

Each morning's published picks are recorded with their price at the close of the pick date.
One week and one month later the close is read again, and the same is done for SPY over the identical window.
A pick "wins" if it is up; it "beats the S&P 500" if its return is higher than SPY's. Averages are simple means, every graded pick counted once.
Losses are listed like wins, nothing is back-filled, and returns are price-only.

The result is published live on the track record page. It is young, so read it with question 5 in mind — and hold us to the same six questions as everyone else.

How to read an AI stock-picks track record