Our track record, graded
We make one call: buy. Every day we record it for every product we track, then grade it against what the price actually did over the next 30, 60, and 90 days. A buy was right if the price never got meaningfully cheaper (more than 2% below) in the window. When we don’t say buy, we say nothing, and you decide. We haven’t found another review site that publishes its own hit rate; we think that’s exactly why you should want one that does.
At 30 days
When we said buy, the price stayed put or rose within 30 days in 3794 of 4160 graded calls (91%). Buying everything, every day, would have been right 68% of the time on the same rows. The difference, +23 pts, is the part our call actually earns.
| Window | Right | Wrong | Accuracy | Always-buy baseline | Lift | Awaiting grade |
|---|---|---|---|---|---|---|
| 30 days | 3794 | 366 | 91% | 68% | +23 pts | 1075 |
| 60 days | 2480 | 286 | 90% | 59% | +30 pts | 2469 |
| 90 days | 1692 | 201 | 89% | 66% | +23 pts | 3359 |
23,830 daily snapshots recorded since 2026-06-21, across every tracked product, buy or not. Grades post automatically the day each window closes, including the misses. A call is graded against the lowest recorded price in the window after it, from the same daily price history that powers every page here. Agents can pull this table raw at /api/v1/track-record.
When we said nothing
Silence is not a prediction, so it isn’t graded. For the record, this is what happened after it. Of the 1,771 no-call days whose 90-day window has closed, a meaningfully lower price showed up in 1,029 (58%).
| Window | No-call days graded | A lower price appeared | Share |
|---|---|---|---|
| 30 days | 11803 | 4710 | 40% |
| 60 days | 6621 | 3547 | 54% |
| 90 days | 1771 | 1029 | 58% |
Where the call is weak
An average hides the products it’s wrong about. Of the 143 products with ten or more graded buy calls at 30 days, 125 score 80% or better and 18 do not. They are listed here, weakest first.
Method, versioned
Method v2, in force since 2026-10-02. A price counts as meaningfully lower when it is more than 2% below the price at the time of the call. A BUY was right if no meaningfully lower price was recorded in the window after the call (buying then cost nothing versus waiting). Baseline = how often 'always buy' would have been right on the SAME graded rows, every tracked product, every day. Lift = our accuracy minus that. When the rule changes, grades already made keep the rule they were made under, and this list says what changed.
- v2 · 2026-10-02 — Retired the public WAIT verdict. Products are either a BUY or carry no call. Its historical grades are kept for the record (37% right at 30 days, 68% at 90) and the site no longer makes that claim. Buy grading unchanged. Baselines and per-product grades published.
- v1 · 2026-06-21 — Every verdict snapshotted daily and graded against the lowest recorded price in the 30/60/90 days after the call, 2% bar.
Retired WAIT verdict, graded under v1 as “a meaningfully lower price appeared in the window”: 4710 of 11803 at 30 days (40%); 3547 of 6621 at 60 days (54%); 1029 of 1771 at 90 days (58%). Kept for the record; the site no longer makes that claim.