We taught The Brick Drop to grade a set four different ways — for the shelf collector, the builder, the parent, and the investor — and refreshed all 28 sets so those grades are actually trustworthy.
Until today, every set on The Brick Drop had one grade. But a set that's a great display piece can be a lousy investment — one letter can't say both. So we added four personal "for you" grades on top of the main score. Along the way we discovered the stored grades had gone stale (the scoring brain had been upgraded since they were last calculated), so we re-graded all 28 sets from fresh reviews. Some grades moved a lot. They're all live now.
One number can't tell four different people whether a set is right for them.
The Millennium Falcon is a stunning shelf centerpiece, a marathon build, a questionable gift for a 7-year-old, and a shaky flip. Same box, four honest answers. So underneath the one headline grade, every set now carries four "lenses" — each one re-weighs what matters to a specific kind of buyer:
Crucially, these aren't four separate opinions we made up. They're the same evidence — the same reviews, prices, and facts — simply weighted differently. That's what made a safe shortcut possible.
Grading a set is expensive: we mine YouTube reviews, pull live prices, score fifteen different traits. Doing that four more times per set — once per lens — would be slow and costly.
A student's grades in each subject don't change. What changes is which subjects you emphasize when you compute their overall GPA. An art school weights studio classes heavily; an engineering school weights math. Same transcript, different "for you" GPA.
The scoring engine works exactly that way: it grades each trait once, then rolls them up with a set of weights. The lens only swaps the weights at the very last step. So in principle we could compute all four "for you" grades from data we already had stored — no re-grading, no cost, and the main grade could never accidentally change. We built and tested that shortcut. Then it caught a problem.
Before writing anything, the shortcut ran a safety check: re-calculate each set's main grade from the stored data and confirm it still matches. All 28 failed the check. That sounds alarming — it was actually the safety net doing its job.
The 28 grades were calculated weeks ago. Since then, the scoring brain got two upgrades it never saw:
1. A build metric was dropped. "Instructions quality" was removed as its own score and folded elsewhere — so the build category is now weighed differently than when these sets were graded.
2. Minifigures got smarter. The engine used to score minifigs on count alone. Now it asks "how many figures for the price?" — so a set with 12 figures in a $650 box reads very differently than it used to.
Because the minifigure change alters a trait's actual value (not just its weight), the free shortcut couldn't faithfully fix it. And there was a bigger implication: the live site was serving grades from the old brain. Refreshing them wasn't a risk to avoid — it was a correction worth making.
Rather than paper over stale numbers, you chose to re-grade all 28 sets properly — pulling fresh reviews and live prices through the current engine, and writing the four "for you" grades alongside each refreshed main grade.
A quick "good enough" refresh would have produced hybrid grades — new math on old minifigure data — that nobody could stand behind. Re-mining takes longer and costs more, but the grades that came out are ones you can actually defend. And from now on, every grade automatically writes all five versions, so this never drifts apart again.
Refreshing under the current engine moved 17 of the 28 overall grades — 10 down, 7 up — and left 11 unchanged. A few swung by two or three letter grades. Here are three that tell the story:
The Lord of the Rings: Minas Tirith slipped from A− to B−. The culprit is the new minifigure math: 12 figures sounds like a lot, but in a $650 box that's a thin ratio, so the figure score came down.
Pikachu and Poké Ball jumped from C to B+. A ~2,050-piece set whose parts alone are worth roughly $543 is a strong value story once the refreshed engine re-weighs the build — though note the low confidence: it rests on thin review coverage.
Shopping Street fell from B+ all the way to D+ — the largest single move. The refreshed value and build reads simply landed much lower than the stale grade had them.
Overall grade shown as was ▸ now, then the four "for you" lenses. The last column is how confident the engine is — most sets are still an "early read," resting on limited reviews (more on that below).
| Set | Overall (was ▸ now) | Display | Builder | Parent | Investor | Confidence |
|---|---|---|---|---|---|---|
| Pikachu and Poké BallPokémon · 2,050 pcs · 0 minifigs | C▲B+ | B+ | B+ | A- | C+ | 16%early read |
| Venusaur, Charizard and BlastoisePokémon · 6,838 pcs · 0 minifigs | A-▼B+ | B+ | B+ | B+ | C | 17%early read |
| Toa Undersea AttackBionicle · 399 pcs · 6 minifigs | C▲B | B+ | B | B | C+ | 22%early read |
| Claude Monet - Bridge over a Pond of Water LiliesArt · 3,179 pcs · 0 minifigs | B+▼B- | B- | B- | B- | C+ | 21%full read |
| Police StationTown · 399 pcs · 4 minifigs | B-=B- | B- | B | B | C | 25%early read |
| Tintin Moon RocketIdeas · 1,283 pcs · 5 minifigs | C▲B- | B- | B- | B- | C | 26%full read |
| Stranger Things: The Creel HouseIcons · 2,593 pcs · 13 minifigs | B-=B- | B- | B- | B | C | 24%full read |
| The Lord of the Rings: Minas TirithIcons · 8,278 pcs · 12 minifigs | A-▼B- | B- | B- | B- | C | 24%full read |
| Porsche 911 RSR and 911 Turbo 3.0Speed Champions · 398 pcs · 3 minifigs | C+=C+ | B- | C+ | B- | C- | 22%early read |
| Douglas DC-3 PAN AM AirlinerIcons · 1,903 pcs · 4 minifigs | C+=C+ | C+ | C+ | C+ | C- | 29%full read |
| Millennium FalconStar Wars · 7,541 pcs · 8 minifigs | C+=C+ | B- | C+ | C+ | C- | 37%full read |
| FIFA World Cup Official TrophyEditions · 2,842 pcs · 1 minifig | B-▼C+ | C | B- | C+ | C- | 21%early read |
| Yoda BustStar Wars · 399 pcs · 1 minifig | C▲C+ | C+ | C+ | C+ | C | 22%early read |
| The Razor CrestStar Wars · 930 pcs · 5 minifigs | C+=C+ | C+ | C | C+ | C- | 22%early read |
| Safari Wildlife Tree HouseCreator · 397 pcs · 2 minifigs | D▲C+ | C+ | C | C+ | C- | 22%early read |
| The Mandalorian's N-1 StarfighterStar Wars · 1,809 pcs · 2 minifigs | C=C | C | C+ | C | C- | 28%full read |
| Avengers: Hulk Lab SmashMarvel Super Heroes · 399 pcs · 5 minifigs | C=C | C | C- | C | D+ | 22%early read |
| S.H.I.E.L.D. HelicarrierMarvel Super Heroes · 3,057 pcs · 6 minifigs | B+▼C- | C | C- | C- | D+ | 29%full read |
| McLaren F1 Team MCL39Technic · 1,675 pcs · 0 minifigs | C-=C- | C | C- | C | D+ | 26%full read |
| Baxter Robot RampageTeenage Mutant Ninja Turtles · 397 pcs · 5 minifigs | C▼C- | C | C- | C | D+ | 22%early read |
| Sheung Wan Western Market— · 397 pcs · 0 minifigs | D+▲C- | C- | C- | C- | D+ | 16%early read |
| X-wing Starfighter (UCS)Star Wars · 1,949 pcs · 2 minifigs | C▼C- | D+ | C- | C- | D | 21%early read |
| Shopping StreetIcons · 3,458 pcs · 7 minifigs | B+▼D+ | D+ | D+ | C- | D+ | 26%full read |
| Power TruckTechnic · 398 pcs · 0 minifigs | F▲D+ | D+ | D+ | D+ | D | 16%early read |
| Technic Control 1Dacta · 397 pcs · 0 minifigs | D=D | D | D | D+ | D | 16%early read |
| TarakavaBionicle · 397 pcs · 0 minifigs | D=D | D | D | D | D | 18%early read |
| Police Command BaseTown · 397 pcs · 4 minifigs | C▼D | D | D | D | D | 27%full read |
| X-wing Starfighter (polybag)Star Wars · 87 pcs · 0 minifigs | C-▼F | F | F | F | D | 18%early read |
This is what the personalization unlocks. Pick a lens, and the set shows a grade for you next to the overall grade — plus a plain-language reason. These are live from the database right now:
The same set, honestly different depending on who's asking. That's the whole point.
Most grades are an "early read." Even after re-mining, many of these older or niche sets have very few written reviews online, so the engine leans on hard facts (part value, piece count) more than on opinion. It flags this honestly as low confidence — but the letter still shows prominently, so read it with that in mind.
One lens has a wording mismatch. Pikachu scores A− on the parent lens, yet its reason says it's "not for the kids." The lens weighting likes it; the written reason disagrees. That's a copy/logic wrinkle to tidy — not a grading bug.
One set (Sheung Wan Western Market, a Hong Kong promo) had no price listing on BrickLink, which crashed its pricing step. We graded it without the price anchor for now and logged a fix so a missing price fails gracefully instead of stopping the grade.