Statomotive.
The proof

Profit is made at the purchase, not the sale.

We found 1897 cars that sold twice, so we could see who actually made money on a collector car. 44% lost. And once we lined them all up, it turned out the winners and losers were mostly decided on the day they bought, not the day they sold.

What we mean by fair value
Worth sixty seconds, because every chart below leans on it
  1. We start with every recorded sale of a model we can find. Price, date, mileage, and the full build sheet: year, body, gearbox, colors, options, trim.
  2. Then we work out what each of those things is actually worth, one model at a time. E46s get compared with other E46s and nothing else. Mileage and the market moving over the years are accounted for, so an option's effect is not just "newer cars happen to have it."
  3. Fair value for one specific car is what the math expects that exact build sheet to sell for, that day. Not a mileage chart, not an average of asking prices. Two cars with the same year and miles can sit $80,000 apart. The market knows. The price guides mostly do not.
  4. It updates itself every time a new sale lands. Nobody's thumb is on it.
The rest of this page is us checking whether that number actually means anything. We ranked real resales by how each car was bought against its fair value, then watched what happened.
What buyers earned, by how they bought
What each car resold for versus what its buyer paid, grouped by purchase price vs. our modeled fair value
Bought 10%+ under fair value398 cars · 74% made money
+12.8%
Bought 3–10% under398 cars · 65% made money
+6.9%
Bought near fair value3% under to 3% over · 398 cars · 52% made money
+0.9%
Bought 3–11% over398 cars · 46% made money
−1.1%
Bought 11%+ over fair value399 cars · 35% made money
−7.2%0% (broke even)
Gross resale profit on the 2086 resales of the 1897 cars we followed. No hold-time filter and no resale excluded: the 1991 resales our model can price are ranked by purchase price against fair value for that exact car and spec, then split into five equal groups of ~398. Group sizes are equal by design; the price bands they span are what vary. The other 95 are cars the model deliberately does not price (modified, incomplete data). No fees modeled.
Does holding longer help?
For each holding period: the typical win (green), the typical loss (red), and under each label, how often owners won
Under 6 months336 cars · 48% won
+7.2%
−8.1%
6-12 months337 cars · 50% won
+11.1%
−8%
1-2 years547 cars · 51% won
+13.8%
−10.2%
2-3 years382 cars · 51% won
+18.3%
−12.6%
3-5 years344 cars · 64% won
+19.9%
−12.3%
5+ years140 cars · 84% won
+40.2%
−10.4%
Same resales. The typical loss holds steady no matter how long the car was kept; the typical win grows with time, and the odds of winning barely move until year three. The 5+ year group is small and mostly bought before the 2021 repricing, so part of its gain is the era, not the wait. Your entry price is what separates 35% winners from 74%.

Here is the pattern we keep running into: the market pays up for whatever cannot be made anymore, and slowly loses interest in everything else. Capped special editions are gaining 11 points of premium a year. Factory manuals, 5.29. Rare colors, 3.31. Meanwhile base trims sit flat, common colors sit flat, and the comfort tech of 2006 keeps fading, because a twenty-year-old navigation system is not a luxury anymore. It is just old.

The two slopes
How fast each attribute's price premium is growing or fading, in points per year, across every car we track
Halo editionrising in 4 of 4
+11
Factory manualrising in 7 of 8 cars
+5.29
Rare paint, any huerising in 48 of 66
+3.31
Front-axle liftrising in 3 of 4 cars
+2.82
Common paintevery hue
+0.0
Audio and comfort tech
−0.34
Manual conversionrising in 0 of 3
−2.35
Power seatsrising in 0 of 4
−3.01
Premium drift measured on every price driver with three or more years of history, aggregated across the cars we cover. We measure these axes separately and they keep agreeing with each other, which is why we take the pattern seriously.
It is rarity, not the color itself
Premium drift of paint colors, grouped by how rare each color is on its own car
Rare colorsunder 8% of the cohort · 66 drivers
+3.31
Mid colors8 to 15% · 33 drivers
−0.51
Common colorsover 15% · 30 drivers
+0.0
The control that proves it: rare subdued colors gain nearly as much as rare bright ones. A common color of any hue gains nothing. Blue looked like the rising color only because blue happens to be rare on most cars we track.
What the manual premium really is
Today's manual premium on every car that offered a choice, shaded by what the automatic actually was
Murcielagovs e-gear single clutch
+160.5%
Ferrari F430vs F1 single clutch
+137.1%
BMW M5 V10vs SMG single clutch
+82.2%
BMW M6 V10vs SMG single clutch
+77.2%
911 Turbo (997)vs torque-converter auto
+63.7%
Gallardovs e-gear single clutch
+56.8%
BMW M3 (E46)vs SMG single clutch
+36.2%
Audi R8 V8vs R tronic single clutch
+33%
BMW M3 (E9x)vs modern dual clutch
+31.1%
BMW M4vs modern dual clutch
+10.6%
Amber bars: cars whose automatic was a jerky single-clutch or an old torque converter. Gray bars: cars with a modern dual clutch. The premium is not nostalgia and it is not rarity alone. It is the market fleeing bad automatics, on a supply of manuals that is fixed forever. That is why it also grows 5.29 points a year and cannot be re-created: a conversion recovers only part of it, and is the one gearbox class losing value.

One thing confused us for a while. The last model year of a generation sells for the most, but the first model year has appreciated best. Both turned out to be true: the last-year premium is already in the price, so there is nothing left to collect. Rare colors work the same way. Small premium today, growing every year. The appreciation lives wherever the premium has not fully arrived yet. And none of it matters as much as the price you pay against fair value, which is the first chart on this page for a reason.

We know how most of this sounds. Buy the rare special edition with the manual, buy it under market, and of course it goes up. Nobody needed a database for that. What the data adds is everything common sense gets slightly wrong. Common sense says holding longer heals a bad buy; it does not, the typical loss is the same at six months and at five years. It says the loaded car is the better car; heavily optioned cars had the worst resale odds of any build choice we measured. It says buy the final, most sorted year; that is the one year whose premium you already paid for. It says a rare color means a loud one; a rare silver quietly beats a common yellow. And it says the manual premium is about loving manuals; measured, it is mostly about avoiding the bad automatics of that era, which is why it nearly disappears on cars that got a good one. The direction was always obvious. The sizes, the exceptions, and the timing are the parts nobody's gut gets right, and they are where the money actually is.

1897
cars followed through resale
55
enthusiast cars covered
20,000+
recorded sales analyzed

Know the number before you bid.

Fair value, price drivers, and every recorded sale, for all 55 cars we cover.

Create your free account
Free. It opens the market charts on every car we cover: every sale plotted over time, price vs mileage, and price by model year.