Tried selling my old laptop last month. eBay sold listings were all over the place. Same model, same year, prices ranging from $180 to $420. Some had keyboards with missing keys, some were "great condition, no charger." Useless for getting a real number.
So I built a script. It pulls hardware info out of WMI, looks up parts in a local pricing database, and spits out an estimate. No admin needed. No internet required for the basic run.
Here's where I got it wrong the first time.
The naive version
My first depreciation model was straight linear. Drop 20% the first year, 15% the next, and so on. Simple. Wrong.
When I tested it against actual sold listings, the script came in 25-40% high on anything older than 2 years. Linear decay overstates value because hardware doesn't lose worth on a flat schedule. It loses worth fast in year one (you opened the box, congrats) then slower after that.
What I switched to
Multiplicative decay. Year 1 retains 70% of original. Year 2 multiplies that by 0.80. Year 3 by 0.85. Floor at 15% so you never get $0 for a working machine.
function Get-DepreciationMultiplier {
param([int]$Years)
if ($Years -le 0) { return 1.0 }
$value = 0.70
if ($Years -ge 2) { $value *= 0.80 }
if ($Years -ge 3) { $value *= 0.85 }
for ($i = 4; $i -le $Years; $i++) {
$value *= 0.88
}
return [Math]::Max($value, 0.15)
}
A 3-year-old system lands at roughly 48% of its original component total. That tracked with what I was actually seeing on completed eBay sales.
Hardware detection
WMI does most of the work.
$cpu = Get-CimInstance Win32_Processor | Select-Object -First 1
$gpu = Get-CimInstance Win32_VideoController | Where-Object { $_.AdapterRAM -gt 0 }
$ram = (Get-CimInstance Win32_PhysicalMemory | Measure-Object Capacity -Sum).Sum / 1GB
Battery is its own thing on laptops. Get-WmiObject -Class BatteryStatus -Namespace root\WMI gives you charge state but not health. For health I parse the output of powercfg /batteryreport which writes an HTML file with design capacity and full charge capacity. Health percentage is the ratio.
That HTML parse is the ugliest part of the script. I used regex against the report because pulling in HtmlAgilityPack felt like overkill for two numbers. It works. If Microsoft changes the report format I'll find out the hard way.
The age problem
Figuring out how old the machine is turned out to be harder than I expected. WMI doesn't expose a clean "manufactured on" date. I tried three sources:
- BIOS release date from
Win32_BIOS.ReleaseDate. Sometimes accurate, sometimes the BIOS was updated and the date is post-manufacture. - OS install date. Useful only if the original owner never reinstalled.
- SMBIOS table parsing. Most accurate. Most annoying to do from PowerShell.
I went with BIOS date as the primary, OS install date as a fallback if the BIOS date is missing or clearly bogus (in the future, before 1995, etc). It's wrong sometimes. Most users sell within a couple years of buying so the error doesn't compound much.
Pricing lookup
The offline pricing database is a JSON file with CPU/GPU model strings mapped to current estimated values. I scraped these once from a hardware pricing site, normalized the model names, and committed it. The script does fuzzy matching because WMI returns CPU strings like "Intel(R) Core(TM) i7-10750H CPU @ 2.60GHz" and my database has "i7-10750H." Levenshtein distance under 5 counts as a match.
The eBay check is optional and slow. It hits the sold listings API, pulls the last 30 results for a query like "$cpu $gpu $ram", filters out wrecks (titles containing "broken," "parts," "no power"), and averages the middle 50% to drop outliers.
What's broken
- Desktop GPU detection is messy when there's an integrated and a discrete card. I take the one with more VRAM but that's hacky.
- No support for AMD APUs as a single unit. The script treats the CPU and GPU separately even though they're the same chip.
- The pricing database goes stale fast. I update it manually every few months. Should probably write a scraper.
- It assumes Windows. PowerShell Core would let it run on Linux but I haven't tested it.
What I'd do differently
If I were starting over I'd skip the JSON pricing database and hit the eBay API every time, with a 24-hour cache. Maintaining the offline DB is the worst part of this project and the eBay numbers are more accurate anyway. The "offline" feature sounded cool when I started. In practice I always run with -UseEbay enabled.
Code is on GitHub. PowerShell 5.1 ships with Windows so there's nothing to install.