Every hobby site will tell you what's hot. None of them will put a date on when to get out. Our Sell-Window board commits to one for every tracked drop — "sell by September 1, holding 60 days past that costs ~26%" — with the decay curve drawn and the confidence stated. Here's the model, its assumptions, and its current weaknesses, in full.
A hyped item's resale price is a collector floor plus a hype premium. The premium ramps into a peak shortly after release, then decays exponentially as attention moves to the next drop. We model the premium at 40% of peak value — a documented assumption, not a discovered truth — and fit the decay with one number per category: the half-life. From that curve, "sell by" falls out mechanically: peak date plus a few days of grace to actually get the sale off, and "cost of holding" is the value the curve sheds over the next 60 days.
The half-lives come from the hobby's well-worn patterns, and they differ by nearly an order of magnitude: sneakers peak ~10 days after drop with a 21-day half-life (SNKRS attention turns over in weeks); cards run 14 and 35 (the hobby-box rip window, then the fade); sealed TCG runs 21 and 60 (booster boxes decay slowly because sealed supply keeps exiting circulation); collectibles sit at 14 and 40; and LEGO is the outlier at 30 and 90 — slow-burn, retirement-driven, and the one category where our own model warns you that the standard sell-fast logic often inverts.
Those category clocks are only the starting point. When an item has its own observed price trajectory — sold-comp history with at least 4 distinct observations spanning 14+ days — we fit the half-life to the item and the confidence tier rises: MEDIUM at that threshold, HIGH at 8 observations across 28+ days with a clean negative fit. Thin data keeps the prior and a wide ±14-day band. Right now the board holds 100 calls: 99 low-confidence, 1 medium, 0 high. We could hide that. Instead it's printed on every page, because a dated call with a stated wide band is still more useful than "this is hot" — and infinitely more gradeable.
The date isn't a prophecy; it's a default against the two mistakes that cost resellers the most. Holding through the decay — the listing that sits "until the price comes back" while the premium half-lives away underneath it. And selling into the ramp — panic-listing day two when the modeled peak is day fourteen. A dated window with a cost-of-holding number turns both into a checkable decision: you can disagree with our curve, but now you're disagreeing with a specific number instead of a vibe, and one of us will be verifiably wrong. That's the entire editorial position of this site: publish the number, grade the number.
The live board, sorted by urgency: dropforecast.com/sellwindow. Model parameters and their documentation ship on every item page.
The Journal is DropForecast's editorial layer: desk-bylined analysis written on top of the site's own tracked data and models, edited and published by the operator. Desk names are house pen names, not individual staff. Every number cites our own boards, and every call referenced here gets graded in public on the scoreboard.
