Is it cheap or expensive right now?

Latest price vs our fair value. A comparison, not a recommendation.

In plain terms: platinum pays nothing and has no cash flows, so its value rests primarily on its ratio to gold (~0.40× today — the 12th percentile of the post-2016 regime, still historically cheap) and the analyst consensus (UBS $1,700–1,800 post-crash; the January LBMA survey's $2,222 was set at the record and is stale), cross-checked against a fourth-straight structural deficit that has cut above-ground stocks to ~11 weeks of cover — the thinnest ever. Fair value NOW is a zone of $1,310–$2,139 (centre ~$1,634), with spot ~$1,608 essentially at the centre after a −42% round trip from January's record — a crash triggered by the surprise Warsh Fed-chair nomination. The near-term cap is an ~18-year-high real yield, a firm dollar and demand destruction (China jewellery −42%); the swing risk is reverse Pt→Pd substitution now that platinum trades at a ~1.3× premium to palladium; the upside is the stock-depletion squeeze re-arming and a 2027 easing cycle.

Key levels

The 5th–95th and 25th–75th percentile bands from the 50,000-path simulation. Static and pre-computed — independent of the fundamental value.

What drives the odds

50,000 paths. The width is sized by the carry-anchored YZ-HAR v3 engine on platinum's own history (Gaussian tail, width 0.853 — a PROVISIONAL first fit, flagged). Over 58 resolved three-month forecasts, the price finished inside the 90% band 95% of the time, against the 90% that band aims at — and inside the 80% and 50% bands 90% and 57% of the time.; drift is pure cost-of-carry (Fed 3.63%, q = 0) — no factor drift, so the fundamental and probabilistic lenses stay independent. The forces that decide which path realizes: real US 10-year yields, the dollar, the gold price & Pt/Au ratio, the palladium discount & reverse substitution, auto-catalyst demand, Chinese jewellery & bar demand, hydrogen/industrial growth, South African supply shocks, lease-rate/physical squeezes, ETF & positioning flows, and the stale-consensus reversion. Details in the methodology.