Change the assumptions. Watch the gold price forecast move.

METHOD AND LIMITS

How this simulator works

A scenario-mixture Monte Carlo model, re-run in your browser on every change (2,500 paths, daily steps, a fixed seed and common random numbers so factor impacts stay stable and correctly signed). Each day, gold's move is built from: structural central-bank and ETF demand, a geopolitical drift, −0.11% per +1bp in the US 10Y real yield (the dominant channel), −1.1× the dollar's % move, no direct Brent term (oil reaches gold through the dollar and real yields, see below), risk-off jump shocks, and a central-bank bargain-buying floor that strengthens below $3,950 and $3,800. Real yields mean-revert to a scenario level, Brent mean-reverts to a scenario target, and in the fiscal-stress regime gold partly decouples from yields (beta damped to −0.05%/bp) because the sell-off is about US creditworthiness itself.

Factor review — every driver's sign was re-validated. Real yields and Fed hikes: bearish, confirmed as gold's dominant headwind. Dollar: bearish when it rises. Central-bank buying: bullish, and it is the dip floor. War escalation: two-faced — bullish where gold is the haven, bearish where the dollar is; on the last eight Brent spikes gold's direction followed the dollar's, not oil's. US fiscal stress: bullish, because gold decouples from yields when the sell-off is about US credit itself. Inflation: two-way — hedge demand against a hawkish Fed reaction, net mildly negative here. Brent crude: no direct line to gold. Over 16 months of daily data, gold has no oil beta once the dollar and real yields are accounted for. Oil works through them: +0.18% on the dollar and +2.4bp on real yields per +10% Brent, so a spike weighs on gold, and a sustained settle above about $120 adds a hawkish-Fed push (+0.6bp per $1) and a growth drag. A ceasefire's effect depends on the dollar and Fed it arrives with: the model pairs it with a hawkish Fed and a small war-premium unwind, which is why the scenario weights matter as much as the oil price. ETF flows are folded into the structural-demand drift rather than given a slider.

Starting inputs (26 Oct 2026), refreshed from live feeds: spot $4,023.68 (28.2% below the 28 Jan 2026 record of $5,602.22) · 10Y nominal 5.00% · 10Y real yield 2.66% (FRED DFII10) · Brent $115.22 · US CPI 3.7% · Next Fed meetings on Kalshi: 28 Oct (~7% hike priced), 9 Dec (~71% hike priced), 27 Jan (~32% hike priced). All Δ figures are measured against the in-browser default run, so they read exactly 0 at the defaults. Sell-side anchors for scale, not targets: Elara $4,200–4,700 rest of 2026, BMI 2026 average $4,400, Goldman $4,650 end-2026.

Why the inputs are live now. An earlier, hand-maintained build of this model carried a 10Y real yield of 2.24% and an all-time high of $5,318. Both were wrong, and the real-yield error alone moved the 3-month median by about $230, because it is the model's dominant channel. Every market input on this page now comes straight from FRED, Kalshi and Binance, and the record high is the intraday spot peak of $5,602.22. We say so rather than quietly repricing, because a forecasting tool that hides its corrections is not worth using.

What the real-yield slider actually controls. The model responds to the gap between today's real yield and where you think it ends up, not to the level itself. Today is 2.66%; the 2.35% default is the model's view of where it settles, which is a fall of 31bp and therefore a bullish assumption, made explicit rather than buried. Set the slider to today's level if you think yields hold.

The weakness to know about before trusting the long horizons. Setting the central-bank pace to zero (which removes both the demand drift and the dip floor) takes out roughly three-fifths of the 1-year median gain at today's inputs. Structural drift compounds across all 252 trading days while the real-yield channel mean-reverts away inside about 63, so the 6-month and 1-year medians rise by construction and the 1-year probability of a new all-time high is correspondingly overstated. The 1-month and 3-month views are unaffected. We are fixing it by making central-bank demand price-elastic — they buy less as price rises, which is what the 2026 data shows, with roughly 130t bought year to date against roughly 160t in the same period of 2025.

The readouts under each control are three different kinds of number and should not be read alike. The Fed ones are genuinely market-implied. The Brent one comes from the model's own simulated paths, so it is a consistency check rather than independent evidence. The rest — real yields, the dollar, CPI, central-bank pace — are plausibility gauges from normal distributions centred on today's level, with spreads we chose.

The central-bank dip floor is currently infinite, which holds the downside tail up more than real central banks would: switching it off moves the 1-year 10th percentile down by roughly $390. Scenario weights and betas are judgement calls calibrated to 2022–26 behaviour and can shift in a genuine crisis. This is a scenario tool for exploring "what if" — not financial advice, not a price guarantee, and not a substitute for an adviser who knows your circumstances.