Commodity & market intelligence. Every call locked before the move, machine-scored after — hits and misses published. 14 metals + Bitcoin. Not financial advice.
Coffee is 23% off its high. We still won't call the direction.
Monday's 80% band: 30,959 – 33,915. Mid 32,437.
MA20 leans up. The 10-day trend leans down. RSI leans nowhere. Verdict: no edge, 2%.
https://t.co/CkEZ90WYjZ
Signals that cancel are still an answer.
#ForgeEdge #Coffee
Monday's fences, posted Sunday.
SILVER (XAG) 66.17 – 72.08, mid 69.13. Friday closed 69.02.
GOLD (XAU) 4,494 – 4,722, mid 4,608. Friday closed 4,604.
Both are 80% bands, published before the open and scored against the close.
There is no direction call inside either one — at this conviction the fence stands on its own width, and we don't claim a direction we can't score.
https://t.co/CkEZ90WYjZ
All prices USD. Not financial advice.
#ForgeEdge #Silver #Gold #Forecasting #Calibration
89% right. Skill score -0.083.
Both numbers are off our own ledger. Below zero means no better than predicting the average — a day usually lands in its band without our help.
Band coverage 83.1% vs 80% claimed. The fences work.
#ForgeEdge#Forecasting
Two numbers off the ForgeEdge public ledger, both of them ours, and only one of them flatters us.
The first is a hit rate of 89% across 1,935 scored forecasts — 1,583 hit, 297 partial, 55 miss, Brier score 0.167 against a coin-flip 0.25. That number looks excellent. It is also close to meaningless on its own.
The second is a skill score of -0.093. Below zero means the calls are no better than always predicting the average.
Here is why both are true at once:
BASE RATE — the market lands inside its band 80.9% of the time by itself. Difficulty 0.1547. That is the market's doing, not ours.
RESOLUTION — 0.0025. The calls barely move away from that base rate, so knowing them leaves a reader roughly where they started.
CALIBRATION — under-confident by about 10.2 points. Well-behaved, which is not the same thing as useful.
Most platforms would publish the 89% and stop. The skill score is the one that decides whether any of this is worth reading, so it goes up as well.
2,214 forecasts logged, every one timestamped at creation, wins and losses both on the record.
https://t.co/CkEZ90WYjZ
Not financial advice.
#ForgeEdge #Forecasting #BrierScore #Calibration #Markets #TransparencyMatters
Our platform is 89% right and has no edge.
Both numbers are ours.
Skill score -0.093.
The market lands in its band 80.9% of the time by itself, so being right a lot proves nothing.
We publish the number that says we're not good yet.
#ForgeEdge#Forecasting#BrierScore
A quick look at why silver, gold and copper all closed lower on 18 August 2026 — and why the usual explanation doesn't fit.
Geopolitics got worse, not better, on Tuesday. Under the textbook, that lifts gold. It didn't.
The reason sits in the bond market: the 30-year US Treasury reached 5.33%, a level last seen in 2007, and real yields climbed toward 18-year highs.
Metals pay no coupon. Against a risk-free 5.33%, that's an expensive place to sit.
Silver -2.75% to 63.87 USD
Gold -1.19% to 4,359 USD
Copper -1.42% to 6.33 USD/lb
Worth noting what didn't cause it: Chinese physical demand is still historically strong, and SHFE warehouse stock is running below its long-run average.
Heavy physical buying and a falling price can happen at the same time. Tuesday was one of those days.
ForgeEdge publishes probability bands before each session and scores them against the close.
90% weighted hit rate across 1,311 scored records, direction skill -0.161.
We publish the weak number as well as the strong one.
Not financial advice.
#ForgeEdge #Silver #Gold #Copper #Yields #Macro
Metals didn't break. Bonds broke them.
Silver, gold and copper all fell on a day the Middle East got worse — the opposite of the textbook.
The 30-year US Treasury hit 5.33%, its highest since 2007.
When the safe thing pays that much, the metal that pays nothing has nowhere to stand.
Silver -2.75% to 63.87 USD
Gold -1.19% to 4,359 USD
Copper -1.42% to 6.33 USD/lb
We publish our bands before the session and score them after the close. 90% weighted hit rate, direction skill -0.161. Both numbers, always.
#ForgeEdge #Silver #BondMarket #Macro #Trading
The Global News Radar gathers public headlines from Google News and the BBC, keyword-scores them every ~30 minutes, and shows the total as a single risk number. Right now: 69/100 across 60 headlines.
Under it sits a line most tools would never print: machine triage, not editorial judgement. We are not claiming the machine is clever. We are claiming it says what it is. Every headline links back to its own source.
Find it at https://t.co/ESlKzjl2xi → Business → Global News Radar
Not financial advice. Capital at risk.
#Shorts #MarketIntelligence #NewsRisk
69/100.
60 headlines, rescored every 30 minutes.
The tab itself says: machine triage, not editorial judgement.
We'd rather tell you what it isn't than dress it up.
Business → Global News Radar
#NewsRisk#Trading#MarketNews
silver goes into monday 1% sure. gold 3%. copper 2%. no direction call on any of them.
monday's 80% fences publish anyway — silver 62.10-67.36 — and lock before the open. record: 90% hit, skill score - 0.161. the RED number stays up.
would you publish a shrug?
#ForgeEdge
wednesday, before the open: silver fence locked at 62.18-67.78. then cpi. the day ran 64.72 to 66.81, closed 65.12 — inside all session. silver's record: 28 locked, 14 hit, 5 miss. the MISSES sit beside the hits. is a forecast worth anything if it can be edited? #ForgeEdge
cpi at 13:30. conviction: silver 2%, gold 3%, copper 3%. no edge anywhere. today's 80% bands went out before the print — silver 62.18-67.78, gold 4,274-4,486, copper 6.32-6.53. scored tonight, MISSES included. is a flat call still worth publishing?
#ForgeEdge#CpiDay
CPI lands wednesday. our bands were drawn before it.
silver ±4.05% → ±6.88% by friday
gold ±2.26% → ±3.85%
copper ±1.55% → ±2.64%
same engine, same hour, three different amounts of DOUBT.
only day one is scored.
narrow before a print, or widen?
#ForgeEdge