Home › Gold forecasts › Saturday 17 October 2026
Gold closed the week at $4,166, down $28 (0.66%) after a week of noise: below $4,100 to its lowest since early August, back up again, and still capped under $4,200 for a fourteenth consecutive session. The weekly call we published last Saturday closed inside its second box — so the order was wrong — but its confirming number was met: we said “Brent at $108 or above”, and Brent closed Friday at $109.86 after a 3.7% jump in a single session. So the verdict is right for the reason, wrong on the order — better than we ourselves forecast in Friday’s map. Six calls were graded this week, and the result carries one piece of good news and one piece of bad. Below: correction number five, and a new weekly call written with all of it applied.
Friday’s close $4,166 · On the week −$28 (−0.66%) · 21K in Suez EGP 6,250 ·
The weekly call (published Saturday 10 October): 40% the rebound continues ($4,200–$4,340) · 35% hot inflation ($4,060–$4,200) · 25% a Hormuz deal ($4,340–$4,460). The week closed at $4,166.39, inside the second. So the order was wrong.
But the second box’s confirming number was: “the yield at 5.28% or above, or Brent at $108 or above.” The yield closed at 5.085% ❌ — but Brent closed at $109.86 ✅, after jumping 3.72% in Friday’s session alone. The condition was met, and the verdict is right for the reason.
And that scenario was named “hot inflation” — and the inflation came in hot on every measure: headline CPI 3.7% on the year against 3.6% expected, core 2.6% against 2.5%, and producer prices +1.3% on the month and 6.3% on the year, the highest since December 2022. The story we wrote is the story that happened. Our only error was ranking it second rather than first.
And here is what has to be said plainly: in Friday’s map, published before the close, we wrote that if gold closed below $4,200 the second box’s condition would need “the yield to jump 0.18 or Brent to rise 2.8%… both remote, so the likely result is right by chance.” Brent rose 3.72%. Our own map ruled out what actually happened, and the result came out better than we predicted for ourselves. We are not going to claim we called it — we called the opposite, and the market was generous.
Friday’s daily call: we published 40% ($4,200–$4,250) · 35% ($4,150–$4,200) · 25% ($4,100–$4,150). The close at $4,166.39 was inside the second, so the order was wrong. Its confirming number was “the 10-year between 5.12% and 5.26%, or a quiet session”. The yield closed at 5.085%, 0.035 below the band, and the quiet rule broke (Brent +3.72%, the dollar −0.70%). So: right by chance.
⚖️ A note on the reading: our source gives two different figures for Friday’s yield — its numeric tables say 5.085%, its written summary says it “edged up to 5.12%”. If the second is correct, Friday’s call was right for the reason, not by chance. We took the harsher reading against ourselves and recorded it as right by chance. Our rule: when there are two numbers, we grade ourselves with the one that does not favour us.
Record: 43 published · 41 graded · 38 correct · 93%. Friday’s map → · How grading works →
| Call | Close | Range | Order | Confirming number | Verdict |
|---|---|---|---|---|---|
| Monday 12 | 4,171.88 | ✅ first | ✅ | ❌ Brent 106.65, dollar 102.59 | By chance |
| Tuesday 13 | 4,161.51 | ✅ second | ❌ | ✅ the new quiet rule | For the reason |
| Wednesday 14 | 4,105.57 | ✅ second | ❌ | ❌ yield touched 5.33, closed 5.20 | By chance |
| Thursday 15 | 4,171.06 | ✅ first | ✅ | ❌ core PPI 0.7%, retail 0.8% | By chance |
| Friday 16 | 4,166.39 | ✅ second | ❌ | ❌ yield 5.085, 0.035 below | By chance |
| Weekly (10 Oct) | 4,166.39 | ✅ second | ❌ | ✅ Brent 109.86 above 108 | For the reason |
Read that table carefully, because it holds two pieces of news.
🟢 The good news: six out of six — every close this week landed inside a box published before it happened. Not one session escaped the map. That is the best range result since the series began.
🔴 The bad news: the order was right in two out of six, and the confirming number was met in two out of six. We know where gold will stop; we do not know which outcome is likeliest, or why.
And that is not a detail. Range alone is easy: widen the boxes and one of them has to win. The value is in naming the most likely outcome, naming the reason, and being graded on both. That is what we were weak at this week, and we are writing it in bold rather than advertising the six out of six.
Correction four (Thursday 15): on a day with a scheduled economic release, the confirming number is the release value plus the price closing inside the range. That is all. The driver’s level at the close cannot be a third condition, because the driver can reverse hours later for an entirely unrelated reason. That is what happened Wednesday: core inflation came in at 0.3% as we said, the yield went to 5.33% as we said, and then equities sold off and pulled it back to 5.20%, breaking our condition.
Correction five (today) — the threshold has to sit away from the price: we lost the condition twice this week by a hair: Monday (Brent 106.65 against a 106 line, the dollar 102.59 against 102.5) and Friday (the yield 5.085 against a 5.12 line). Looking at both, the cause was the same: we had put the threshold right next to the current price. That is not a test of a reading. That is a coin toss.
From today: a confirming threshold has to sit a reasonable distance from the price at the time of writing — at least 1% for oil and the dollar, and 0.06 for the yield. Any closer and it is not a condition, it is luck. You will find the rule applied in the new weekly call below, with the distance printed next to every number.
The earlier rules still stand: 1 thresholds are “at or above” · 2 yield bands are at least 0.12 wide · 3 a quiet scenario is confirmed by the size of the move, not by a level.
| Now | Reference | |
|---|---|---|
| Spot gold, weekly close | $4,166 | −$28 on the week |
| 24K, per gram, we sell | EGP 7,142 | unchanged |
| 22K, per gram, we sell | EGP 6,547 | unchanged |
| 21K, per gram, we sell | EGP 6,250 | unchanged |
| 18K, per gram, we sell | EGP 5,357 | unchanged |
➡️ 21K steady at EGP 6,250 (buy 6,200), from our counter reset today. If it is reset again, the live number is at today’s prices. 📍 The ounce fell from $4,187 to $4,166 while the gram held, so the implied dollar rate rose from 53.06 to 53.32 — the gram is sitting slightly above its value again. The same movement we described all week: the local market rises by lift and falls by stairs, and the gap settles over two or three days rather than within the hour. 📅 The week in numbers: the ounce from $4,194 to $4,166 (−0.66%) · the 10-year from 5.22% to 5.085% (−14 basis points) · the dollar −0.4%, its first weekly fall after three weekly gains · and Brent +5.1% to $109.86. Oil was what moved this week, not gold.
A week empty of data and empty of the Fed: the blackout began yesterday ahead of the 27–28 October meeting, and there is no inflation and no jobs report. Two drivers remain: oil and geopolitics, and risk appetite in equities.
⚠️ And the first thing we did was change the order. Last week we ranked the upside first in three calls out of five, and the market stayed locked under $4,200 for fourteen consecutive sessions. So this week we rank the cap continuing first, not the break. That is the week’s lesson applied, not described.
🇪🇬 21K in Suez at the end of next week: EGP 5,970 – 6,495, most likely EGP 6,120 – 6,300, assuming the implied dollar rate holds at 53.32.
✏️ Every threshold in the three scenarios above has its distance from the current price printed beside it. That is correction five applied. If you find one sitting close to the price, that is our failure and you should tell us.
Six out of six inside the range. A pretty number, the kind that goes in an advertisement. We are not going to do that.
The reason is simple: the range is the easy part. Put three boxes across $400 and the odds that the close lands in one of them are very high — that is not skill, it is arithmetic. The hard part is saying which box is likeliest, and the harder part is saying why and being graded on the reason.
This week we won the easy part and lost the hard one: order two out of six, reason two out of six.
So what are we doing about it? Three things, all written down and all in force from today:
1️⃣ Stop favouring the upside. We ranked the break higher first in three of five calls while the market has been capped under $4,200 since 29 September. From this week, the order is built on what the market is actually doing, not on what logic says it ought to do.
2️⃣ Thresholds sit away from the price (correction five). We lost two conditions by a hair because the line was pressed against the price.
3️⃣ The second chain is now in the model. Gold does not always run on the rate chain. On frightened days it runs on the haven chain, and then hot inflation can lift it rather than sink it. Next week’s call contains a whole scenario built on that chain.
And what this means for you: when someone shows you a forecast, do not ask whether it was right or wrong. Ask: how wide was the range, what did it say was likeliest, what number did it put down to confirm it, and how far was that number from the price? Four questions that separate analysis from talk in about a minute.
| If the ounce is | 21K per gram works out near |
|---|---|
| $3,980 (floor of the new weekly call) | EGP 5,970 |
| $4,080 (the dividing line) | EGP 6,120 |
| $4,100 | EGP 6,150 |
| $4,166 (the weekly close) | EGP 6,250 |
| $4,200 (the line) | EGP 6,300 |
| $4,250 | EGP 6,375 |
| $4,330 (ceiling of the second scenario) | EGP 6,495 |
| $4,460 | EGP 6,690 |
📍 Raw metal at today’s local conversion factor (1.5001), before making charge, hallmarking and tax.
| $4,330 | Ceiling of the new call’s second scenario |
| $4,230 | The neckline |
| $4,200 | The line gold has been capped under since 29 September |
| $4,166 | The weekly close |
| $4,100 | Support |
| $4,080 | The dividing line in the new call |
| $4,066 | The low of the past two weeks |
| $3,980 | Floor of the new call |
Next week there is no inflation, no jobs report and no Fed. The only drivers are Brent and equities. Brent above $108 is inflationary pressure that supports gold as a haven. An equity bounce removes gold’s support. And the big decision lands on 28 October.
The Fed blackout began Saturday — no Fed official speaks all week, which by
itself makes the market quieter and less reactive.
Monday 19: the NAHB housing index (31 expected after 32).
Tuesday 20: September building permits and housing starts · pending home sales.
Wednesday 21: mortgage applications · US crude inventories (last week
−8.56 million barrels, the biggest draw since February — now a number that matters to gold through
oil).
Thursday 22: jobless claims (around 199K expected).
Friday 23: flash PMIs · the final University of Michigan survey with its
five-year inflation expectations (3.4% expected after 3.5%) — the one number all week that could
move the yield.
Friday midnight: the weekly close on which the new weekly call is graded.
The week after: the rate meeting on 27–28 October.
Friday’s close and the weekly close for gold ($4,166.39), Treasury yields (5.085%, with Thursday’s close at 5.07%), Brent ($109.86 after $105.92), the dollar (101.78–101.84 after 102.55), the September CPI, PPI and retail sales releases, hike odds, the Warsh, Hammack, Goolsbee and Daly remarks, the detail on disrupted Hormuz shipping and the sinking of the GFS Pride, strait flow volumes, US crude inventories and next week’s calendar from Trading Economics; the detail on the yield reversing after Wednesday’s equity selloff from Investing.com and BabyPips; local prices from our own published price list, reset on 17 October.
⚖️ A note on precision: the source sometimes shows two different values for the same figure on the same page (Brent 109.86 or 109.69 · the dollar 101.78 or 101.84 · the yield 5.085 or 5.12). In each case we have used the figure its numeric tables give, and in grading we chose the reading that does not favour us.
This report is the English edition. The Arabic edition is primary and was published first, at the same date’s Arabic page. Every figure here appears there.