Forty-five days before the midterms — Iran, a base in Poland, Canada's turn to Europe, and whether Trump backs down again

1. The week in four decisions

Between Wednesday 16 and Friday 18 September 2026 the White House produced, in this order: a claim that the war with Iran is "hopefully toward the end" and that Tehran has been in direct contact (Al Jazeera); a warning, one day later, that "I have a big decision coming up. Do I want to go in and annihilate them, or do I not?" (The Jerusalem Post); an announcement of "major progress" toward a permanent US Army base in Poland, with the location to be named "very soon" (RFE/RL); a report, the same day, that the Pentagon is weighing the withdrawal of 25,000 to 40,000 troops from Europe (NBC News); the word "laughable" for the European Commission's offer to make Canada the EU's first associate member, with a threat of "very serious tariffs" on Europe if the move is judged hostile (NPR); and a security agreement with Denmark over Greenland that leaves the island Danish (CNN).

Behind the week sit three numbers that the White House does not control. On 16 September the Federal Reserve, chaired by Trump's own nominee Kevin Warsh, raised its policy rate by a quarter point to 3.75–4.00%, the first increase since July 2023, by a vote of 12–0, against the president's public demand for cuts; sixteen of eighteen officials expect another increase this year (CNBC). On 11 September the Bureau of Labor Statistics reported consumer prices up 3.4% over the year, with gasoline up 3.9% in a single month and responsible for more than a third of the monthly increase (BLS). And on 18 September Brent crude closed at $103.21 a barrel (Trading Economics), against about $84 on 10 August (Al Jazeera). The war, the price of fuel and the cost of money are now the same story, and it is the story of the election.

2. The midterms, 45 days out

The arithmetic has not changed since our analysis of 24 August: Republicans hold the Senate 53–47 and the House by a handful of seats; Democrats need a net gain of four in the Senate and only a few in the House. What has changed is the direction of every signal. The Reuters/Ipsos poll published on 14 September put the president's approval at 35%, up two points from a record low, and gave Democrats a seven-point lead on the generic congressional ballot, the widest of the second term (Reuters/Ipsos via U.S. News). The Silver Bulletin average stood at D+7.5 on 15 September, D+8.2 after adjustment for likely voters (Silver Bulletin). The same poll found 63% of Americans, including four in ten Republicans, disapproving of the president's convention promise of a $5,000 "dividend" if his party keeps Congress.

The prediction markets, which aggregate money rather than opinions, moved with the polls. On 16 September CNBC reported that Kalshi traders had flipped the Senate to Democrats for the first time since April, at 52–55% (CNBC); Polymarket priced a Democratic House near 90%. Nate Silver's model gave Democrats 58.6% for the Senate on 14 September, driven by better polling in Texas, Ohio and Michigan. The rating houses are more cautious, and in one case moved the other way: on 15 September the Cook Political Report shifted New Hampshire from Lean Democrat to Toss-up, in the Republicans' favour, because former senator John E. Sununu is running closer than expected against Chris Pappas (Cook Political Report). Texas and Iowa remain toss-ups after the 20 August moves; Emerson's Texas poll of 12–14 September has Talarico and Paxton within one point of each other.

One thing that was supposed to happen before the election will not. Congress passed a continuing resolution weeks early, 90–6 in the Senate and 370–48 in the House, and the president signed it on 2 September, funding the government through 11 December (The Hill). There will be no shutdown on 1 October. The next cliff falls five weeks after the vote, in a lame-duck session that will also have to find the roughly $200 billion the Pentagon says the Iran war still needs.

3. Iran: the war Washington wants to end and cannot

The war began on 28 February with the US–Israeli strikes that killed Supreme Leader Ali Khamenei; his son Mojtaba was named successor on 8 March. Since then the president has issued deadlines and withdrawn them with a regularity that markets now trade on: the "8 PM Tuesday" ultimatum of 7 April, withdrawn about an hour before expiry in exchange for a two-week ceasefire; the ceasefire extended indefinitely on 21 April; the Islamabad Memorandum signed on 17 June and declared "over" on 8 July after Iranian attacks on shipping; strikes halted again on 27 July; a "last chance" on 4 August with no follow-up; and a strike on the underground "Pickaxe Mountain" site threatened on 14 July, 22 July and 5 September and, as of this writing, not carried out. Strikes did resume on 1 September, then slowed: by 19 September there had been no Iranian missile launch at a Gulf host state for ten days, and the last US strike reported by an open-source tracker, on 14 September, hit two IRGC boats. The lull is not a truce: on 17 September US officials said Iran had shot down at least two MQ-1 drones in the preceding days (CBS News).

The Strait of Hormuz is the war's real front. The United States is running a naval blockade of Iranian ports, and the president says "not one ship" has reached Iran since it began. On 1 September CENTCOM said it had escorted about 40 vessels carrying some 18 million barrels through the strait in a single day, a wartime record (CNN); ship-tracking data from Kpler counted between five and eleven transits a day over the same days (Al Jazeera). Both sides can close the strait; neither can open it alone. That is why Brent trades above $100 while the fighting is paused.

Washington's problem is not military. The war's support at home fell to 31% in late August, from 37% in March, with only 69% of Republicans still in favour and 83% of Americans expecting it to drag on (Reuters/Ipsos via Al Jazeera). On 15 September the House voted for the third time to end US hostilities, 220–204, with seven Republicans joining every Democrat (U.S. News); the Senate passed its own version 50–48 in June. No authorisation for the use of force has been voted. Tehran's position is simpler than Washington's: it says it will talk only if the United States honours the June text, and it has a new supreme leader who is, by every account, harder than his father.

4. Poland: a base announced in the week a third of the army may leave Europe

The Polish announcement is real but not yet a base. What the president announced on 17 September, the 87th anniversary of the Soviet invasion of Poland, is progress toward one: a US delegation inspected sites the previous week; Poznań and Powidz, in the west, are the two names most cited; Poland has said it will spend around $4 billion on infrastructure, including schools for soldiers' children; the base would host several thousand troops with their families — figures around 5,000 have circulated since June — on top of the roughly 10,000 who rotate through Poland today (Stars and Stripes). President Karol Nawrocki, whom Trump credited by name, said Warsaw "will do its utmost to provide the best possible location". The legal frame is the 2020 defence cooperation agreement; the working groups date from consultations in Warsaw on 10 September.

The same day, NBC News reported that the Pentagon is weighing the withdrawal of roughly a third of US forces in Europe, about 25,000 troops, possibly as many as 40,000, with Germany, Italy and Spain the most likely to lose them; the head of European Command is to submit an assessment, and Defense Secretary Hegseth is due to receive a final recommendation by 6 November, three days after the election. The 2026 defence authorisation act forbids funding a European presence below 76,000 for more than 45 days without certifications to Congress and a 60-day wait, so a cut of that size cannot be quiet or quick. Read together, the two announcements describe a posture, not a contradiction: fewer Americans in Europe, concentrated on the eastern flank, in countries that pay. Romania has already lived the first half of that sentence. Since the October 2025 decision not to replace the 101st Airborne's rotational brigade, roughly 1,000 US troops remain in the country according to press estimates, most of them at Mihail Kogălniceanu.

5. Canada: the ally that stopped waiting

The trade war with Canada is the one file on which the president has followed through. On 20 July he invoked Section 338 of the Tariff Act of 1930, never used before, to impose 50% duties on about $20 billion of Canadian cars, alcohol and dairy; they took effect on 19 August. Talks collapsed on 24 August, and he announced that the 50% rate will extend to all Canadian vehicles and parts from 1 January 2027. Ottawa answered on 8 September with counter-tariffs of 15% to 50% on C$27.6 billion of US goods, matched dollar for dollar (CNBC). At the mandatory review of the USMCA on 1 July the United States declined to confirm the agreement's extension, which triggers annual reviews from now on; Canada and Mexico both wanted a sixteen-year renewal.

Canada's response has been to build alternatives rather than to wait. Mark Carney's government reached the NATO 2% target in March, committed to 5% by 2035, committed to raise its brigade in Latvia to 2,600 troops and keep it there until 2031, signed a strategic partnership with China in January and launched trade talks with India in March, struck a deal in July for a pipeline to move a million barrels a day to the Pacific coast, and on 16 September received from Ursula von der Leyen the proposal to become the EU's first "associate member", a status that does not yet exist in the treaties (CNBC). Carney's line to the European Parliament was that Canada and Europe "are not fair-weather allies" and that economic integration "is now being weaponized". The domestic numbers explain his freedom to say so: Abacus Data measured 59% approval of his government in early September, the economy grew 3.3% annualised in the second quarter, its fastest pace in more than three years (CBC), and 81% of Canadians describe the relationship with the United States as negative (The Hub). Canada is the case that shows what happens when the target of a threat is small enough to hit: the threat is executed, and the target leaves.

6. The pattern, measured

"TACO" — Trump Always Chickens Out — was coined by the Financial Times in May 2025 as a description of tariff threats that were withdrawn when markets fell. The 2026 record is more interesting than the slogan. Bloomberg Economics counted 49 tariff threats and reviews from November 2024 to January 2026 and found that roughly one in four was carried out (Bloomberg). Our own count of the year's decisions, below, sorts them by who was on the other side.

Date	Threat or deadline	Counterparty	Outcome
3 Jan	"You got to surrender" (Maduro)	Venezuela	Followed through: strike, Maduro captured
20 Jan	10% tariff on eight NATO allies over Greenland	Denmark, Europe	Withdrawn at Davos; 18 Sep security deal, island stays Danish
20 Feb	Iran: deal in "10 to 15 days" or consequences	Iran	Followed through: strikes on 28 Feb
20 Feb	Supreme Court voids IEEPA tariffs, 6–3	Courts	Pivot the same day to a 10% Section 122 surcharge; ruled unlawful in May, stayed on appeal
7 Apr	"8 PM Tuesday": reopen Hormuz or "a whole civilization will die"	Iran	Withdrawn ~1 hour before expiry; two-week ceasefire
15 Apr	Fire Powell if he stays past his term; fire Lisa Cook	Federal Reserve	Powell stayed on the Board; Supreme Court blocked the Cook firing 5–4 (29 Jun); Warsh hiked on 16 Sep
14 Jul, 22 Jul, 5 Sep	Strike on "Pickaxe Mountain" "very soon"	Iran	Not carried out as of 19 Sep
20 Jul – 24 Aug	Section 338 50% tariffs; 50% on all autos from 1 Jan 2027	Canada	Followed through on the first (19 Aug); the second is pending
17 Sep	Additional 7.5% "excess capacity" tariff on China	China	Delayed until after Xi's state visit of 24 Sep; truce runs to 10 Nov
17 Sep	"Very serious tariffs" on Europe if the Canada deal is hostile	EU	Open

Two regularities stand out. First, the president follows through when the cost to him is not yet visible or falls on someone else: Venezuela in January, Iran in February before any market had priced a war, Canada's first round of tariffs in August. He backs down once the counterparty has shown it can raise the price of oil, the yield on Treasuries or the price of an American car: Iran over Hormuz since April, Europe over a trade deal that moves the S&P 500, China over rare earths — and, on our numbers, probably Canada's cars in January. Second, the one institution that has held him to nothing is the one he does not negotiate with. The Fed answered a year of pressure with a rate increase, and the Supreme Court answered the attempt to fire a governor with a 5–4 ruling. "Chickening out" is the wrong metaphor. This is a president who prices his threats, and the price is set by the other side.

7. The causal chain: why a president who backs down abroad hits harder on allies

The cause is a president who cannot afford the price of his own escalation before an election. Every marginal day of war adds to gasoline prices, which add to the inflation print, which keeps the Fed hiking, which raises mortgage rates in the suburbs that decide the House. The 3.9% monthly jump in gasoline in August is the whole mechanism in one number.

The mechanism is substitution. A president who needs a visible win but cannot buy one from Iran, the Fed or the courts buys it where the bill is smaller: a base in Poland that Poland pays for, a Greenland deal that changes no border, a tariff on Canada that Canadian voters, not American ones, will punish. The announcements of 17–18 September are not scattered; they are the cheapest available wins, taken in a week when the expensive one was not for sale.

The effect is a two-speed alliance. Allies that are exposed and pay — Poland at 4.8% of GDP, the Baltic states — get more of a shrinking American presence; allies that are exposed and negotiate — Canada, the EU — get tariffs and turn elsewhere. The same logic produces a paper ceasefire with Iran, because a ceasefire costs the president nothing and buys him a lower oil price before 3 November, whereas a strike campaign costs him the very thing he is short of.

The counter-hypothesis deserves weight. The escalation path is not empty: the president's "annihilate them or not" line was said in public, Israel has said the war has "no time limit", the supreme leader is a hardliner, and a president who is going to lose the House anyway might prefer to lose it as a wartime president than as a defeated one. The February strikes are the proof that the deadline can be real. We keep this path at roughly a quarter of the distribution, not a rounding error.

8. The simulation: Monte Carlo and the olivLaw panel

We ran two instruments on the same fact sheet. The first is a Monte Carlo model with 50,000 trials over four Iran end-states for 31 December — a ceasefire or agreement announced, a continued stalemate, a major new strike campaign, and strikes followed by a ceasefire — with the price of oil, the president's approval on election eve, the ten-year yield and the two chambers of Congress drawn conditionally on each. Six further files are modelled as separate events with explicit priors: the Canadian auto tariff, the European troop cut, the Polish base site, a further Fed increase, the China truce and a December shutdown. Every prior is a dated newsroom assumption, listed in the methodology, and the model can be rerun with different ones.

The second is an olivLaw deliberation with 14 virtual persons, each anchored on a real vantage point — the White House NSC, the Speaker's office, a Senate Democrat, an Ohio America First voter, a Pennsylvania independent, a Wall Street rates desk, a former Fed economist, the Prime Minister's Office in Ottawa, the Polish defence ministry, Iran's national security council, a Gulf sovereign fund, Beijing's trade negotiators, a foreign-policy analyst and a Pentagon strategist — arguing over three rounds from the same verified facts. The panel answered two questions: whether a formal US–Iran ceasefire or agreement is announced by 31 December, and whether the 50% Canadian auto tariff takes effect at 50% on 1 January 2027. Its aggregate probability on each is pooled with the model's prior at a weight of 40%. The results are in the table below; the panel's own numbers and the individual arguments are reported in the methodology, including where the panel disagreed with the model.

Quantity, 31 December 2026 unless stated	Model prior	Panel	Pooled
US–Iran ceasefire or agreement announced	49%	42%	46%
Major new US strike campaign on Iranian territory	28%	—	29%
50% tariff on Canadian autos in force at 50% on 1 Jan 2027	30%	40%	34%
Formal announcement of a ≥25,000 troop cut in Europe	35%	—	34%
Site of the Polish base named	65%	—	65%
Another Fed increase (28 Oct or 9 Dec)	60%	—	60%
Brent above $110 on the last trading day	19%	—	20%
Trump approval below 40% on election eve	85%	—	85%
Democrats win the House / the Senate	87% / 54%	—	87% / 55%

The model's median for Brent on 31 December is $96, with an 80% band of $78 to $119: the oil market is the variable through which every Iran path is priced. The model also gives 71% odds that at least one of the three "hard" follow-throughs — a major strike campaign, the Canadian auto tariff at 50%, a European cut of 25,000 or more — happens by year-end. A president who "always chickens out" would score near zero on that line.

9. Scenarios to 31 December, with probabilities that add up to 100

Scenario	Probability	What it looks like
(a) The paper ceasefire	36%	A halt announced before or just after 3 November, Gulf-brokered, presented as Iran "wanting a deal"; Hormuz partially reopens; Brent falls toward $85; the Fed still hikes once
(b) The long stalemate	36%	Blockade, escorts and sporadic fire through the election; no text signed; Brent near $100; the $200 billion war bill and the 11 December cliff collide in the lame duck
(c) The strike campaign	18%	Pickaxe Mountain and a wider campaign; Iranian launches at Gulf hosts resume; Brent above $115; the war-powers majority in the House grows; no deal by year-end
(d) Strike, then deal	10%	A short campaign framed as the final blow, followed within weeks by the ceasefire of scenario (a)

Scenarios (a) and (b) are tied, and that is the finding. A paper ceasefire is the only outcome the president can announce alone, at no cost, before an election in which the price of gasoline is his main liability; a stalemate is the outcome Tehran can produce without agreeing to anything. The panel pulled the ceasefire three points below our prior, mostly on one argument: the June memorandum lasted three weeks, and Tehran's condition — honour it — is one Washington cannot accept before 3 November. The two escalation scenarios together are a little over a quarter of the distribution, which is the honest weight of a threat made in public by a man who executed the previous one.

10. What we got wrong on 24 August

In our midterms analysis of 24 August we gave the Federal Reserve "likely (55–80%)" to hold rates at its September meeting, with a hike as the residual risk at 26–31%. The Fed hiked. The prediction resolves against us and will be scored as such at its deadline. The error was one of anchoring: we read the July minutes as a committee split between holding and hiking and gave the incumbent outcome the benefit of the doubt, when the August inflation print and a new chair with a credibility problem to solve pointed the other way. The rest of that forecast has aged better than we feared: the House was "likely" Democratic and is now priced near 90%; the Senate was "roughly even" and is now 52–59% Democratic depending on the instrument; we said a Democratic sweep had become at least as likely as a divided Congress, and the markets now agree.

11. What would invalidate this analysis

First: a confirmed Iranian launch at a Gulf host state, or a confirmed coalition strike on Iranian territory, before the Xi visit on 24 September. That would end the lull that scenarios (a) and (b) rest on, and would move the escalation paths above a third.

Second: a reading of Brent below $90 for a week. It would mean the strait is functionally open, which removes the president's main reason to sign anything, and would move weight from (a) to (b).

Third: a public exemption for Canadian autos before 3 November. It would show that the follow-through on Canada is itself a bargaining position, and would cut the tariff prediction below 20%.

Limits of the analysis

Limits of the factual base: casualty figures for the war differ by source and date (US deaths reported between 18 and 23), and we do not cite a single number. The state of the strait is described differently by different trackers on the same day; we use CENTCOM's escort figure and Kpler's vessel counts side by side, because they disagree. The Polish base figures — $4 billion, 5,000 troops — come from Stars and Stripes and earlier Polish statements, not from a signed document. The panel's probabilities come from language models playing roles; where an extracted number contradicted the persona's own text, we say so in the methodology rather than average it away.

12. Falsifiable predictions

Ten predictions on the calibrated ICD-203 scale, each with an explicit deadline and a public source of verification. They will be scored at resolution with the Brier score, through a separate record.

#	Prediction	Horizon	Probability	How to check	Source
F1	The United States and Iran announce a formal ceasefire or agreement (a signed text or a White House-announced halt to hostilities lasting at least 30 days)	31 Dec 2026	even chance (46%)	White House statement or signed text, reported by two wire agencies	White House; Reuters; AP
F2	The United States conducts a major new strike campaign on Iranian territory (a named operation or strikes on at least three sites in one week)	31 Dec 2026	unlikely (29%)	CENTCOM statement	US Central Command
F3	Democrats win a majority in the House of Representatives	30 Nov 2026	very likely (87%)	At least 218 seats called	AP; Clerk of the House
F4	Democrats win control of the Senate (at least 50 seats plus the vice-president, or 51)	15 Jan 2027	more likely than not (55%)	Seats called, including any runoff	AP; Senate
F5	The Silver Bulletin approval average shows the president below 40% on 2 November 2026	3 Nov 2026	very likely (85%)	Published average on that date	Silver Bulletin
F6	The 50% tariff on Canadian vehicles and parts is in force at 50% on 1 January 2027, without delay, reduction or replacement by a deal	2 Jan 2027	less likely than not (34%)	Proclamation in force and CBP guidance	Federal Register; US Customs and Border Protection
F7	The Department of Defense formally announces a reduction of at least 25,000 US troops in Europe	31 Dec 2026	less likely than not (34%)	Official announcement or notification to Congress	Department of Defense; Armed Services Committees
F8	The site of the permanent US Army base in Poland is officially named	31 Dec 2026	more likely than not (65%)	Joint US–Polish statement naming the locality	White House; Polish Ministry of National Defence
F9	The FOMC raises the federal funds target range again at its 27–28 October or 8–9 December meeting	9 Dec 2026	more likely than not (60%)	FOMC statement	Federal Reserve
F10	The US–China tariff truce is extended or renewed beyond 10 November 2026	15 Nov 2026	likely (75%)	White House fact sheet or joint statement	White House; Ministry of Commerce of China

Internal consistency can be checked: F1 and F2 are not exclusive (scenario (d) satisfies both), so they sum above the scenarios; F3 and F4 are the model's chamber probabilities, which sit between the markets and the Silver model; F6 and F7 are the two follow-throughs the president has announced but not executed, and both are priced below one half for the same reason the Bloomberg count found one in four.

13. What it means for Europe and Romania

Three consequences follow for this side of the Atlantic. One, the presence is being repriced, not withdrawn. A permanent base in Poland alongside a cut of 25,000 or more elsewhere means the American footprint in Europe will be smaller and further east, and that the countries which keep it will be the ones that pay for it. Romania, at about 1,000 US troops and a Mihail Kogălniceanu expansion financed largely by itself, is already inside that logic; the Polish precedent of "$4 billion for 5,000 soldiers" is now the going rate. Two, the oil path is the fiscal path. Every Iran scenario above translates into a Brent price, and for a country financing itself at the highest yields in the Union, a year-end near $100 rather than $85 is the difference between a rating review that passes and one that does not. Three, Canada is the template for the EU. The president's threat of "very serious tariffs" on Europe over an associate membership for Canada is, on the 2026 record, the kind of threat he does not execute against a counterparty that can move the S&P 500; the Canadian auto tariff is the kind he does. European negotiators should assume the difference is deliberate.

14. Methodology and sources

Public sources used: the Iran statements of 16–17 September — Al Jazeera and The Jerusalem Post; the Polish base — RFE/RL and Stars and Stripes; the European troop review — NBC News; Canada — CNBC, NPR and CBC; the Fed — CNBC; prices — BLS; the continuing resolution — The Hill; polls and markets — Reuters/Ipsos, Silver Bulletin, CNBC on Kalshi, Cook Political Report; the war-powers vote — U.S. News; war support — Reuters/Ipsos via Al Jazeera; Hormuz — CNN and Al Jazeera; Greenland — CNN; the threat scorecard — Bloomberg.

Anchor data, verified on 18–19 September 2026: Fed target range 3.75–4.00% (16 Sep, 12–0); CPI +3.4% y/y, core +2.4%, gasoline +3.9% m/m (August, released 11 Sep); Brent $103.21 (18 Sep); House 220–204 (15 Sep); continuing resolution signed 2 Sep, funding to 11 Dec; Reuters/Ipsos approval 35% and generic ballot D+7 (14 Sep); Silver Bulletin D+7.5 / D+8.2 (15 Sep); Kalshi Senate D 52–55% (14–16 Sep); Canadian counter-tariffs C$27.6 billion (8 Sep); Section 338 tariffs in force since 19 Aug; Cook New Hampshire Toss-up (15 Sep). The number of days is counted from 28 February to 19 September, and from 19 September to 3 November.

Model priors (19 September): Iran end-states — ceasefire or agreement 38%, stalemate 34%, major strike campaign 17%, strikes then ceasefire 11%; Canadian auto tariff at 50% on 1 January 30%; European cut of at least 25,000 announced 35%; Polish site named 65%; further Fed increase 60%; China truce extended 75%; December shutdown 22%. Conditional distributions for Brent, approval and the ten-year yield are truncated normals whose means and standard deviations are listed in the model file. The panel's answers are pooled with the priors at a 40% weight. The panel had 14 virtual persons and three rounds, and argued from the same verified facts. On the Iran question its aggregate was 41.6% (median 42%), and the spread narrowed from 20 points in the first round to 7 in the third — a convergence we read partly as herding rather than new evidence. On the Canadian tariff the aggregate was 39.6% (median 35%, range 25–60%). Every extracted probability was checked against the persona's own stated number, and all 28 matched. Two defects are declared rather than hidden: several personas on the tariff question described the August Canadian tariffs as Section 232 measures, when they were imposed under Section 338, and we discount the arguments that rest on that confusion; and the fact sheet the panel received dated the 40-tanker escort to 17 September instead of 1 September, which changes no argument's direction. The persona texts were written by a language model; they are arguments to weigh, not the views of the institutions they imitate.

Method. The 2026 record was compiled by dated event, with the counterparty named, and the sorting by "who can raise the price" is the newsroom's reading, offered as a hypothesis the next quarter will test. The probabilities are the newsroom's own, expressed on the ICD-203 scale and checked for consistency between scenarios and predictions. Where the panel's aggregate diverged from the model, both numbers are printed; where an extracted probability contradicted a persona's own argument, the persona is excluded from the aggregate and the exclusion is reported.

Disclaimer: This material is for information and analysis. It is not investment advice, a political recommendation or a legal assessment, and it expresses no electoral preference. The probabilities are explicit estimates, formulated so that events can contradict them. Poll averages, prediction-market prices and race ratings are revised weekly; the figures quoted reflect the public information available on 19 September 2026.