On this page · 7 sections▾
Q3 2026 produced 2058 scored forecasts from 36 banks across 11 markets, and the collective result reads like an argument against conviction trading: a pooled direction hit rate of 40%. When the crowd gets the sign wrong more often than right, precision on magnitude stops being a virtue.
The quarter split the field sharply between one bank that worked and a long tail that didn't. There was little middle ground, and the per-market detail shows the pain was not evenly distributed.
The Q3 2026 leaderboard
UOB sat alone at the top with a 1.01% average absolute error and 66% direction hits across 155 forecasts — the rare combination of volume and accuracy. Below it, TD and TD Securities posted low error figures on tiny samples (3 and 5 forecasts) while hitting direction 0% of the time, a reminder that a small n flatters the magnitude column and punishes nothing else. Crédit Agricole and Mizuho were the only others to clear 55% direction on a meaningful book.
| Rank | Bank | Avg. error | Direction hit rate | Forecasts scored |
|---|---|---|---|---|
| 1 | UOB | 1.01% | 66% | 155 |
| 2 | TD | 1.41% | 0% | 3 |
| 3 | Crédit Agricole | 2.28% | 58% | 12 |
| 4 | TMGM | 2.49% | 20% | 15 |
| 5 | TD Securities | 2.56% | 0% | 5 |
| 6 | Mizuho | 3.43% | 55% | 51 |
| 7 | Société Générale | 3.65% | 37% | 86 |
| 8 | RBC Capital Markets | 3.67% | 43% | 30 |
| 9 | Rabobank | 3.84% | 51% | 39 |
| 10 | CIBC | 3.99% | 30% | 10 |
Ranked by average absolute error across all Q3 2026-dated forecasts; minimum three scored forecasts to qualify.
Closest call in every market
The closest single calls by market tell their own story about where the sign was easy and where it wasn't. USD/JPY, AUD/USD, USD/BRL and the oil contracts carried the higher market hit rates; EUR/USD, USD/CAD, USD/CNY, USD/CHF and NZD/USD all sat below 20%, meaning even the market's closest misses came from a field mostly facing the wrong way.
| Market | Closest call | Their target | Where it settled | Miss | Banks graded |
|---|---|---|---|---|---|
| EUR/USD | CACIB | 1.1395 | 1.1345 | 0.44% | 29 |
| USD/JPY | CIBC | 157.06 | 157.25 | 0.12% | 23 |
| GBP/USD | Rabo | 1.3269 | 1.3267 | 0.02% | 22 |
| AUD/USD | Danske | 0.694 | 0.6952 | 0.17% | 25 |
| USD/CAD | Citi | 1.42 | 1.421 | 0.07% | 26 |
| USD/CNY | ING | 6.72 | 6.7045 | 0.23% | 19 |
| USD/CHF | Citi | 0.84 | 0.8348 | 0.62% | 21 |
| USD/BRL | BofA | 5.15 | 5.1818 | 0.61% | 19 |
| NZD/USD | Citi | 0.55 | 0.564 | 2.48% | 20 |
| BRENT | GS | 112.31 | 115.91 | 3.11% | 7 |
| WTI | Westpac | 87.50 | 97.18 | 9.96% | 13 |
Call of the quarter
Rabobank called GBP/USD at 1.3269 (published 2026-07-31); it settled at 1.3267 — a 0.02% miss.
Miss of the quarter
Citi called BRENT at 62.8975 (published 2026-07-11); it settled at 115.91 — a 45.74% miss.
Who leads, who follows
Accuracy is only half the grade — the other half is whether a desk thinks for itself. For every pair of forecast revisions on the same currency and target quarter, a desk either led (a peer moved the same direction within the next four weeks), followed (it moved after a peer already had), or went contra (opposite direction within two weeks). The herding index is the share of follows — the classic career-concern pattern where being wrong with everyone beats being wrong alone. Rolling 18 months, all pairs.
| Desk | Herding index | Leads | Follows | Contras | Interactions |
|---|---|---|---|---|---|
| Scotiabank | 49% | 17 | 31 | 15 | 63 |
| UBS | 47% | 14 | 28 | 18 | 60 |
| Danske Bank | 44% | 16 | 28 | 19 | 63 |
| Citi | 44% | 28 | 33 | 14 | 75 |
| J.P. Morgan | 43% | 14 | 20 | 12 | 46 |
| Mizuho | 41% | 23 | 33 | 24 | 80 |
| Crédit Agricole | 40% | 5 | 10 | 10 | 25 |
| Morgan Stanley | 38% | 19 | 27 | 25 | 71 |
| Goldman Sachs | 37% | 43 | 40 | 25 | 108 |
| UOB | 36% | 59 | 74 | 73 | 206 |
| Rabobank | 35% | 8 | 9 | 9 | 26 |
| MUFG | 33% | 19 | 14 | 10 | 43 |
A low index is not automatically good — a desk can lead the pack in the wrong direction. Read it against the leaderboard above.
What it means
The gap between UOB's 66% and the pooled 40% is the headline, but the spread in sample sizes is the subtext. TD's 1.41% error on three forecasts and TD Securities' 2.56% on five are not comparable to Société Générale's 3.65% on 86 or UOB's book of 155. Reading the leaderboard without the n column invites exactly the wrong conclusions.
For anyone consuming bank forecasts, the dispersion is the warning. When the crowd is right on direction 40% of the time, and the best single call (Rabobank, GBP/USD, 0.02%) shares a quarter with the worst (Citi, Brent, 45.74%), the aggregate is close to noise. Pick your bank, check its sample, and distrust the consensus.
---
Scored against realized values at each forecast's target date. Live per-bank accuracy, updated daily, is on the track record and each bank's forecast page.
Frequently Asked Questions
Which bank had the best Q3 2026 forecasting record?
UOB led with a 1.01% average absolute error and 66% direction hits across 155 scored forecasts.
What was the overall direction accuracy for banks in Q3 2026?
The pooled direction hit rate was 40% across 2058 scored forecasts from 36 banks.
What were the best and worst single calls of the quarter?
Rabobank's GBP/USD call missed by just 0.02%, while Citi's Brent call missed by 45.74%.
Read next
Firms covered in this article
Bank Forecast
Uob →
Bank Forecast
Td →
Bank Forecast
Creditagricole →
Bank Forecast
Tmgm →
Bank Forecast
Tdsecurities →
Bank Forecast
Mizuho →
Bank Forecast
Societe Generale →
Bank Forecast
RBC →
Bank Forecast
Rabobank →
Bank Forecast
Cibc →
Bank Forecast
Danskebank →
Bank Forecast
Citi →
Bank Forecast
ING →
Bank Forecast
Bank of America →
Bank Forecast
Goldman Sachs →
Bank Forecast
Westpac →
Bank Forecast
Scotiabank →
Bank Forecast
UBS →
Bank Forecast
JPMorgan →
Bank Forecast
Morgan Stanley →
Bank Forecast
MUFG →
More from Forecast Scorecard
Q1 2026: RBC nails direction 88% of the time, but commodities buried everyone
RBC Capital Markets hit direction on 88% of its calls, yet the best single forecast belonged to Citi — a 0.12% miss on GBP/USD. Gold and WTI wrecked the rest.
- USD/MXN
USD/MXN Consensus Check: Spot at 18.49, Median Target 17.75 — Week of October 16, 2026
USD/MXN trades at 18.49, roughly 4.2% above the 18-firm Dec-26 consensus median of 17.75, with a 2.20-point spread separating the most and least bearish desks.
- EUR/USD
EUR/USD Consensus vs Spot: 3.1% Gap — Week of Oct 16, 2026
EUR/USD trades at 1.1239, roughly 3.1% below the 31-firm median Dec-26 target of 1.16, with a 0.155 dispersion range signalling deep disagreement on the path.
Share