The call
The model leans bearish on USD/BRL: its score is −29 on a scale of -100 to +100, below the -25 line, and it has held this view for 11 days. Confidence is medium. A bullish call means the first currency (USD) is expected to gain against the second (BRL).
- 76% of the weighted components agree with the call.
- Volatility is in the top fifth of its past year, so moves are less reliable.
Calls change only when the score moves clearly across the line: a bullish or bearish call stays until the score falls back inside ±10, so it does not flicker. A call is a view of direction over weeks, not a forecast of tomorrow.
Why: what makes up the score
| Ingredient | Reading | Weight | Points |
|---|---|---|---|
| Trend | Down 0.2% over three months and down 5.6% over twelve; price is below its 50-day and below its 200-day average. | 53% | −19.2 |
| Value (vs 5-year average) | Price is 4% below its five-year average; large gaps tend to close slowly, so this signal leans against the move. | 12% | +3.8 |
| Carry (interest-rate gap) | Holding USD against BRL earns about -9.88 points a year in interest (policy rates 3.88% against 13.75%). | 23% | −23.0 |
| Central-bank drift | Over six months the rate gap moved 1.25 points in favour of USD (central banks tightening or easing at different speeds). | 12% | +10.0 |
| Macro drivers | How the instrument has reacted to the dollar, equities and US rates over the past year; see the drivers table below. Shown for context; not in the score (it did not help in testing). | 0% | – |
| Score | −29 | ||
Each ingredient is scaled from -1 to +1 using only past prices and published rates; points = weight × reading × 100. The weights were set from published research before testing; an ingredient that clearly failed in the 2010-2017 test period was taken out of the score. Full method.
Price chart
Risk levels
Levels for a short idea, from recent volatility (the 14-day average true range, ATR, is 0.0526 or 1.04% of the price):
The "wrong beyond" level is where the call's reasoning would no longer hold on current volatility; it is a way to think about risk, not an order. Objectives are simple multiples of that distance. Volatility now: 18.9% a year , higher than 100% of the past year's days . Moves can gap past any level.
Interest rates behind the pair
A positive gap means holding the first currency earns more interest than the second costs. Rates come from the central bank rates page (BIS and central banks). Some currencies are managed by their central banks, so their prices do not move freely .
How it reacts to the big drivers
| Driver | Link over the past year | Driver over 3 months | Pushes the instrument |
|---|---|---|---|
| US equities | −0.42 | rising | down |
| US 2-year yield | +0.28 | rising | up |
"Link" is the correlation of 5-day moves with the driver over the past year (between -1 and +1; below 0.15 in size is ignored). The driver's 3-month direction is scaled by its own volatility. For currencies this table is context only: in testing it did not improve the call.
Seasonal pattern
In October, USD/BRL has averaged −0.18% and risen in 54% of the 28 years we have.
Context only, not in the score. For most markets a calendar pattern is weak and unstable; farm goods have the clearest seasonal logic (planting and harvest).
Track record: the model replayed on USD/BRL
From 2010 to today (17.4 years), one position, rebalanced monthly, using only the data of each day, including the interest earned or paid, after costs of 2.5 basis points per switch.
| Strategy | Return a year | Risk | Per unit of risk | |||
|---|---|---|---|---|---|---|
| Follow the model's calls | +2.5% | 11.2% | 0.22 | |||
| Trend alone | +1.7% | 13.9% | 0.12 | |||
| Always long USD | −2.5% | 15.1% | -0.16 | |||
| The model's calls, in two halves | ||||||
| 2010 to 2017 | +1.8% | 11.1% | 0.16 | |||
| 2018 to today | +3.2% | 11.4% | 0.27 | |||
On a single instrument luck dominates: a model that is right a little more than half the time still has long losing stretches. The pooled result across all pairs is on the forex page. This is a replay of rules, not a live record. The live log of every daily call started on 16 Oct 2026.
| Date | Changed to | Score | Price then |
|---|---|---|---|
| 5 Oct 2026 | Bearish (was neutral) | −32 | 4.9936 |
| 23 Sep 2026 | Neutral (was bearish) | −7 | 5.1703 |
| 2 Sep 2026 | Bearish (was neutral) | −32 | 5.0912 |
| 14 Aug 2026 | Neutral (was bearish) | −6 | 5.2231 |
| 21 Jan 2025 | Bearish (was neutral) | −26 | 6.0225 |
| 12 Aug 2024 | Neutral (was bullish) | +4 | 5.4981 |
| 12 Jun 2024 | Bullish (was neutral) | +25 | 5.4128 |
| 16 Jan 2024 | Neutral (was bearish) | −6 | 4.9263 |
| 15 Dec 2023 | Bearish (was neutral) | −25 | 4.9376 |
| 12 Dec 2023 | Neutral (was bearish) | −3 | 4.9636 |
*Price move afterwards in the direction of the call (positive = the call was right); blank for neutral calls and for recent changes without enough days yet. Shows hits and misses alike.
About this call
USD/BRL is a currency pair: the price of one USD in BRL. The call comes from one scoring model used for every instrument in the same way, described on the method page. It reads prices, interest rates and (for context) trader positioning; it does not read news or predict events, and it can be wrong for long periods.
Risk warning. Educational research, not personal advice or an invitation to trade. Forex and commodity trading, especially with leverage, carries a high risk of losing money quickly; most retail accounts that trade leveraged products lose money. Past behaviour of a model does not predict the future.