NZD to ZAR
Track record: Ensemble, 30 days ahead
Pair statistics
- 14-day change
- —
- 30-day volatility
- —
- Annualised
- 1-year volatility
- —
- Annualised
- RSI, 14 days
- —
- 30-day range
- —
- Lowest and highest daily rate
Popular pairs
| Pair | Rate | Day | 30 days |
|---|---|---|---|
| — | — | ||
| — | — | ||
| — | — | ||
| — | — | ||
| — | — | ||
| — | — | ||
| — | — | ||
| — | — | ||
| — | — |
| Pair | Rate | Day | 30 days |
|---|---|---|---|
| — | — | ||
| — | — | ||
| — | — | ||
| — | — | ||
| — | — | ||
| — | — | ||
| — | — | ||
| — | — | ||
| — | — |
How it works
Data
Rates come from Frankfurter, which blends the daily mid-market rates that central banks and other official sources publish. There is one rate per calendar day, weekends included, and Frankfurter can revise a day's rate as more sources report.
Nothing here is live. The newest figure is the latest daily rate, and the tracker keeps a copy for up to an hour. Only active ISO 4217 currencies are listed.
Indicators
Moving averages. The simple average of the last 5 or 20 daily rates.
Bollinger bands. Two standard deviations either side of the 20-day average.
RSI. The 14-day relative strength index, from simple averages of daily gains and losses. Readings above 70 or below 30 are often treated as stretched.
30-day range. The lowest and highest daily rate over the last 30 days.
Volatility. The root mean square of daily log returns, annualised at the series' own frequency of about 365 rates a year.
Forecasts
Each model is fitted to the last 365 daily rates for the pair and projected forward over the chosen horizon. The fan shows the model's 50%, 80% and 95% prediction intervals: the deeper the shade, the likelier the rate is to land there if the model's assumptions hold.
Ensemble. The average of the trend, linear regression and exponential smoothing forecasts, weighted equally.
Trend. Carries the recent average daily change forward, weighting the last 5 days over the last 20, and lets it fade over about a month.
Linear regression. Fits a straight line through the past year of daily rates and extends it.
Exponential smoothing. Holt's method: a smoothed level and trend, updated with each day's rate.
ARIMA-lite. An autoregressive model of daily changes that uses the last three changes, fitted by least squares.
Kalman filter. Tracks the rate's level and a slope that fades over time, with the noise in each estimated from the data.
Track record. Every week over the past year, the model is refitted on the 365 days before that week and asked for a forecast at the chosen horizon, which is then compared with the rate that followed. The average miss is set against the simplest forecast there is: that the rate stays where it is. The counts show how often the actual rate landed inside the 50% and 95% ranges; a well-calibrated fan catches about half and about 95% of them. Neighbouring forecasts overlap in time, so treat these figures as a guide, not a precise score.
Tested this way on NZD/ZAR, EUR/USD and USD/JPY over the year to November 2026, none of these models was clearly more accurate than assuming no change, which is the usual result for exchange rates. The models only see past rates and know nothing about rate decisions, elections or anything else that moves a currency.