This page presents my forecasts, for all countries of the world with major numbers of COVID-19 cases, of the future trajectories of total confirmed cases, along with predicted daily new cases (synchronized in the lower plot). Predictions are least-squares fits to three-parameter Gompertz models using current case data from Johns Hopkins University’s Center for Systems Science and Engineering.
As these models rely on available confirmed cases, predictions are sensitive to any substantial changes to testing policy, such as intensity of testing (Ireland is a case in point, see the plot and note the jump on April 10th, when thousands of samples were outsourced to Germany; or Poland’s mass testing of miners in mid-May), classification of outcomes. The models are appropriate for a single epidemic wave. To navigate to a particular country of interest, you can use the top menu, or you can just hit Ctrl-F and search for the country name within your browser. A (small) number of countries have been excluded with very poor data quality.
At the top you’ll find several bird’s-eye view plots: my projections of how many cases more or less each country will end up with at the end of this epidemic wave: both in absolute numbers and relative to population, then an estimate of the stage of the epidemic by country, and normalized historic trajectories. Individual country trajectories follow. I conclude with two-month numeric predictions for Poland, where I’m based. Following a related but slightly different approach, here’s an interactive dashboard, and another one for US states.
Note that since the majority of carriers are asymptomatic, intensive testing will elevate confirmed cases (e.g. Bahrain)!
Scores in excess of 100 percent are indicative or a resurgence of cases (what they call a “second wave”), or an accummulation over a longer term of a relatively low-level but steady daily incidence
This is the only plot that’s not a forecast. Rather, each country’s last data point is the present, and the time scale is shifted individually, so that the starting points (and courses) are comparable, both being scaled to population size (a common error that I’ve seen is normalizing the starting point to absolute numbers, which effectively captures different epidemic stages in more and less populous countries).
Note: isolated extreme daily counts and negative daily counts (artefacts in the JHU data) are not shown on the lower plots.
Date | Predicted | Daily New |
---|---|---|
2021-02-16 | 1551300 | 2388 |
2021-02-17 | 1553600 | 2318 |
2021-02-18 | 1555900 | 2250 |
2021-02-19 | 1558000 | 2183 |
2021-02-20 | 1560200 | 2119 |
2021-02-21 | 1562200 | 2056 |
2021-02-22 | 1564200 | 1995 |
2021-02-23 | 1566100 | 1936 |
2021-02-24 | 1568000 | 1879 |
2021-02-25 | 1569800 | 1823 |
2021-02-26 | 1571600 | 1769 |
2021-02-27 | 1573300 | 1716 |