Past Updates

Here you will find a comprehensive collection of our past updates.

• March 24: We compiled a list of all of our COVID-19-related Tweets here. Over the past year, we have posted quite a bit of original research on Twitter, so feel feel to take a look and give some of them a read.
• March 8: We made our last daily update on March 7 (containing infections estimates through February 21, 2021). Read Youyang Gu’s One Year Later blog post for a detailed explanation and for a list of alternate resources. Follow @youyanggu on Twitter for continued COVID-19 insights. Thank you for your support over the past year. See our downloadable data on GitHub.
• February 26: After almost a year since covid19-projections.com was launched, we will be making the last daily update on Sunday, March 7. See Youyang’s blog post here.
• February 24: See our latest thoughts on why we expect a return to normal by summer.
• February 22: See our latest analysis on the impact of in-person university instruction on community transmission.
• February 18: See our thoughts on the UK B.1.1.7 variant here.
• February 17: Is political leaning a better predictor of the severity of the fall 2020 wave than existing immunity and levels of intervention? See our latest findings here.
• February 11: See our latest thoughts on why the recent attention on herd immunity may be overblown, and why we should shift our focus away from this topic. Our modeling suggests that it is increasingly unlikely that we will reach theoretical herd immunity in 2021.
• January 28: We are releasing raw and adjusted CDC vaccination time series data for download on GitHub. It contains every CDC vaccination update since they started releasing data on Dec 20, 2020.
• January 25: In December, we reported on why the CDC overestimates true infections in the US. They have since adjusted their estimates, which now closely matches covid19-projections.com.
• January 20: Our new county-level summary page is now searchable, allowing for fast lookup of specific counties. Originally created on Jan 5, we present the best and worst-impacted counties in a sortable, downloadable table format.
• January 14: We broke down vaccine doses administered into 1st and 2nd doses. You can find this on our Path to Herd Immunity page, or on all of our plots.
• January 12: We added vaccination plots to the infections page for every US state (and nationally) based on the latest CDC data. For example, see the vaccination progress for the US. Scroll down or click here to find your state. See a map of the vaccination progress for all 50 states here.
• January 7: See our new thread for an analysis of existing COVID-19 population immunity and its correlation with the severity of this current wave.
• January 5: See our new county-level summary page, where we present the best and worst-impacted counties in a sortable table format.
• January 5: We made a major update to our Path to Herd Immunity page, incorporating the latest vaccination data from the CDC and preliminary information on the new COVID-19 strain.
• December 29: See why our total infected estimates differ from the CDC.
• December 24: We added the latest vaccination data from the CDC to our Path to Herd Immunity page.
• December 16: See our analysis of the effectiveness of the Kansas mask mandate.
• December 14: I’ve self-funded this project over the past 9 months. Donate to help me keep covid19-projections.com running.
• December 9: See our latest writeup, Path to Herd Immunity: 2021 Outlook of COVID-19 in the US, which models how the COVID-19 vaccine can help the US achieve herd immunity.
• December 3: We added plots of reported tests and hospitalizations to the bottom of every page. Maps are also available on the Maps page.
• December 1: New updates this week: maps over time, plots of reported cases and deaths on every page, and downloadable daily infection estimates for every US county.
• November 25: We have released a writeup of our methodology: Estimating True Infections Revisited: A Simple Nowcasting Model to Estimate Prevalent Cases in the US.
• November 23: We released detailed county-by-county estimates for every US county. You can view them below. You can view maps of these estimates on the new Maps page.
• November 18: We are re-launching this website with nowcasting of infections estimates in the US (what has happened/is happening). You can download all of our estimates here.
• November 2: Our final October 5 projections estimated 231,000 (221-245k) US deaths by November 1, 2020. The true November 1 death toll was 230,995 (as reported by Johns Hopkins University). See how our final forecasts tracked the true results here. See our updated historical performance here.
• October 5: We have released our final update. Read Youyang Gu’s blog post to read more about this decision. See our Weekly Update on Twitter. Thank you for your support over the past six months.
• September 28: We will no longer be extending our projections past November 1. Our last forecast update will be on Monday, October 5. Read Youyang Gu’s blog post. Thank you for your support over the past six months.
• August 25: In our update this week, we’ve lowered our projection for additional deaths in the US by ~10% due to the steady decrease in cases and hospitalizations nationally. See our Twitter thread. We are now also displaying Rt estimates for the future, as well as daily deaths per 1M (on hover).
• August 13: Youyang Gu is giving a virtual talk about the model today (Thursday, August 13) at 2pm ET/11am PT as part of the Marketplace Algorithms and Design seminar. You can rewatch the recording on YouTube.
• August 10: See our new findings regarding the role of immunity, behavior, and interventions in the spread of COVID-19. We use Louisiana as a case study.
• August 5: We released a report, Estimating True Infections: A Simple Heuristic to Measure Implied Infection Fatality Rate that explains our findings regarding the relationship between true infections, reported cases, test positivity rate, and infection fatality rate for COVID-19. You can view a summary on our Twitter.
• July 31: We added lower and upper bounds to our reproduction number (Rt) estimates. You can now download these values for all of our projections on GitHub.
• July 23: We made a major update that tries to better account for the recent increase in cases and deaths. Read our update notes on Twitter.
• July 13: View our updated historical performance (commentary on Twitter).
• July 8: We have extended our projections through November 1, 2020. As we predicted in our June 17 update, deaths decreased until the 4th of July weekend and are now gradually increasing. We are now incorporating a potential for a fall wave as a result of school reopenings and the beginning of influenza season, leading to a wider confidence interval.
• July 1: We updated our infections estimate to closer match the observed data. We now estimate there to be around 200k new infections per day in the US, 4-5x higher than the number of reported cases. Because we use only deaths in our model, we believe this estimate may still be an underestimate of the true prevalence.
• June 24: We have open-sourced our SEIR simulator. This is the underlying SEIR model without the machine learning layer to learn the parameters. If your system supports Python, you can generate your own simulations in under 5 minutes. No prior Python experience is needed.
• June 17: Our latest model update features a tighter confidence interval and improved modeling of reopenings. We incorporated a slower reopening, a greater lag between cases and deaths, and a lower proportion of severe cases. As a result, we now expect deaths to continue to decrease until July 4th weekend.
• June 16: We have open-sourced our code to evaluate COVID-19 models. The goal of this project is to evaluate various models’ historical point forecasts in a transparent, rigorous, and non-biased manner.
• June 11: We have extended our projections to October 1, 2020.
• June 10: We launched the C19Pro Score for both cases and deaths to show where weekly cases and deaths are changing the most. They are shown in the first two maps on our new Maps page.
• Jun 5: We made a map of our estimate of the current R_t for every US state and 70+ countries. Scroll down below to see it.
• Jun 1: We have lowered our US projections after an analysis of recent studies showing promising treatment results, possible seasonality effects, high mask-wearing compliance, and slow reopenings.
• May 29: We ran a simulation on what would happen if just 20% of infected individuals reduced their own transmission by 25% (such as by self-quarantining immediately after showing symptoms). US deaths would be 30% lower by May 29, and up to 50% lower by September. See the results here.
• May 27: We added 7 new countries (Australia, Belarus, Bolivia, Cuba, Honduras, Kuwait, UAE), 2 Canadian provinces (Alberta and British Columbia), and 20 US counties. We now have projections for 71 countries (>99% of all global COVID-19 deaths), 4 Canadian provinces (99% of Canadian COVID-19 deaths), 56 US states and territories, and 34 US counties (including the top 30 most populous counties).
• May 26: We have extended our projections to September 1, 2020.
• May 17: We added a tests-per-day target for each region based on Harvard Global Health Institute’s study that recommends 10 contacts per infected individual. You can see this estimate above the “Newly Infected” graph on each projection page (e.g. we estimate 1.3 million tests/day are needed for the US). Download the raw data here.
• May 16: We added a plot of our estimate of the effective reproduction number (R_t) over time to all of our projections. If R_t is greater than 1, it means that the number of cases is growing. The ultimate goal is to keep R_t under 1. BBC provides a simple explanation for understanding the importance of the R value.
• May 14: See how the US would have fared if everyone began social distancing one week earlier or one week later.
• May 13: If you add -noreopen to the end of any projections page URL, you can see our projections with the assumption that there are no reopenings. E.g. covid19-projections.com/us –> covid19-projections.com/us-noreopen. Some caveats here.
• May 12: We added 23 additional countries. The countries our projections cover encompass 6.3 billion people and account for over 99% of all global COVID-19 deaths. View our global projections here.
• May 1: We updated each US state’s reopening timeline according to The New York Times. To date, we are the only model referenced by the CDC that factors in individual state re-openings.
• April 25: We now added likelihoods of death milestones in the US and R value estimates.
• April 15: We now assume social distancing will be relaxed starting on June 1, leading to higher projections than prior iterations. Learn more about our assumptions here.
• April 5: covid19-projections.com launched.