How will you resolve this?

@Akzzz123 When we find out the origin of covid, I'll resolve this to the date.

If it's a lab leak, this will be the actual (or best guess) lab accident date.

If it's a natural virus that spilled over at the Huanan market, this will be the actual (or best scientific consensus) first spillover date in Wuhan.

If it's a natural virus that originated outside Wuhan, this would be the date of the first patient infected in Wuhan.

Since the Rootclaim debate discusses various theories with potential starting dates anywhere from August to December, the purpose is to see if viewers develop confidence in any of those theories.

@PeterMillerc030 And how would you decide if we have found the origin?

@Akzzz123 Maybe I could wait until Isaac King resolves his market?

Obviously it's hard to know exactly what evidence will finally settle this debate -- if it is proven to be a natural virus, I think there will still be some people that believe it's a lab leak. But I think there will eventually be enough evidence that moderates will believe it.

What evidence would settle the debate for you, either for lab leak or for natural origin?

There were no COVID cases in Wuhan in 2019 because nothing was called COVID yet.

@zcoli First infection with the virus that came to be known as SARS-CoV-2.

@PeterMillerc030 Perhaps I should reconsider my huge wager on the technicality. I don’t know if I’ve seen an estimate of dates of first human infection inclusive of the likelihoods of lineages that did not persist until being sampled. Could infer it poorly from Pekar 2022 discussion (95% CI of 2 to 23 introductions with the two observed infections separated by days to a few weeks). Lots of uncertainties and model dependence, but at least it bumps likelihood of October up a bit.

Perhaps I’m looking at it from the wrong direction, but multiple introductions spread over a long time seems less plausible for lab leak… so I don’t understand insistence of people promoting lab leak to push things back when genomes of what’s sampled point so clearly at November whether or not you buy assumptions going into conclusions of multiple introductions.

@PeterMillerc030 To clarify, which date are we talking about?

(1) Date of first transmission of SARS2 to a human that results in a productive infection? If so, how much difference is allowed for the transmitted viral genome from lineages A and B?

(2) Earliest transmission to human that leads to unbroken chain of infection for years (for A) or until today (for B or alternative progenitor).

(3) Earliest transmission to human that results in symptomatic COVID-19, whether or not it was diagnosed.

(4) Earliest transmission date to a human eventually diagnosed with COVID-19.

@zcoli I'm looking for first provable infection. So in the event of a lab leak, that'd be the date that the first person infected in the lab got sick. Proof and resolution in that case would come from documentation of a lab leak, including the date of the accident.

In the possible case that the virus is natural, that would be the date that the first person was infected with that natural virus. Proof and resolution in that case would likely come from serology data and/or interviews at the Huanan market confirming that a wild animal trader was infected earlier than any other known case at the market. While it could theoretically be possible that someone else in Wuhan was infected earlier, if there's no antibody evidence or reliable interview with that person, this can not be used as proof of an earliest covid case.

To further clarify, let's say that the date here would resolve to the date of symptom onset, not the date of infection. That would maintain consistency with existing covid cases (i.e. Wei Guixian had symptom onset on December 10, 2019, but she was infected at an unknown earlier date).

If, for some reason, we gather evidence proving the exact date that a person was infected, but not the date of symptom onset, I will add an estimated incubation period to find the proper date for market resolution.

@PeterMillerc030 A Q with a definitive resolution is something like: “On January 1, 2027, what will be the median estimated date of first infection of a human with SARS2 in a meta analysis of estimates in papers published in 2024, 2025, and 2026? Any SARS2 variant without a major difference in human-to-human transmissibility counts. Only the most recently published manuscript will be taken into account for senior authors with multiple qualifying estimates. Only manuscripts published in peer reviewed journals with 2023 impact factors above X will be considered. Manuscripts in journals with eLife-type peer review models will be considered on a case-by-case basis evaluating whether reviewer criticism was addressed that quantitatively impacts estimated dates of infection.”

I hate journal impact factors, but a reasonably low-but-not-too-low cutoff avoids needing to be subjective in ways that are worse. I assume more papers will be published with updated estimates in this 3 year period, but that’s entirely a guess.

The main point of my question was whether earlier infections prior to A, B, or something preceding A and/or B that did not lead to sustained transmission qualify. In an animal intro scenario, October might be underpriced if you think animals are housed elsewhere in a place with more animals per human. A model with a distribution of likelihoods that an infection leads to sustained transmission would have market introduction at the high end compared to other scenarios.

@zcoli Market serology would be helpful as would serology linked to high confidence onset dates with IgM and IgG data… especially multiple samples from same patients with December onset.

@zcoli In the case that Covid spilled over from an animal at HSM, there's surely an earlier covid case in a farmer or an animal trader, but it's unclear to me if that person got infected in Wuhan.

I don't know how the animal distribution worked in Wuhan, if there is some warehouse or something that the animals reached first, then I suppose there could be an earlier case there. If that case is proven, some day, I would count it. But if it's not, the earliest proven case would likely be at the market.

If we never get more data from the market (or the lab), I suppose I can just resolve this based on the current scientific consensus in the future, i.e. the mean index case date from whatever the best epidemic simulations say at the close date.

Also, 2040 is ridiculously far away, maybe I should bump this up to 2029, or so? If China hasn't shared more data in 5 years, then they probably never will.

Mostly I just wanted to see if anyone had a preference for covid's starting time, since that was discussed so heavily in the debate. Maybe I should have posted a poll, instead, but then there's no disincentive for people to just guess a month.

In the 3rd debate, I tried to get Saar and Yuri to create some model for case growth rates, under their lab leak scenario, and they did come up with some specific numbers. I can dig up the spreadsheet if you're interested.