Subjective resolution, judged by me, don't bet if that's not your cup of tea. I will not bet on this market.
For the purpose of this market, a loss is defined as at least one judge concluding that the natural origin is correct. Note that this is different than the definition in the linked market and also includes a split decision. I chose this definition because I think it includes all the scenarios which would challenge Sa'ar's extreme confidence. This market will N/A if both judges vote for lab leak or stay undecided.
I'll have to judge by the actual response, but I expect Saar to congradulate Peter or at least write a nice rather than defensive post concluding the debate, and I also expect him to update the website and the tone he uses to reflect the fact that he's lost the debate.
It's fine if he doesn't do everything I expect, I'm mostly just looking for sincere acceptance, not resorting to trying to pretend the debate didn't exist or blaming the judges of not underatanding Bayesian inference or whatever.
1,000
3.00Resolving yes. The main reason is the doubling down that is apparent in the updated COVID origins estimate. One of the things that always rubbed me wrong with Rootclaim is their grandiose claims about "never being wrong" and "exceeding human reasoning". To any reasonable observer, the fact that Saar has lost the debate means that there is at least 0.2% chance that he is wrong. However, he made a new estimate that suggests that he truly is that sure he is right in the debate. I think even a minimum level of humility would mean giving 1% to the chance that they lost fair and square. Alternatively, some introspection on the veracity of Rootclaim's process could be sufficient, but the only introspection was about explaining themselves well enough. They did not update their beliefs at all. Being able to update your beliefs is a crucial part of being a good Bayesian.
I still think the tone was fine, even the delay in the results I could have forgiven, but this is crucial.
Entirely fair to consider the format wasn't ideal in hindsight given various incorrect claims weren't picked up. https://twitter.com/AntGDuarte/status/1760452313510969526?t=vncc_OgJHhYe6Y3CeIgkaA&s=19
https://twitter.com/Dissenting2020/status/1760968618240536657?t=PyRjhuo6yOL00zv2EaNmqQ&s=19
https://twitter.com/ban_epp_gofroc/status/1759478663412859379?t=OvTjutVvgYcNkkxrj6aOQw&s=19
@MikePa67d That first thread is basically pandemic fanfic.
Antonio has invented a theory where there was an earlier (Lineage A) covid outbreak which was successfully contained (how exactly? the only thing that's contained covid successfully is full city lockdowns) but then still spreads quietly at a low level and then shows up again at the Huanan market.
Okay, first, did it only show up at the Huanan market? Because, if so, you're stuck with the same crazy coincidence that the lab virus only shows up at the single most likely location in Wuhan for a natural origin. You need to put those odds in your probabilistic calculation, and then if you do, lab leak loses.
Or, was it everywhere in Wuhan, but only seen at the market? Well then, you've again defeated your theory. It can't be everywhere by December 1st (violates the exponential growth model I presented, and other reasonable models on case numbers). It doesn't match the Wuhan serology data or the excess pneumonia deaths, or any of the other data pointing to the north side of the river. It also violates the theory that Antonio just created where Lineage A was successfully contained. But then Lineage B is not contained somehow and it's everywhere?
Then Antonio says, "top brass in Wuhan & Beijing know the outbreak is research-related"
Except that... the Wuhan government did nothing to control the outbreak for weeks. They allowed lunar new years celebrations weeks after the market outbreak was public. They allowed a 40,000 family banquet designed to set a record. Then they freaked out and reversed course a few days later, instituted a mask mandate, shut down the entire city of Wuhan. Then Beijing shut down most of China soon after. Then shut down the farms and killed 10's of millions of animals (for show, of course).
Meanwhile, they censored the early whistleblowers who did say covid was spreading at the market. Which is the exact opposite of what you'd expect, if they were trying to promote a false market narrative to cover up a lab leak.
I don't doubt that lab leak Twitter can come up with confusing stories like this by stringing together random facts, nor that they will continue to do so for years to come. But I also think that any one of these stories will unravel and fall apart given any serious examination, like what was done at the Rootclaim debate.
The theories will fall apart just the same, whether it's a verbal or written debate. Most of them would be a lot easier to debunk in writing, honestly -- that's what I originally asked for when we were negotiating the contest.
I won't be doing a second written debate, but that's for a variety of other reasons. More than anything, there are trust issues after Rootclaim spent 2 months trying to overturn the outcome of the last debate. I don't see how I could possibly bet against them again after that happened. But I also just don't want to spend another 6-12 months reading and debunking every obscure new version of the lab leak theory.
@PeterMillerc030 A big strength of the debate was requiring Saar, outsourced to Yuri, to actually present a model at the end of the debate that was at least qualitatively consistent with data showing that B came before A which came before lineages derived from A. You can imagine an infinite capacity for coverups in China and it doesn’t work because you can ignore all data from China and this is still what you find in SARS2 genomes sequenced outside of China.
The only option at that point is to posit an unlikely set of circumstances (not the direction you want to go in a Bayesian analysis to support your hypothesis) or to posit that Huanan market magically had far higher transmission rates than anywhere else in Wuhan. A handful of anecdotes about outbreaks in markets elsewhere doesn’t cut it; this contention is absurd and totally unjustified. Not consistent with onset date curves for cases linked and unlinked to the market. Not consistent with documented early transmission epidemiology where you find, as expected from SARS and MERS outbreaks, healthcare-associated transmission before infection control ramped up through January.
***Antonio has invented a theory where there was an earlier (Lineage A) covid outbreak which was successfully contained (how exactly? the only thing that's contained covid successfully is full city lockdowns) but then still spreads quietly at a low level and then shows up again at the Huanan market.***
Yeah, I don't know what he is referring to there. 'What Really Happened in Wuhan' refers to some school closures in November 2019 and doctors being told of a particularly severe flu season but otherwise there isn't anything about wider bio containment measures.
***Okay, first, did it only show up at the Huanan market? Because, if so, you're stuck with the same crazy coincidence that the lab virus only shows up at the single most likely location in Wuhan for a natural origin.***
The earliest Lineage A cases had no market link so this scenario seems unlikely. There's an argument they were closer than would randomly be expected although close here is referring to one which was 2.32 km away. There's also a potential undercount in the WHO cases if Demaneuf is correct. Stoyan and Chiu summarize some of the shortcomings with the data too.
***Or, was it everywhere in Wuhan, but only seen at the market? Well then, you've again defeated your theory. It can't be everywhere by December 1st (violates the exponential growth model I presented, and other reasonable models on case numbers). It doesn't match the Wuhan serology data or the excess pneumonia deaths, or any of the other data pointing to the north side of the river. It also violates the theory that Antonio just created where Lineage A was successfully contained. But then Lineage B is not contained somehow and it's everywhere?***
These are good points although the second twitter thread I linked above suggests A was in fact detected in various other places (outside Wuhan) before B?
The excess death data is interesting. There is an analysis by Mei He and Dunn that looked at unofficial cremation data fwiw. https://papers.ssrn.com/sol3/Delivery.cfm/SSRN_ID3616277_code78166.pdf?abstractid=3540636&mirid=1
@MikePa67d You need to be specific about which of their back of the envelope estimates of cremation data and similar that you are referring to because they’ve published multiple, wildly different, estimates. None are particularly credible i.e. consistent with other data. Early estimates of the range of the likely number of infections in China circa January 2020 based exclusively on data from outside of China match data reported from China and models of spread based on that data from both inside and outside of China. The genome sequences do not lie. It’s simply impossible that there was a large number of SARS-CoV-2 infections prior to December 2019. The only scenarios that are possible to find plausible without abandoning data have very little SARS2 in humans prior to December 2019.
Stoyan and Chiu do not know what “epidemiological link” means and misrepresent the paper they criticize twice just in the abstract. Citing the latest paper that can’t even get the abstract right is not a great argument unless your audience has no attention span.
@MikePa67d Regarding Lineage A. The original Rootclaim analysis gives a 2X advantage for zoonosis over lab leak based on Huanan coincidence. This low number is based largely on Lineage A not being found in Huanan market or market-linked patients. I disagree with the strength of this evidence — more customers than employees and the primary lineage A case was almost certainly not found— but OK.
Subsequently, there’s a Lineage A sample reported in Huanan market environmental sequencing. The second earliest collection date of any Lineage A sample.
So, how did this loss of the key supporting evidence impact the Rootclaim analysis? It did not impact it at all. The Huanan coincidence still gives
a 2X advantage to zoonosis over lab leak.
@zcoli Here‘s where the Oct/2020 analysis implicitly appeals to lack of lineage A in market:
Several samples from patients with exposure to the Huanan market had identical virus genomes, suggesting that they may have been part of a cluster. However, the sequence data also showed that some diversity of viruses already existed in the early phase of the outbreak in Wuhan, suggesting that the virus had already begun to spread and evolve from another location prior to this.
Here’s where the new evidence is dismissed. Note that lineage A is 1 in 4 sequenced genomes from Huanan market environmental sampling. Is that different enough from one third?
All infections are lineage B, as are all but one environmental sample. While elsewhere in Wuhan, lineage A was a third of cases.
@MikePa67d I used to wonder if the death counts in Wuhan were accurate. After studying more of the data, like the seroprevalence, I realized that China is not suppressing larger numbers of deaths. I wrote a little bit about that here:
https://medium.com/p/8b3d8fd262dd
If you backtrack the growth rate from those final numbers you come up with the conclusion that @zcoli mentioned -- there just weren't many covid cases at the time the market outbreak was first detected. The only lab leak scenario that fits the data is one where someone goes straight from lab to the market within roughly the first 10 to 40 cases.
In any case, I think you've seen the problem with your initial statement that the debate "format wasn't ideal in hindsight given various incorrect claims weren't picked up".
I think it was a pretty good format that Saar invented, where both sides introduced claims, tried to correct each other, and the judges asked questions of both. There was a strong incentive for both sides to challenge any false claims, and both sides did challenge many things.
Other people on Twitter may now invent completely different claims which were not brought up in the debate. And most of those claims will probably be false, just like most things on Twitter.
Saar has now proposed a different written contest format. It might be better or worse, it's hard to say. If it's well designed, it should also do better than Twitter at sorting through true and false claims.
In either case, the great part about the contest is that the discussion ends eventually, with some judgment and resolution, as opposed to Twitter where people just keep having the same arguments for years on end.
@SaarWilf By what possible definition do "hospitals" lack the required conditions to become a superspreading location?
And what data even shows that HSM was a superspreading location? See here - https://www.nejm.org/doi/10.1056/NEJMoa2001316?url_ver=Z39.88-2003 - in the debate you presented a model in which incredibly high transmission rates within HSM are invoked to explain quite a bit; the problem is that there is no evidence that I am aware of to support this.
One of the features of SARS and MERS outbreaks is heterogeneity in transmissibility, and in particular the occurrence of super-spreading events, particularly in hospitals.16 Super-spreading events have not yet been identified for NCIP, but they could become a feature as the epidemic progresses.
@SaarWilf Here, with "spillovers were repeatedly occurring in restaurants" you are comparing to a situation that wasn't reported to exist in 2019 in Wuhan. How is this appropriate?
See: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3367621/
Epidemiologic investigations showed that 2 of 4 patients with severe acute respiratory syndrome (SARS) identified in the winter of 2003–2004 were a waitress at a restaurant in Guangzhou, China, that served palm civets as food and a customer who ate in the restaurant a short distance from animal cages.
@SaarWilf Last question for now. How is it not improper cherry picking to note what you consider evidence against raccoon dog infection here (you misunderstand the evidence badly and I suspect this was covered in the debate already), and not note that one of the animals from which virus was isolated in Shenzhen was a raccoon dog? Also, regarding "phylogenetically distinct" you may not be aware that these are sequences from viruses isolated from a cell type known to select for mutations in culture: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7166449/
clean insert - insertion with nothing around it that would make it look like a result of an evolutionary process. For example, in the debate Peter tried to claim an insert near the HKU1 FCS is just a big a coincidence. A quick look at the HKU1 alignment with its closest relative is enough to understand the difference. Attaching the SARS2 alignment for comparison.
Hospitals are definitely a likely location for early clusters, but they're not as good as HSM as they have no cold surfaces, and far better hygiene, especially with respiratory viruses.
Important to remember our claim is not that HSM is the most likely early cluster. Just that it is not some random location as the judges concluded. If we give it >1% probability of forming an early cluster it has near zero weight as evidence.
More here: https://twitter.com/Rootclaim/status/1753353163178913815
What is the importance of the infected raccoon dog in Shenzhen?
Can you elaborate on the relevance of this study to the claim? https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7166449/
@SaarWilf The relevance of my points is clear and responding by pretending it's unclear so that you can pretend to win if I don't return is a boring filibuster.
Restaurants in Wuhan in 2019 didn't have customers dining next to civet cages. Expectation of repeat of restaurant infections is absurd.
If SARS isolated from a raccoon dog where SARS was first detected in a wildlife industry reservoir was not important, you wouldn't have omitted it.
This is the same cell line used to isolate the Shenzhen samples; a few mutations are totally unsurprising even without further passage. "The same study detected multiple infected animals with phylogenetically distinct strains." is deriving some meaning from this but it's unclear how this adjustment impacts the guesstimate.
These "clean inserts" during the SARS-CoV-2 pandemic happen countless times every single day. Sometimes they're transmitted and sequenced and these events are part of the evolutionary history since 2020 of just about every variant of concern during the pandemic.
This gets us to the last point on HKU1. Whoever is advising you on this is trying to sabotage your reputation. Your alignment appears to be S sequences from human coronaviruses HKU1 and OC43. I can't vouch precisely for this reconstruction of history, but it's safe to say that however long it has been, there's been quite a bit of drift relative to SARS-CoV-2 and RaTG13.
(https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9320361/)
How much does the likelihood drop from 99.8% now that we've figured out what "clean insert" means to you?
@zcoli please edit your response to remove any accusations of dishonest behavior. i will respond then. thank you.
@SaarWilf Your entire case requires that countless people are lying to cover up the cause of death of many millions of people for years. Your case failed and now you're doubling down on your contention that all of the people you won't name are liars. It appears that you perfectly balance all of the evidence that could exist and doesn't with something about a database at a date that can't plausibly be associated with human-to-human transmission of SARS-CoV-2 and a conspiracy theory about a picture on a website.
@zcoli Your tone in every single Covid thread is dispiriting. I don’t know why anyone who doesn’t antecedently agree with you would want to continue arguing with you. You hide behind anonymity to cast aspersions on anyone who doesn’t toe the party line. I don’t know what your personal relation to Peter is but he’s doing better without you. He doesn’t need a guard dog.
@NicoDelon Peter got his $100k and needs no defense. And yes it's dispiriting to see Saar blow up an opportunity to prove that his method works and pivot to act on what judges concluded was very likely to be the truth. I hope that changes and my bet on this doesn't pay off.
@zcoli will gladly respond, as long as it's an open discussion assuming both side are honest and willing to correct.
@zcoli the market origin theory assumes people are covering up evidence of positive animals. Not sure why this entails any different level of cover up to someone getting infected in a lab. Particularly given such infection could have been asymptomatic.
@MikePa67d That's not true at all. This is a black market that survives twice-annual surprise inspections run by people who've stayed ahead of regulators since the post-SARS wildlife crackdown. News media and inspectors showed up on 31/Dec; susceptible animals were gone and potentially related wildlife stalls were closed. This isn't some implausible grand cover up here; it's people who necessarily conceal aspects of their business twice a year as a precondition of being in business.