Options
Why the early results on the night might be deceptive – politicalbetting.com
A week to go and perhaps a good moment to pass on a observation about how information might come out on election night because of the different ways that each state will out count.
0
This discussion has been closed.
Comments
Unless Texas or Florida flip.
Actually happened it seems, although the tweet is a joke
I thought I'd have a look at the false positives issue while using actual maths rather than Tobymaths.
Assuming:
Sensitivity set to 80% as reported (so 20% false negatives from actual cases)
Specificity is under discussion, so using the logic that you can't have more false positives than measured positives and looking at the positivity rate at the lowest prevalence (4th July, by the cases measured, was 0.349%, and this wasn't uniquely low - 12 of the days between the start of July and mid-August had sub-0.4% positivity) to set an upper bound. Assuming that the UK wasn't totally covid free and some people (of those singled out as most at risk of having it from symptoms and exposure) would get true positives, I assigned 0.3% as the highest plausible false positive rate; so specificity set to 99.7%.
Using the Bayes theorem and using the measured positivity as a reasonable prior for the prevalence in the sample, and then iterating by taking the calculated true positives, dividing them by the number of samples for a calculated prevalence to improve the prior, and iterating 30 times until the prevalence in was equal to the prevalence out to within less than a thousandth of a percentage point.
We get this:
It's quite hard to present enough information in this sort of range, but:
- The dark gold/brown columns are the measured cases (the reported cases from testing)
- The black columns are the expected false positives
- The green columns are the expected false negatives
- The red columns are the net error (the amount by which the measured and true cases differ - take away the false positives (people without covid who were incorrectly reported as having it) and add the false negatives (people with covid who were tested and incorrectly reported as being clear)
- The see-through columns are the calculated true cases, when false positives and negatives are taken into account.
Conclusions:
- When the virus is at very low levels, we'll see up to a few hundred extra cases per day incorrectly reported (if the false positive rate is about as high as is reasonably plausible)
- When it starts to increase, the false positives and false negatives balance out
- When it gets to a reasonably high spread (ie the start of September), the false negatives start to outweight the false positives, and we end up underreporting cases
- The false negatives will be currently outweighing false positives to the tune of a few thousand cases
One further worrying conclusion: false positives and false negatives mean that the rate of true cases has been increasing faster than reported cases (I doubt that Toby and co meant for this to be the case):
If you have been involved in a motoring accident, the law requires you to stop at the scene and exchange correct personal details with any person who has reasonable grounds to request them. If details are not exchanged you must report to the police as soon as reasonably practicable and in any event within 24 hours. Failing to stop at the scene of an accident involving injury or damage is an offence which carries a maximum penalty of 6 months imprisonment.
https://cartwrightking.co.uk/areas-of-practice/motoring-law/failing-stop-after-accident
POLICE VERSION
"A Metropolitan Police spokesman told MailOnline: 'Police were alerted by LAS at around 12.20hrs on Sunday 25 October to a report of a road traffic collision between a cyclist and car in Grafton Road, NW5.
'The driver of the car had stopped at the scene and exchanged details with the cyclist but had left before officers arrived.
'Officers later attempted to contact the driver of the car and left a message advising him to report the matter to police.
'The driver of the car subsequently attended a north London police station"
& SIR KEIR'S VERSION
"A spokesperson for Sir Keir said that the Labour leader was involved in a minor road traffic accident and spoke to a police officer who attended the scene and swapped details with the officer and others involved.
Sir Keir stayed at the scene until the ambulance arrived then reported the incident to a police station, and has since been in touch with the other person involved, the spokesperson added."
https://www.dailymail.co.uk/news/article-8882229/Police-launch-investigation-Labour-leader-Sir-Keir-Starmer-involved-road-accident.html
Shame no Keir exit date market on betfair!
So what are you going on about ?
I can't believe he wouldn't have acted appropriately to be honest. Quite possible he contacted the police station as claimed, and they were contacting him at the same time
https://www.investors.com/news/biden-vs-trump-poll-joe-biden-lead-shrinks-donald-trump-tops-2016-vote-share-ibd-tipp-presidential-poll/
The bolded segment is a very lawyerly statement. Most people will read that and envision meaning speaking to the plod as they arrived at the scene, however that doesn't actually have to mean that. Furthermore, most normal people would say that they spoke to a police officer AT the scene, if that is what they did.
I did look to lay him for next PM! 3.95, I didnt bother
Indeed, the greatest betting opportunities (Brexit, US Presidential election 2016) come when the market is slow to respond to new evidence because of preconceived ideas about the likely result.
From memory in 2016, the Nytimes had a really good election needle which auto-compared county results with previous county results and spat out a prediction (which was pretty accurate) as to how well either candidate was doing in a given state.
This is great for journalism, not so helpful to pundits as it makes it rather easy for the rest of the market to be well informed. I was considering trying to make my own version, but suspect it's not worth it as there will be enough data journalists doing this and doing it better than I can.
Last time, astoundingly, I was able to recoup about half of my losses on the popular vote market because California was slow in counting and it was obviously going for Hilary.
I'm hoping a similar such opportunity will present itself on the night... I suspect turnout %/total number of votes for either candidate might be a profitable market *if* you can get the math right.
"'Officers later attempted to contact the driver of the car and left a message advising him to report the matter to police."
However that can be at a police station within 24 hours ,as many slight injuries are done everyday.
The case evidence shows that even in the most isolated spots in the country, there were cases bubbling away asymptomatically - every now and then cases popping up. The virus never went away.
I used to live in KT. My local was The Vine, although I used to go on quite a few pub crawls
https://www.dailymail.co.uk/news/article-8884757/British-Transport-Police-no-record-speaking-Keir-Starmer-scene-crash-cyclist.html
As I pointed out the statement they then put out is much more lawyerly. It may well be that the confusion via the spokesperson, is due to what most people would have done and interpret a carefully worded statement as speaking to them AT the scene, not speaking to them AFTER the incident.
https://twitter.com/RobertCahaly/status/1321103067933249536?s=20
Some time ago the police change the naming of accidents from road traffic accidents to road traffic collisions.
As in theory some one is at fault to cause a collision.
The moment they lose either the threat is gone. There is a very, very narrow window of opportunity here to act. Once the GOP contain a blocking majority in the Senate what would constrain the SCOTUS?
Do not know if the RTC occurred near one.
It will be good cannon fodder for Boris at PMQ.
The account came from the LOTO’s office, which apparently provided the badge number of the officer Starmer says he spoke to.
"Argentina locked down early and hard. Now cases are exploding.
When the coronavirus first reached Argentina, Andrés Bonicalzi steeled himself for the sacrifices to come. A lawyer in Buenos Aires, he started working from home, canceled his weekly visits with his parents and vowed to keep his son inside. The government announced one of the world’s strictest lockdowns. The next few weeks would be difficult.
But those hard weeks have turned into seven months, and much of Argentina’s quarantine, believed to be the world’s longest, is still dragging on.
So much sacrifice, Bonicalzi sometimes thinks, and for what? The South American country has become one of the coronavirus’s most explosive breeding grounds. In early August, fewer than 200,000 Argentines had contracted the virus. That number has since surged to 1.1 million — 1 out of every 44 people — and 28,000 are dead."
https://www.washingtonpost.com/world/the_americas/coronavirus-argentina-million-quarantine-lockdown/2020/10/26/65eefde2-149c-11eb-bc10-40b25382f1be_story.html
If Trump scrapes all three I don't think we'll know the result as Biden will be depending on winning PA/ MI/ Wis and they all count the early votes later. Very frustrating.
"But the rallies" is going to be the "we won the argument" line in a week's time after the 🌊 hits.
Haha who is "my boy"? Boris? Farage? Burnham?!
Well worth staying up for the first reports - and indeed the PB Zoom meet if possible.
Perhaps you could explain to me exactly how you are "smelling" what's happening across the US right now.
I can't keep up with it!
https://slate.com/news-and-politics/2020/10/brett-kavanaugh-mail-ballots-trump-fraud.html
... George W. Bush’s 2000 election legal team—which included Barrett, Kavanaugh, and Roberts—argued during that contested election that ballots arriving late and without postmarks, which were thought to benefit Bush, must be counted in Florida...
Or are they just shameless hacks ?
(tbf to Roberts, he’s clearly not completely impervious to the occasional pang of conscience.)
Colorado
Michigan
Arizona
Maine
North Carolina
Iowa
Montana
South Carolina
Alabama
Georgia
They have Iowa and Montana down as the toss ups, everything above leaning Dem and everything below leaning Rep
We cannot set it to higher than that, because we'd be in the impossible position of saying that more than 400 out of those 400 were false positives.
The ONS data is separate; we're looking only at false positives for these.
It is arguable that the true prevalence had to have been at least 0.2% at random and would necessarily have been higher in the sample (and setting the false positives at 0.3% ends up with a true prevalence needing to be 0.064%), the false positive rate selected for this analysis must be way too high.
Running it again until the lowest true prevalence I get in the test sample is 0.25% gives a false positive rate of 0.15%.
Always be the bigger man, I could learn a lot from you