Regarding new variants, what is the relevant metric to think about generation of relevant mutations?
Infected person days (somewhat weighted by severity since a person with a more severe infection has probably more cells getting infected)? But you probably often need multiple mutations that work well together, so depth of the phylogenetic tree probably matters as well (as opposed to width, which Omicron certainly generates a lot of).
In the end, the distance and especially the path to a new optimum are of course the important factors, but for this we’d could (as a first order approximation) use the difficulty going from “wild type” to Alpha.
I feel like estimating the breadth and depth of phylogenetic “search in the genomic space” lends itself so well to mathematical modelling that it should be a pretty well-covered topic. Given that (and current infection estimates), one should be able to get a good estimate of the likelihood per day that a new interesting variant occurs. That is of course unless Omega found a local optimum that is very hard to escape from.
So, sorry for what's probably a dumb-ass question if one is familiar with trading, but I am not, so here goes:
When you say you want to 'sell' any of these, it's clearly that means you think there's a lower chance of these things happening than the market prices imply, so if you were actually holding contracts that would yield $1 if they happened, you'd sell them at this price, and down to some lower price (either your sell-to or honest-guess price, depending on what you're up to). I get that. However, I do not currently own any such contracts, and it doesn't seem like you do either. So, when you say you want to 'sell' these, is that equivalent to saying you want to buy the no contracts at their current price and then up to whatever the mirror of your sell-to price for yes is?
This, to me, seems like it would be the case, but I wanted to check if there's some crucial distinction between selling yes and buying no that I'd failed to pick up on (likely something to do with how fees are going to work out, I'd guess, or maybe with what seeing the prices does to your estimates of the probability?)
Well. You do want to make sure to always buy/sell the least expensive side because the liquidity fee (generally around 2%) is applied only to the shares you buy/sell. For instance, say a market is priced at precisely 98/2 (Y/N). If you buy the Y shares at 98, you will be paying nearly 100 after fees! However, if you instead sell N shares, you only pay a negligible amount in fees (2% of 2 cents instead of 98 cents). If you don't already have shares, you will want to split shares (generate equal numbers of both sides using a determination amount of cash, so e.g. $1000 to get 1000 shares of each side). The method to do this on the UI is to add liquidity and then remove liquidity. If you add e.g. $1000 to liquidity and then remove it, you'll have 1000 shares of each side in the market (presuming no one bought shares in the seconds in between). The fees saved by splitting will become larger the further price is away from 50/50. Of course, if you already want the cheaper side you just buy it normally.
See this bit has left me a tad in the shade. I'm thinking I should get into actually doing this, because I'm convinced it's good rationality training and whatnot, but I need to understand *exactly* what I'm doing, or I'll likely find some way to mess it up.
I realise explaining this kind of thing very slowly to complete trading newbies is likely not fun, and if your answer is just "read some basic trading glossaries or introductions to how it all works" then, well, fair enough I probably should do that at this point. But, if you feel like it, do you mind expanding slightly on what you mean whe nyou talk about adding liquidity?
It seems to me that you're suggesting that, to essentially bet the more expensive side, you're better off buying *both* sides then selling the cheaper side, but how does that get you into a better place re: fees than just buying the more expensive side. Isn't it just three transactions instead of one to get you to exactly the same place (holding the more expensive side)? Is it that adding liquidity attracts no fees? That's the only way I can make that make sense.
Also, I am thinking that, even if you fancy explaining this to me right now, I do need to read some sort of 101 content so I just know what all the terms in posts like these are (what is "picking off", for instance?) When it's a simple "bet one side or the other", I can generally just back-form what the terms mean from context based on what the probability estimates given imply about what needs to happen, and generally when there's a simple arbitrage, it's obvious to me what the correct strategy is that I can just sort of work out the terminology by assuming it means what it needs to mean to make it make sense. I think I've grasped "put", "call", and "straddle" by this method as well, but some other stuff I see just leaves me with too few finger-holds to get a sense of anything. Is there a common reference for this stuff from the perspective of prediction markets, or is pretty much everyone who is into prediction markets already familiar enough with trading terminology that no such resource has been considered valuable enough to create, and I just need to hit some basic "trading for dummies" content?
So basically, splitting shares means for X amount of dollars you generate X amount of each bracket in a contract (So X shares of Yes and X shares of No). This does not cost any fees. There is no direct option on the UI/Website to split shares, but you can do this indirectly by adding liquidity (There is an Add Liquidity Button) and then removing all the liquidity (Remove Button will appear once liquidity has been added and the page automatically refreshes with the option). The end result is the same as splitting shares. So if you added $1000 in liquidity and then remove it immediately, you'll have 1000 Y shares and 1000 N shares. You can then sell the cheaper side and save on fees. This assumes no one bought shares in the seconds your liquidity was in the market. Usually, this is the case. But you should not do this during live events like sports betting when trades are going through every second.
Likewise, if you own shares worth more than 50 cents and want to exit, it will be cheaper to buy the other side for less than 50 and hit the "Merge Shares" option afterwards. Merge shares is the opposite of splitting shares and takes equal amounts of all sides and turns that into dollars.
Note that while I believe everything aenews says here is accurate, getting this wrong is a small mistake unless you are trying to bet large amounts with a small edge. It's simply about executing the best way you can on principle, rather than there being so much at stake.
I don't agree. One should always buy/sell the side with cheaper fees. If the price is close to 50/50, then the benefit will be relatively small. If price is e.g. 80-100 cents, then it's a massive difference. At 90 cents, you'd be paying 9X the fees if not using the liquidity trick. At 95 cents, 19X the fees. At just over 98 cents, you can't make money at all without splitting shares.
The only exception would be for a fast-moving market.
This is not a complicated procedure. There is an "Add Liquidity" button on every market page. You just need to add liquidity (add in dollars however many shares you wish to sell of the other side), and then hit the "Remove Liquidity" option once the liquidity is added and the option appears. Then you can sell shares at your discretion when you want.
I recommend trying this out with a couple cents or a dollar to see what I mean and familiarize yourself with the options.
And if you already have the currently more expensive side of a contract and are looking to exit, then there's pretty much no benefit to selling directly. You'd definitely want to buy the other side and then merge out. It's the same thing but cheaper.
I’m curious how that mask requirement question will resolve if it remains permanently in effect but with vague and unenforceable conditions like “everyone must wear a mask if they believe they have been exposed to an airborne illness in the last two weeks,” which seems like a plausible outcome to me.
That’s probably reasonable. I think the intent of such a mandate would be to disable mask requirements without appearing to take a political side, so your interpretation captures the spirit.
Assuming what we care about for the purposes of this question is the necessity of masks due to real danger vs. the state of the politics surrounding covid.
While I agreed with your impression that France will likely stay above ES/IT/US, I think you jinxed it: https://imgur.com/a/0paoS6Y
The very top of the graph is the last data point available when when you published the post.
It reminds me of a french philosopher known for his work on the topic "COVID modelling doesn't work, forget about it".
Maybe the French gave up and stopped testing. Or it's really related to the graph with substructure model from Lemoine, with France having maybe sparser clusters but higher connectivity between them (why that would be the case I have no idea).
I think this is an artifact in the data and people have sold off in a panic. OurWorldInData seems to get their numbers from CSSE and they have 177,895 new cases for 19/1 which is far far lower than what other sites and more importantly the WHO report, which is around 430k cases. Not only is this not lower than previous, it's the highest number of daily cases they've ever had. I don't know if the discrepancy is because of the update time for the CSSE data or something else. My impression is that the low numbers for the 17th & 18th coupled with the above have created a false decline in the graphs. I'm around 90% confident at the moment that France is still a buy and betting accordingly.
I just checked again and it seems I was indeed very wrong.
I kind of assumed that that these data-pipelines are stable (almost 2 years into the crisis) but now that I thought about it a little more I wonder how I arrived at that impression.
Even the graph from my picture looks very wrong (a 40% decrease in a 7 day rolling avg - wtf).
Regarding new variants, what is the relevant metric to think about generation of relevant mutations?
Infected person days (somewhat weighted by severity since a person with a more severe infection has probably more cells getting infected)? But you probably often need multiple mutations that work well together, so depth of the phylogenetic tree probably matters as well (as opposed to width, which Omicron certainly generates a lot of).
In the end, the distance and especially the path to a new optimum are of course the important factors, but for this we’d could (as a first order approximation) use the difficulty going from “wild type” to Alpha.
I feel like estimating the breadth and depth of phylogenetic “search in the genomic space” lends itself so well to mathematical modelling that it should be a pretty well-covered topic. Given that (and current infection estimates), one should be able to get a good estimate of the likelihood per day that a new interesting variant occurs. That is of course unless Omega found a local optimum that is very hard to escape from.
Also: I am too late and someone who agrees with you made all those trades :P
Given the poor liquidity, it probably only took a single person with some small k$s in funds for this to happen.
So, sorry for what's probably a dumb-ass question if one is familiar with trading, but I am not, so here goes:
When you say you want to 'sell' any of these, it's clearly that means you think there's a lower chance of these things happening than the market prices imply, so if you were actually holding contracts that would yield $1 if they happened, you'd sell them at this price, and down to some lower price (either your sell-to or honest-guess price, depending on what you're up to). I get that. However, I do not currently own any such contracts, and it doesn't seem like you do either. So, when you say you want to 'sell' these, is that equivalent to saying you want to buy the no contracts at their current price and then up to whatever the mirror of your sell-to price for yes is?
This, to me, seems like it would be the case, but I wanted to check if there's some crucial distinction between selling yes and buying no that I'd failed to pick up on (likely something to do with how fees are going to work out, I'd guess, or maybe with what seeing the prices does to your estimates of the probability?)
Yeah, you can simply buy the No contract, that was what I intended there.
Well. You do want to make sure to always buy/sell the least expensive side because the liquidity fee (generally around 2%) is applied only to the shares you buy/sell. For instance, say a market is priced at precisely 98/2 (Y/N). If you buy the Y shares at 98, you will be paying nearly 100 after fees! However, if you instead sell N shares, you only pay a negligible amount in fees (2% of 2 cents instead of 98 cents). If you don't already have shares, you will want to split shares (generate equal numbers of both sides using a determination amount of cash, so e.g. $1000 to get 1000 shares of each side). The method to do this on the UI is to add liquidity and then remove liquidity. If you add e.g. $1000 to liquidity and then remove it, you'll have 1000 shares of each side in the market (presuming no one bought shares in the seconds in between). The fees saved by splitting will become larger the further price is away from 50/50. Of course, if you already want the cheaper side you just buy it normally.
See this bit has left me a tad in the shade. I'm thinking I should get into actually doing this, because I'm convinced it's good rationality training and whatnot, but I need to understand *exactly* what I'm doing, or I'll likely find some way to mess it up.
I realise explaining this kind of thing very slowly to complete trading newbies is likely not fun, and if your answer is just "read some basic trading glossaries or introductions to how it all works" then, well, fair enough I probably should do that at this point. But, if you feel like it, do you mind expanding slightly on what you mean whe nyou talk about adding liquidity?
It seems to me that you're suggesting that, to essentially bet the more expensive side, you're better off buying *both* sides then selling the cheaper side, but how does that get you into a better place re: fees than just buying the more expensive side. Isn't it just three transactions instead of one to get you to exactly the same place (holding the more expensive side)? Is it that adding liquidity attracts no fees? That's the only way I can make that make sense.
Also, I am thinking that, even if you fancy explaining this to me right now, I do need to read some sort of 101 content so I just know what all the terms in posts like these are (what is "picking off", for instance?) When it's a simple "bet one side or the other", I can generally just back-form what the terms mean from context based on what the probability estimates given imply about what needs to happen, and generally when there's a simple arbitrage, it's obvious to me what the correct strategy is that I can just sort of work out the terminology by assuming it means what it needs to mean to make it make sense. I think I've grasped "put", "call", and "straddle" by this method as well, but some other stuff I see just leaves me with too few finger-holds to get a sense of anything. Is there a common reference for this stuff from the perspective of prediction markets, or is pretty much everyone who is into prediction markets already familiar enough with trading terminology that no such resource has been considered valuable enough to create, and I just need to hit some basic "trading for dummies" content?
So basically, splitting shares means for X amount of dollars you generate X amount of each bracket in a contract (So X shares of Yes and X shares of No). This does not cost any fees. There is no direct option on the UI/Website to split shares, but you can do this indirectly by adding liquidity (There is an Add Liquidity Button) and then removing all the liquidity (Remove Button will appear once liquidity has been added and the page automatically refreshes with the option). The end result is the same as splitting shares. So if you added $1000 in liquidity and then remove it immediately, you'll have 1000 Y shares and 1000 N shares. You can then sell the cheaper side and save on fees. This assumes no one bought shares in the seconds your liquidity was in the market. Usually, this is the case. But you should not do this during live events like sports betting when trades are going through every second.
Likewise, if you own shares worth more than 50 cents and want to exit, it will be cheaper to buy the other side for less than 50 and hit the "Merge Shares" option afterwards. Merge shares is the opposite of splitting shares and takes equal amounts of all sides and turns that into dollars.
Note that while I believe everything aenews says here is accurate, getting this wrong is a small mistake unless you are trying to bet large amounts with a small edge. It's simply about executing the best way you can on principle, rather than there being so much at stake.
I don't agree. One should always buy/sell the side with cheaper fees. If the price is close to 50/50, then the benefit will be relatively small. If price is e.g. 80-100 cents, then it's a massive difference. At 90 cents, you'd be paying 9X the fees if not using the liquidity trick. At 95 cents, 19X the fees. At just over 98 cents, you can't make money at all without splitting shares.
The only exception would be for a fast-moving market.
This is not a complicated procedure. There is an "Add Liquidity" button on every market page. You just need to add liquidity (add in dollars however many shares you wish to sell of the other side), and then hit the "Remove Liquidity" option once the liquidity is added and the option appears. Then you can sell shares at your discretion when you want.
I recommend trying this out with a couple cents or a dollar to see what I mean and familiarize yourself with the options.
And if you already have the currently more expensive side of a contract and are looking to exit, then there's pretty much no benefit to selling directly. You'd definitely want to buy the other side and then merge out. It's the same thing but cheaper.
I’m curious how that mask requirement question will resolve if it remains permanently in effect but with vague and unenforceable conditions like “everyone must wear a mask if they believe they have been exposed to an airborne illness in the last two weeks,” which seems like a plausible outcome to me.
If I can choose to not wear a mask on a plane, there's no mandate. If I can't, there is one?
That’s probably reasonable. I think the intent of such a mandate would be to disable mask requirements without appearing to take a political side, so your interpretation captures the spirit.
Assuming what we care about for the purposes of this question is the necessity of masks due to real danger vs. the state of the politics surrounding covid.
While I agreed with your impression that France will likely stay above ES/IT/US, I think you jinxed it: https://imgur.com/a/0paoS6Y
The very top of the graph is the last data point available when when you published the post.
It reminds me of a french philosopher known for his work on the topic "COVID modelling doesn't work, forget about it".
Maybe the French gave up and stopped testing. Or it's really related to the graph with substructure model from Lemoine, with France having maybe sparser clusters but higher connectivity between them (why that would be the case I have no idea).
If all it takes to get case counts to come down is a probability assessment, Viva la France.
I think this is an artifact in the data and people have sold off in a panic. OurWorldInData seems to get their numbers from CSSE and they have 177,895 new cases for 19/1 which is far far lower than what other sites and more importantly the WHO report, which is around 430k cases. Not only is this not lower than previous, it's the highest number of daily cases they've ever had. I don't know if the discrepancy is because of the update time for the CSSE data or something else. My impression is that the low numbers for the 17th & 18th coupled with the above have created a false decline in the graphs. I'm around 90% confident at the moment that France is still a buy and betting accordingly.
There we go, OurWorldInData just updated
https://i.imgur.com/t3Y7sRQ.png
I just checked again and it seems I was indeed very wrong.
I kind of assumed that that these data-pipelines are stable (almost 2 years into the crisis) but now that I thought about it a little more I wonder how I arrived at that impression.
Even the graph from my picture looks very wrong (a 40% decrease in a 7 day rolling avg - wtf).
I take the L and the shame.