In "My 2020 Calibration", the graph is wrong. Your x-axis has 95 inserted between 90 and 100, but the spacing for these gaps of 5 is the same as the spacing for gaps of 10 elsewhere in the graph, making the axis abruptly nonlinear.. Your black line going "straight" from (50, 50) to (100, 100) therefore does not pass through (60, 60) or (70, 70) etc.; in order to do that, it would need to bend sharply at (90, 90) when the x-axis suddenly stretches out.
Also, in the text above the graph, you said "Of my 50%..." over and over, instead of changing the number.
Medvedev seems to be making a bid for the role of the court jester, which was long held by a certain Vladimir Zhirinovsky, who died last year. He'd been decently popular due to his absurd antics (and ultra-nationalist rhetoric), one of the more notable ones was "predicting" in 2002 that elite Iraqi soldiers would rip Bush's army to shreds.
I always found political prediction markets interesting. Suppose you were a generic Republican. Why not ask yourself how much you will need to be ambivalent about Dems winning the next election and put money on them? Either way, you seem to win.
"Stock" status of 0%-100% could work if you conceptualize each person as having good and bad status points. If Elon Musk was sitting at 50% pre-Tesla, I could imagine him post-Tesla with 10x as many good points and 1/2 as many bad points, which by my math would put him at 95%. Someone with no status quanta either way would be undefined, which seems reasonable.
Legality of prediction markets. CFTC won't let a commodities market engage in gaming, but, sports betting has leaked out an been legalized pretty much everywhere in the country as is evidenced by the advertisements on NFL games this season. This all occurred because of a Supreme Court overturning a federal ban on sports betting https://www.espn.com/chalk/story/_/id/23501236/supreme-court-strikes-federal-law-prohibiting-sports-gambling
The states responded by enacting legal frameworks for sports betting with the approval and connivance of the major professional sports leagues.
Given the size of that hole in the bans, I would guess that we will see British style omnibus bookmaking in the US in the next few years. Election bets will be a very important product. Could prediction markets be far behind.
Don't be shocked by the CFTC's behavior. It is pretty typical of the alphabet agencies that I had the displeasure of dealing with when I was active in those fields before retirement.
It is a violation of their charters and of administrative law, but they know that it is extraordinarily expensive and time consuming to hold them to account in the Federal courts.
What is the chance the Department of Justice will enforce the national mandatory training of police in appropriate ways of handling criminal suspects?
1. Police can't be exempt from prosecution for crimes. James Bond is fiction; no one has a license to kill.
2. Police can't be exempt from prosecution for violating a criminal suspect's civil rights.
If the feds hold individual departments' feet to the fire, and the training they've provided is truly rigorous, departments should support officers following humane and appropriate treatment of criminal suspects, even when they're sued.
If the officers do not follow their DOJ-directed training and handle criminal suspects humanely, the department should have no responsibility to defend them.
Request to have the 2023 contest results post out before the 2024 contest deadline - being able to learn from my mistakes in last year's contest would be helpful for the next contest, but as is I have to wait a year for that.
(I understand that writing the results post is probably the kind of surprisingly large amount of work post that's hard to write to a short deadline so this may be hard or impractical)
The part about scandal markets reminds me of Robin Hanson's arguments for the legalization of blackmail, as well as some arguments against such a position. "The median scandal is extremely stupid, but can still ruin someone’s life.", I think this is very much true and not said enough, but I'm not sure about how to solve the general problem, that is even without scandal markets, this and other related issues surrounding false allegations and such, are already such a perverse problem. And I don't think its reasonable to expect the general public to raise their epistemic standards especially on such issues. There is also the cost not just to the person who has been accused, but also to people who haven't been accused modifying their behaviour to reduce the chance of being accused.
I don't think Medvedev's joking, I think he's coming up with pleasant stories of what the future holds because anything more realistic is unpleasant for him to think about. "Of course, the EU will collapse and we will come out victorious". It's just desperation.
A stock could evaluate to the result of a large survey. "Do you like Elon Musk? Y/N." Then in the same way futures expire quarterly, the stock could expire quarterly by fielding the survey once-per-quarter.
Did anyone here use Metaculus Prediction Updates for Ukraine, if not, why not (given that you're interested in the topic at all)? If you used it, how, and why ... and how satisfied are you with what you got?
I've gone through phases of: useful, what's going on?, probably not working at all, it's back?, to name some major ones. My previous exposure to PM was limited, and I'd appreciate your thoughts on this.
I basically woke up with this question this morning (not really, but almost), and planned to post it on the next OT, but that just fits too well. :) Congrats to Metaculus.
I've noticed in manifold that when a market is surely going to resolve YES the percentage gets to something like 90~95% then stops getting better, and then maybe gets to 99+% when it's about to close and resolve. This happens because people have limited mana and would rather put it to use on other markets that would pay better. What I'm saying is very obvious, I realize, but I'm pointing it out as a problem for markets like the one about Coinbase going bankrupt. It could mean 10% is a good estimate, but it could also mean people are not interested in putting their mana in a market that closes in late 2024 for some measly gains. And the closer to the max/min value, the worse it gets.
I kinda dislike the metac.'s "example" - oops it was Scott's ... err, - I still dislike the nuclear strike take here, not that it is wrong, but it's far too short: "a 10% chance of a strike, which comes from a 15% chance if the war in Ukraine continues vs. a 5% chance if it doesn’t. And they think there’s a 50% chance the war will continue, which comes from a 60% chance if the US stops arms shipments and a 33% chance if it doesn’t - and so on." This sounds too much like: we can substantially reduce the chance of a strike if we stop to send arms. - I see the logic, but one might mention the probability that the US/West will be seen as insincere and weak then - with massive rise of risk of worse conflicts in other places, incl. those who might go nuclear + that much bigger chunks Ukraine will be under Russian rule, that decades of unrest in the region will follow, that Putin and his clone (or is it clown) Medvedev will continue their aggressions in other areas of the former CCCP. Even an average-acx-forecaster (as me) might see that "no arms for Ukraine" goes "less peace". - So, not a good example to pick. Sorry, but I am prickly about Ukraine, esp with public support so fickle. Btw: Excellent forecasting by Scott, it was a hard year to forecast.
Even restricted scandal markets seem terrible. "Will John Doe be found to have committed fraud?" strengthens the "John Doe <-> fraud" link in people's minds (including subconsciously), in search engines, and in future AIs' training data. It will also contribute to a "no smoke without fire" effect among normies and indeed anyone who stumbles across John Doe's fraud-scandal market in isolation without knowing the broader context.
> Current law includes the following provision on event contracts, [banning]:
I think this is a slightly bad / misleading summary. The whole point of that section is that the current law does *not* ban these things, but merely that it gives the CFTC the power discretion / authority to ban contracts which involve these.
With the current summary, the follow-up sentence after the list, namely "So the CFTC may ban certain prohibited categories.", makes little sense.
>If you have a better idea for how to run stocks, leave it in the comments here and they’ll probably see it.
How about sentiment analysis using a hard-to-manipulate source like Google News coverage, highly viewed social media, Wikipedia, etc.?
You could define a "sentiment index" equal to, say, log(sum([content.views * content.sentiment for content in pieces_of_content]))
You could also impute an index by randomly polling users on which things they like better than other things, then backsolve numbers which fit with the rank orderings that users give.
Once you've identified a method to create an index for any person/thing/etc. then you'd need a method to bring its price in line with reality while also allowing speculation. Perpetual futures are one approach. Or overlapping quarterly/annual futures.
Or, idea I just had -- make it so every day, there's an 0.1% chance that the asset gets a 24-hour "liquidity event" where people are allowed to buy/sell an unlimited amount of the person's stock, at their current sentiment index (as computed by an algorithm). Of course, people should be able to set up buy/sell orders in advance to take advantage of liquidity events, so they don't need to log into Manifold every day. You could configure the % daily chance of a liquidity event based on the rough timescale that you'd like to be forecasting on.
>When I first looked into this a few weeks ago, a few conspiracy-related prediction markets gave pretty high predictions, because people weren’t sure if the market creators would resolve them honestly, which naturally pushes the price towards 50.
Maybe Manifold should add an option so that when you create a market, you can nominate one or more other users to be the resolvers for that market.
At the beginning of your post you mention hostility to prediction markets from communities that handle classified nuclear information. Perhaps germane to this, a (now deceased) friend was a physicist who worked at a government research establishment that handled highly-classified information. He once told me that the quality of the science done in the unclassified part of the institution was FAR higher than that in the classified part, a phenomenon he attributed to the lack of open critique and review.
What does Scott mean by alpha? He's used it in the Bay Area posts a few times as well, but my attempts to get Google to answer this question just get me links about alpha males, which I don't think is relevant.
The 'What percentage of US electricity will be produced by solar power in 2030' is slightly misleading. It's something that segues into 'What percentage of US energy.....' in many people's minds.
But instead of solar power being four percent of US energy, it is currently nearer one percent.
Solar enthusiasts aren't keen on the distinction, though I think it's important to realise that solar is unlikely to produce more than two percent (of US energy) by 2030. A wild success would be three percent, so its impact will remain trivial, at least for a few decades.
"My 2022 Calibration" Now that it is fixed it and understandable, would it be correct to say that you really don't have a good understanding of the difference between being 60% confident and 90% confident in your prediction?
I'd liked to see your "Of my 40% predictions" (and 30%, 20%, etc) results, to see if your blind spot is symmetric.
Also how often was did the event happen for all binary predictions? Is there a bias in how prediction questions resolve - for example, mostly they happen or mostly they don't or happening or not happening is random?
The reality is that "prediction", no matter how well it works is often wildly inappropriate for the task at hand. The fact that we are getting better at it makes it all the more dangerous.
A lot of times, we let our brains think we are predicting something, when what we are supposed to be doing is gathering all the actual evidence about a given possibility, and seeing where it stands. The distinction is huge.
Idea for stocks: team (maybe with some outside input) gets together once a year to order all the figures in order of status or influence or whatever. Market is on what the index of figure X will be, from (say) 1-10 if there are 10 stock markets.
It seems quite probable that the "unmoored" status stock eventually won't track the status, but only the perceived future probability of the stock. Very vulnerable if majority of Manifold users at some point decides that is its about "lolz", not Musk popularity, or that small values are now good instead of large values, or something equally silly..
Not sure if the cap at 100% is a good thing or a bad thing. On one hand, it doesn't make obvious sense: One can sort of imagine that a person may have a zero status (lowest possible), but can they have a full 100% status?
On the other hand, a hard cap at 100% might have an useful purpose as an inbuilt check against bubble -> crash behavior. Every market participant must ask themselves "is the current growth sustainable" because at some point it must it hit 100%. And if growth is very fast, it will hit it very soon.
On the third hand, if MUSKSTATUS2000 (launched in 2000) is already at 90%, but you still think Musk will become many times more popular when Mars colony succeeds, is there anything stopping you launching another token, MUSKSTATUS2023, initially trading at 2%
The original sin here is the notion that policy makers objectively choose and promote policies based on transparent public benefit or some such textbook naïveté.
The real world of policy making involving pushing for agendas in conflict with other agendas.
Nobody involved in policy wants accuracy or objectivity or whatever. Only people outside of policy making want to break this monopoly - much like the crypto banking idiots. And much like the crypto banking idiots who refuse to understand that the speed and function of the present system is a benefit to incumbents - not an obstruction for them - their path to resort will never materialize.
I don't think we blame normies enough, hating someone over a Halloween costume or something they did as a teenager is a massive moral flaw.
You see it most clearly with women in revenge porn cases, the ex-boyfriend who releases images is rightly seen as the villain but the boss who fires a teacher who is the victim of an abusive ex, isn't seen to have blame.
Normie mobs hating on minor infractions means we can't have nice things like prediction markets on scandals and we should be angry at normie mobs.
>Having a scandal market on someone incentivizes people to dig up every scandal they’ve ever been involved in or accused of (including false allegations). Outside Encyclopedia Dramatica and KiwiFarms, most people don’t have a convenient list of every embarrassing and reputation-lowering thing they’ve ever done in their lives in one place.
Do I get any points for predicting the existence of scandal market short sellers 3 months ago?
The person who set up Scott's scandal market replied to me and said it would be neat if somebody hired a private detective to stalk people who have scandal markets, but there wasn't enough volume. I feel like we saw entirely different things in this hypothetical!
Re: making stocks more than a Ponzi scheme. It could be tied to some real measurement of the person's "status", e.g., https://en.wikipedia.org/wiki/Q_Score. The stock would have to be a prediction of some standardized measurement guaranteed to be made at regular intervals by a trusted party.
Each stock represents the question 'will questions in {this list} pay out?'. have a corresponding list of questions relevant to that person. Ex: Elon Musk would have questions such as 'will humans be on mars by 2035' and 'will the twitter deal wind up being profitable', where a more influencer-type might have questions about their follower count etc. The important thing is to break down why each public figure has their social clout, and judge them on those metrics with lower-scope questions.
you can make stocks 'reset' once per year, or devise a new model that pays off 'dividends' but is persistent.
“some people might genuinely need to hedge political events”
There is enough there to last far beyond the next cycle, especially considering the real question of when a derivative is actually second derivative rather than something else.
Here's an idea for “stock ratings” of people, given that percent ratings seem like they would not work.
What about rating them the way the rating agencies do: AAA . . .D? You can anchor each rating with descriptors appropriate to people rather than stocks: “ Reliable,” “clever”, “a good bet,””likely/unlikely to fuck up” etc. There are various ways to turn this ordinal scale into a numerical one — I’m sure many have played around with ways to do this with stock ratings, so as to make them more amenable to calculations of various kinds. Look and see what all those clever people have done.
Seems to me there are 2 factors that are determinants an individual’s “stock,” as we normally think of it: good/bad and strong/weak. (By the way, those are the 2 main factors that fall out of factor analysis of adjectives describing people.)
-Good/bad= right/wrong, kind/cruel, helpful/destructive, sane/crazy etc.The stock rating measure seems like a reasonable approximation of this.
-Strong/weak= how big a deal someone is, how much influence they have, how well known. Seems like things like Google searches and Twitter mentions are a good measure of this.
So maybe use strong/weak as a multiplier of good/bad measure. Makes pretty good intuitive sense. No matter how good or bad someone’s impact is, their personal stock isn’t valuable if they have very little impact.
> Many come up with futuristic hypotheses, as if competing to single out the wildest, and even the most absurd ones.
I would interpret it as mockery of most of these yearly predictions, which (unlike your competition, or something like the GJP) are seeing attention rather than accuracy.
To the point on Congress having created the CFTC to regulate derivatives contracts: let's not forget that there have been exactly two cases where Congress did, in fact, intervene to ban futures contracts explicitly: onions, and motion picture box office ticket sales (see https://en.wikipedia.org/wiki/Onion_Futures_Act). Clearly the height of well-considered market regulation.
Is the substack susceptible to spamming? I got this email purporting to be from Astral Codex Ten ( <forum@mg1.substack.com>) regarding Astral Codex Ten replied to a comment on Mantic Monday 1/30/2023.:
"Let's discuss👆I've got something to introduce to you 💚📈 Text㈩[number removed]"
Creators of scandal markets on Manifold could set resolution criteria that attempt to mimic the Character and Fitness process for people applying to be lawyers in the US. This process varies by state.
For example, California has these Moral Character criteria: https://www.calbar.ca.gov/Admissions/Moral-Character/Factors-and-Conduct. Will failing/passing those criteria closely match a yes/no on 'a significant number of Scott's friends, readers, and/or professional aquaintances [sic] are angry at him or want to stop associating with him'? Probably not.
By the way, are any of the blogs in these interlinked communities written by lawyers?
umm ok. This thread stuff is going beyond my marginal computer literacy. I understand email, not so much modern media. No idea whether this will respond to your note.
If you want to have "stocks" on a prediction market, there should be dividends associated with them. The economics of that are clearly going to be easier with fake-money, but a system where the top people in terms of "positive buzz" get a return should be possible.
The problem of getting experts to consider the value of prediction markets strikes me as similar to the problem of getting baseball players to use analytics to improve their game. This came about through two mechanisms:
1. Coaches talked to the statisticians, and figured out how to translate the data into something the players could understand and accept.
2. Younger players grew up in an analytics world, and were more accepting of the data from the start.
I suspect right now prediction markets need to identify the coaching layer, the people in the expert community who understand the data, but speak the language of the experts.
Medvedev's joke (which I found rather droll) might be that there were similar predictions about Russia last year.
You forgot to change your percents
"Of my 50% predictions, 10 were right and 0 wrong, for a score of 100%"
In "My 2020 Calibration", the graph is wrong. Your x-axis has 95 inserted between 90 and 100, but the spacing for these gaps of 5 is the same as the spacing for gaps of 10 elsewhere in the graph, making the axis abruptly nonlinear.. Your black line going "straight" from (50, 50) to (100, 100) therefore does not pass through (60, 60) or (70, 70) etc.; in order to do that, it would need to bend sharply at (90, 90) when the x-axis suddenly stretches out.
Also, in the text above the graph, you said "Of my 50%..." over and over, instead of changing the number.
Medvedev seems to be making a bid for the role of the court jester, which was long held by a certain Vladimir Zhirinovsky, who died last year. He'd been decently popular due to his absurd antics (and ultra-nationalist rhetoric), one of the more notable ones was "predicting" in 2002 that elite Iraqi soldiers would rip Bush's army to shreds.
I always found political prediction markets interesting. Suppose you were a generic Republican. Why not ask yourself how much you will need to be ambivalent about Dems winning the next election and put money on them? Either way, you seem to win.
"Stock" status of 0%-100% could work if you conceptualize each person as having good and bad status points. If Elon Musk was sitting at 50% pre-Tesla, I could imagine him post-Tesla with 10x as many good points and 1/2 as many bad points, which by my math would put him at 95%. Someone with no status quanta either way would be undefined, which seems reasonable.
Legality of prediction markets. CFTC won't let a commodities market engage in gaming, but, sports betting has leaked out an been legalized pretty much everywhere in the country as is evidenced by the advertisements on NFL games this season. This all occurred because of a Supreme Court overturning a federal ban on sports betting https://www.espn.com/chalk/story/_/id/23501236/supreme-court-strikes-federal-law-prohibiting-sports-gambling
The states responded by enacting legal frameworks for sports betting with the approval and connivance of the major professional sports leagues.
Given the size of that hole in the bans, I would guess that we will see British style omnibus bookmaking in the US in the next few years. Election bets will be a very important product. Could prediction markets be far behind.
Don't be shocked by the CFTC's behavior. It is pretty typical of the alphabet agencies that I had the displeasure of dealing with when I was active in those fields before retirement.
It is a violation of their charters and of administrative law, but they know that it is extraordinarily expensive and time consuming to hold them to account in the Federal courts.
Have we considered the possibility that the CFTC is secretly run by the Vorin church?
What is the chance the Department of Justice will enforce the national mandatory training of police in appropriate ways of handling criminal suspects?
1. Police can't be exempt from prosecution for crimes. James Bond is fiction; no one has a license to kill.
2. Police can't be exempt from prosecution for violating a criminal suspect's civil rights.
If the feds hold individual departments' feet to the fire, and the training they've provided is truly rigorous, departments should support officers following humane and appropriate treatment of criminal suspects, even when they're sued.
If the officers do not follow their DOJ-directed training and handle criminal suspects humanely, the department should have no responsibility to defend them.
What is the chance?
Request to have the 2023 contest results post out before the 2024 contest deadline - being able to learn from my mistakes in last year's contest would be helpful for the next contest, but as is I have to wait a year for that.
(I understand that writing the results post is probably the kind of surprisingly large amount of work post that's hard to write to a short deadline so this may be hard or impractical)
(My self-review post
https://shakeddown.substack.com/p/2022-prediction-contest-performance?sd=pf )
The part about scandal markets reminds me of Robin Hanson's arguments for the legalization of blackmail, as well as some arguments against such a position. "The median scandal is extremely stupid, but can still ruin someone’s life.", I think this is very much true and not said enough, but I'm not sure about how to solve the general problem, that is even without scandal markets, this and other related issues surrounding false allegations and such, are already such a perverse problem. And I don't think its reasonable to expect the general public to raise their epistemic standards especially on such issues. There is also the cost not just to the person who has been accused, but also to people who haven't been accused modifying their behaviour to reduce the chance of being accused.
I don't think Medvedev's joking, I think he's coming up with pleasant stories of what the future holds because anything more realistic is unpleasant for him to think about. "Of course, the EU will collapse and we will come out victorious". It's just desperation.
A stock could evaluate to the result of a large survey. "Do you like Elon Musk? Y/N." Then in the same way futures expire quarterly, the stock could expire quarterly by fielding the survey once-per-quarter.
Did anyone here use Metaculus Prediction Updates for Ukraine, if not, why not (given that you're interested in the topic at all)? If you used it, how, and why ... and how satisfied are you with what you got?
I've gone through phases of: useful, what's going on?, probably not working at all, it's back?, to name some major ones. My previous exposure to PM was limited, and I'd appreciate your thoughts on this.
I basically woke up with this question this morning (not really, but almost), and planned to post it on the next OT, but that just fits too well. :) Congrats to Metaculus.
I've noticed in manifold that when a market is surely going to resolve YES the percentage gets to something like 90~95% then stops getting better, and then maybe gets to 99+% when it's about to close and resolve. This happens because people have limited mana and would rather put it to use on other markets that would pay better. What I'm saying is very obvious, I realize, but I'm pointing it out as a problem for markets like the one about Coinbase going bankrupt. It could mean 10% is a good estimate, but it could also mean people are not interested in putting their mana in a market that closes in late 2024 for some measly gains. And the closer to the max/min value, the worse it gets.
I kinda dislike the metac.'s "example" - oops it was Scott's ... err, - I still dislike the nuclear strike take here, not that it is wrong, but it's far too short: "a 10% chance of a strike, which comes from a 15% chance if the war in Ukraine continues vs. a 5% chance if it doesn’t. And they think there’s a 50% chance the war will continue, which comes from a 60% chance if the US stops arms shipments and a 33% chance if it doesn’t - and so on." This sounds too much like: we can substantially reduce the chance of a strike if we stop to send arms. - I see the logic, but one might mention the probability that the US/West will be seen as insincere and weak then - with massive rise of risk of worse conflicts in other places, incl. those who might go nuclear + that much bigger chunks Ukraine will be under Russian rule, that decades of unrest in the region will follow, that Putin and his clone (or is it clown) Medvedev will continue their aggressions in other areas of the former CCCP. Even an average-acx-forecaster (as me) might see that "no arms for Ukraine" goes "less peace". - So, not a good example to pick. Sorry, but I am prickly about Ukraine, esp with public support so fickle. Btw: Excellent forecasting by Scott, it was a hard year to forecast.
Even restricted scandal markets seem terrible. "Will John Doe be found to have committed fraud?" strengthens the "John Doe <-> fraud" link in people's minds (including subconsciously), in search engines, and in future AIs' training data. It will also contribute to a "no smoke without fire" effect among normies and indeed anyone who stumbles across John Doe's fraud-scandal market in isolation without knowing the broader context.
> Current law includes the following provision on event contracts, [banning]:
I think this is a slightly bad / misleading summary. The whole point of that section is that the current law does *not* ban these things, but merely that it gives the CFTC the power discretion / authority to ban contracts which involve these.
With the current summary, the follow-up sentence after the list, namely "So the CFTC may ban certain prohibited categories.", makes little sense.
>If you have a better idea for how to run stocks, leave it in the comments here and they’ll probably see it.
How about sentiment analysis using a hard-to-manipulate source like Google News coverage, highly viewed social media, Wikipedia, etc.?
You could define a "sentiment index" equal to, say, log(sum([content.views * content.sentiment for content in pieces_of_content]))
You could also impute an index by randomly polling users on which things they like better than other things, then backsolve numbers which fit with the rank orderings that users give.
Once you've identified a method to create an index for any person/thing/etc. then you'd need a method to bring its price in line with reality while also allowing speculation. Perpetual futures are one approach. Or overlapping quarterly/annual futures.
Or, idea I just had -- make it so every day, there's an 0.1% chance that the asset gets a 24-hour "liquidity event" where people are allowed to buy/sell an unlimited amount of the person's stock, at their current sentiment index (as computed by an algorithm). Of course, people should be able to set up buy/sell orders in advance to take advantage of liquidity events, so they don't need to log into Manifold every day. You could configure the % daily chance of a liquidity event based on the rough timescale that you'd like to be forecasting on.
>When I first looked into this a few weeks ago, a few conspiracy-related prediction markets gave pretty high predictions, because people weren’t sure if the market creators would resolve them honestly, which naturally pushes the price towards 50.
Maybe Manifold should add an option so that when you create a market, you can nominate one or more other users to be the resolvers for that market.
At the beginning of your post you mention hostility to prediction markets from communities that handle classified nuclear information. Perhaps germane to this, a (now deceased) friend was a physicist who worked at a government research establishment that handled highly-classified information. He once told me that the quality of the science done in the unclassified part of the institution was FAR higher than that in the classified part, a phenomenon he attributed to the lack of open critique and review.
What does Scott mean by alpha? He's used it in the Bay Area posts a few times as well, but my attempts to get Google to answer this question just get me links about alpha males, which I don't think is relevant.
The 'What percentage of US electricity will be produced by solar power in 2030' is slightly misleading. It's something that segues into 'What percentage of US energy.....' in many people's minds.
But instead of solar power being four percent of US energy, it is currently nearer one percent.
Solar enthusiasts aren't keen on the distinction, though I think it's important to realise that solar is unlikely to produce more than two percent (of US energy) by 2030. A wild success would be three percent, so its impact will remain trivial, at least for a few decades.
Hi Scott and ACX community!
We at Futuur are fans, and created a category of markets inspired by the ACX 2023 prediction contest:
https://futuur.com/q/tag/astral-codex-ten-2023-prediction-contest
These markets are available in both real money (crypto) and play money (note that forecasts
generated for each are independent).
We'd love to hear any feedback folks might have, as well as suggestions for new markets.
Thanks & happy forecasting!
Hey Scott,
Would you be interested in a chat with Polymarkets founder Shayne Coplan?
"My 2022 Calibration" Now that it is fixed it and understandable, would it be correct to say that you really don't have a good understanding of the difference between being 60% confident and 90% confident in your prediction?
I'd liked to see your "Of my 40% predictions" (and 30%, 20%, etc) results, to see if your blind spot is symmetric.
Also how often was did the event happen for all binary predictions? Is there a bias in how prediction questions resolve - for example, mostly they happen or mostly they don't or happening or not happening is random?
The reality is that "prediction", no matter how well it works is often wildly inappropriate for the task at hand. The fact that we are getting better at it makes it all the more dangerous.
A lot of times, we let our brains think we are predicting something, when what we are supposed to be doing is gathering all the actual evidence about a given possibility, and seeing where it stands. The distinction is huge.
(1) Medvedev on Irish reunification - according to ST: TNG, it'll happen next year:
https://memory-alpha.fandom.com/wiki/Irish_Unification_of_2024
(2) Scandal markets cause scandal - whoever could have guessed? Unless you're a tabloid newspaper, this kind of venture makes little sense.
(3) "Instead, we serve as the sole liquidity provider and counterparty." Well, that cannot possibly go wrong!
https://www.coindesk.com/price/ftx-token/
Idea for stocks: team (maybe with some outside input) gets together once a year to order all the figures in order of status or influence or whatever. Market is on what the index of figure X will be, from (say) 1-10 if there are 10 stock markets.
Scott, for a market resolution on Manifold - could you confirm if this is your old mastodon account? https://schelling.pt/@scottalexander
It seems quite probable that the "unmoored" status stock eventually won't track the status, but only the perceived future probability of the stock. Very vulnerable if majority of Manifold users at some point decides that is its about "lolz", not Musk popularity, or that small values are now good instead of large values, or something equally silly..
Not sure if the cap at 100% is a good thing or a bad thing. On one hand, it doesn't make obvious sense: One can sort of imagine that a person may have a zero status (lowest possible), but can they have a full 100% status?
On the other hand, a hard cap at 100% might have an useful purpose as an inbuilt check against bubble -> crash behavior. Every market participant must ask themselves "is the current growth sustainable" because at some point it must it hit 100%. And if growth is very fast, it will hit it very soon.
On the third hand, if MUSKSTATUS2000 (launched in 2000) is already at 90%, but you still think Musk will become many times more popular when Mars colony succeeds, is there anything stopping you launching another token, MUSKSTATUS2023, initially trading at 2%
Re: policy
This entire area is nonsense.
The original sin here is the notion that policy makers objectively choose and promote policies based on transparent public benefit or some such textbook naïveté.
The real world of policy making involving pushing for agendas in conflict with other agendas.
Nobody involved in policy wants accuracy or objectivity or whatever. Only people outside of policy making want to break this monopoly - much like the crypto banking idiots. And much like the crypto banking idiots who refuse to understand that the speed and function of the present system is a benefit to incumbents - not an obstruction for them - their path to resort will never materialize.
Amazon is already delivering "some products" by drone.
https://arstechnica.com/gadgets/2022/12/amazon-begins-drone-deliveries-in-california-and-texas/
I don't think we blame normies enough, hating someone over a Halloween costume or something they did as a teenager is a massive moral flaw.
You see it most clearly with women in revenge porn cases, the ex-boyfriend who releases images is rightly seen as the villain but the boss who fires a teacher who is the victim of an abusive ex, isn't seen to have blame.
Normie mobs hating on minor infractions means we can't have nice things like prediction markets on scandals and we should be angry at normie mobs.
>Having a scandal market on someone incentivizes people to dig up every scandal they’ve ever been involved in or accused of (including false allegations). Outside Encyclopedia Dramatica and KiwiFarms, most people don’t have a convenient list of every embarrassing and reputation-lowering thing they’ve ever done in their lives in one place.
Do I get any points for predicting the existence of scandal market short sellers 3 months ago?
https://astralcodexten.substack.com/p/mantic-monday-twitter-chaos-edition/comment/10616137
The person who set up Scott's scandal market replied to me and said it would be neat if somebody hired a private detective to stalk people who have scandal markets, but there wasn't enough volume. I feel like we saw entirely different things in this hypothetical!
Re: making stocks more than a Ponzi scheme. It could be tied to some real measurement of the person's "status", e.g., https://en.wikipedia.org/wiki/Q_Score. The stock would have to be a prediction of some standardized measurement guaranteed to be made at regular intervals by a trusted party.
Metaculus stocks as aggregate scores:
Each stock represents the question 'will questions in {this list} pay out?'. have a corresponding list of questions relevant to that person. Ex: Elon Musk would have questions such as 'will humans be on mars by 2035' and 'will the twitter deal wind up being profitable', where a more influencer-type might have questions about their follower count etc. The important thing is to break down why each public figure has their social clout, and judge them on those metrics with lower-scope questions.
you can make stocks 'reset' once per year, or devise a new model that pays off 'dividends' but is persistent.
“some people might genuinely need to hedge political events”
There is enough there to last far beyond the next cycle, especially considering the real question of when a derivative is actually second derivative rather than something else.
Here's an idea for “stock ratings” of people, given that percent ratings seem like they would not work.
What about rating them the way the rating agencies do: AAA . . .D? You can anchor each rating with descriptors appropriate to people rather than stocks: “ Reliable,” “clever”, “a good bet,””likely/unlikely to fuck up” etc. There are various ways to turn this ordinal scale into a numerical one — I’m sure many have played around with ways to do this with stock ratings, so as to make them more amenable to calculations of various kinds. Look and see what all those clever people have done.
Seems to me there are 2 factors that are determinants an individual’s “stock,” as we normally think of it: good/bad and strong/weak. (By the way, those are the 2 main factors that fall out of factor analysis of adjectives describing people.)
-Good/bad= right/wrong, kind/cruel, helpful/destructive, sane/crazy etc.The stock rating measure seems like a reasonable approximation of this.
-Strong/weak= how big a deal someone is, how much influence they have, how well known. Seems like things like Google searches and Twitter mentions are a good measure of this.
So maybe use strong/weak as a multiplier of good/bad measure. Makes pretty good intuitive sense. No matter how good or bad someone’s impact is, their personal stock isn’t valuable if they have very little impact.
Walk into a betting shop in England and you can wager on just about anything sports, horses, elections, what have you.
https://www.oddschecker.com/politics/british-politics/next-uk-general-election/most-seats
I think Medvedev states what he's doing:
> Many come up with futuristic hypotheses, as if competing to single out the wildest, and even the most absurd ones.
I would interpret it as mockery of most of these yearly predictions, which (unlike your competition, or something like the GJP) are seeing attention rather than accuracy.
To the point on Congress having created the CFTC to regulate derivatives contracts: let's not forget that there have been exactly two cases where Congress did, in fact, intervene to ban futures contracts explicitly: onions, and motion picture box office ticket sales (see https://en.wikipedia.org/wiki/Onion_Futures_Act). Clearly the height of well-considered market regulation.
Doesn't Musk doubling move him from 50% (1:1) to 66,(6)% (2:1)?
Is the substack susceptible to spamming? I got this email purporting to be from Astral Codex Ten ( <forum@mg1.substack.com>) regarding Astral Codex Ten replied to a comment on Mantic Monday 1/30/2023.:
"Let's discuss👆I've got something to introduce to you 💚📈 Text㈩[number removed]"
Creators of scandal markets on Manifold could set resolution criteria that attempt to mimic the Character and Fitness process for people applying to be lawyers in the US. This process varies by state.
For example, California has these Moral Character criteria: https://www.calbar.ca.gov/Admissions/Moral-Character/Factors-and-Conduct. Will failing/passing those criteria closely match a yes/no on 'a significant number of Scott's friends, readers, and/or professional aquaintances [sic] are angry at him or want to stop associating with him'? Probably not.
By the way, are any of the blogs in these interlinked communities written by lawyers?
umm ok. This thread stuff is going beyond my marginal computer literacy. I understand email, not so much modern media. No idea whether this will respond to your note.
Most likely Medvedev writes comments for an audience of 1. It's an exercise in predicting what Putin would like to hear.
If you want to have "stocks" on a prediction market, there should be dividends associated with them. The economics of that are clearly going to be easier with fake-money, but a system where the top people in terms of "positive buzz" get a return should be possible.
In other news, I started a prediction market on Biden ending the COVID emergency: https://manifold.markets/AlexPower/will-the-covid-emergency-in-the-us . Once the odds stabilize, I might be interested in real-money action at the betting line.
The problem of getting experts to consider the value of prediction markets strikes me as similar to the problem of getting baseball players to use analytics to improve their game. This came about through two mechanisms:
1. Coaches talked to the statisticians, and figured out how to translate the data into something the players could understand and accept.
2. Younger players grew up in an analytics world, and were more accepting of the data from the start.
I suspect right now prediction markets need to identify the coaching layer, the people in the expert community who understand the data, but speak the language of the experts.
Leaving Robin Hanson’s idea here re status tracking
https://www.overcomingbias.com/2022/12/status-app-concept.html
Thoughts about status stocks.
INTEREST rates are bounded between 0 and 1 for the term of the instrument.
We PAY more attention to (are INTERESTed in) people with higher STATUS.
Short term (~ 2weeks to 11 months) INTEREST instruments are DISCOUNTED.
Long term INTEREST instruments PAY a coupon rate periodically.
The (English) vocabulary of bills, notes, and bonds corresponds more to celebrity status than stocks.