Nate I want to echo your claim people read your work for your voice. When FiveThirtyEight had fewer Silver articles — and a metric ton of content I can’t imagine you’d ever approve — I lost all interest in the site. I stopped visiting it long before you left because it seemed like a totally different site. Also because everything was a fucking video or “podcast.”
I’ll read whatever you put out. I don’t even like poker but I read those too.
Seconded. I very slowly stopped reflexively typing fiverhirtyeight.com as fewer and fewer articles were Nate, and it’s been a breath of fresh air to have SB, with (almost) purely Nate’s perspective and writing style.
Nate—I think I’m a founding member. I’m here for what you think is interesting, and I simply don’t read the articles in which I have little interest. For instance, in terms of your election modeling, I just want to know the time, I don’t need to know how you are building (and upgrading) the watch. I love the March Madness model (I leveraged it to win my office pool) and would love anything that will give me an edge in my family’s Fantasy Football league (my younger sister annoyingly keeps winning). I like and appreciate your quirkiness: I read or skim most of the posts, I bought the e-book version of your new book, and I listen to Risky Business (when you and Maria have time to record). Thanks for the radical transparency and I look forward to an interesting fall.
I completely agree with this. I’m one of the small sub-group of readers that isn’t a gambler, although my profession involves a high degree of taking calculated risks.
But I’m probably also a founder and I came here to read Nate’s wild, random thoughts. One of my favorite things in life is when an author makes me suddenly realize there’s a whole new way to think about something. Nate has done this several times over the years. If he puts himself in a box, I think I’m less likely to get the same experience.
So please be random Nate! Post about whatever you want. That’s when things get most interesting for me.
Same. While my masters in Operations Research does touch on probability and forecasting, it's a pretty minor part of it. Personally I don't enjoy gambling and never have.
Same, I hate gambling, but have a slightly libertarian lean toward what people (like Nate) do for fun. Why am I then a subscriber? I like sports, I follow politics relatively intensely, and as a researcher and teacher, I actually love statistics! :) Would love to see the percentage of subscribers who gamble.
I actually don’t care for the basketball ones at all, but Nate’s general approach to breaking down, analyzing and articulating the data trends is always refreshing.
As a data professional, it’s something that’s very difficult to find and cultivate in others and Nate is generally one of my first examples of best-in-class.
I’ve loved the new book so far and there are some sections of signal and the noise that I quote and recommend often. The housing market chapter sticks out as one for all my twenty-something peers who think house prices (or stonks 😝) only go up.
Nate, my question is about book/author economics. You mentioned that your publisher “probably cares the most about print hardcover sales” which got me thinking about broader book economics. Is it better for you, the author, if I buy a hardcover print version than an ebook? Than an audiobook? Is it beneficial at all to you if I get your book at the public library? Would it hurt your chances for a next book (with a publisher) if the audiobook did better than the hardcover? In short, what format is most important to you as the author and what are the dynamics of the importance of all the different formats?
It's probably outmoded, but the NYT hardcover list and combined ebook/hardcover list just have a lot of prestige in the industry. But please order whichever version you like!
Or some other somewhat esoteric or niche sport. Rugby, curling, dinghy sailing (I’d suggest ILCA or sunfish), or if you get the NHL model redone you can add one for the PWHL’s second season.
I may or may not be in the majority, but I'm mostly here for the political modeling. Initially, I didn't plan to sub very far past the election. However, a lot of the "tougher call" politics stuff here would pull me in. I'm a bit of a politics junkie, so the more coverage you gave it, the more likely you are to get my dollars.
+1 from me for making sure the newsletter doesn't lose your voice in the clutter of a huge expanded stable of contributors. I'm completely ignorant about the economics of scaling up an independent media business, so I can't speak to the trade-offs, but as a reader I think you need to be very careful about diluting your centrality here. Of course your models have excellent proprietary value, but I'm not convinced the value isn't directly tied to and/or subservient to your write-ups of them. Your actual underlying biggest comparative advantage in the market might be your writing style/skill/voice. You aren't boring or hard to read.
My experience with previous iterations of 538 was that it was totally hit or miss with the other contributors. I can't even really explain why that was true. One answer is that you are just plain more fun to read. Another is that I liked your beat/lane better than theirs. But that wasn't strictly it. If I came to the Times site looking for your content and got something from, say, Micah or a freelance guest, it just felt off in a weird way, even if the content was really interesting. Like, it could be an article that I would have been perfectly happy reading if I happened upon it somewhere else, but it just didn't fit into my expectation of what I'd get from you. A mismatch for the mood/reason/whatever for why I happened to show up. And when you moved from the Times to Disney that feeling multiplied. Some stuff felt too slate-pitchy; some stuff felt too newsy; later-on some stuff felt too Vox-y. But a lot of it just felt...off.
I also think---again, admittedly ignorant of the economics side here---that Substack is much more suited to blogger-style engagement, where voice and quirkiness are part of the value. I get the hesitation to do too many "WHERE THE'S BEST FOOD?!?!?!" and "WHAT IS UPSTATE NY?!??!?!" posts, but building a mostly-genuine persona to your writing is definitely a feature, not a bug. Calibrating that is obviously hard, but it's almost impossible for freelance voices or occasional contributors. And I say that as an occasional contributor, LOL.
H-E-double-hockey-sticks yeah for more NHL (or cawlidge hawkey!) content.
I'd be interested in your thoughts on using "AI" in modeling. Rather, using machine learning to come up with models and make predictions through brute force.
For example, there are quite a lot of statistics about basketball. Almost 3000 games per year, each player having all these stats. At some point (if not already) the upper River academic folks will find the time to aim the "machine learning AI laser" at a huge NBA dataset, and see if it can come up with a predictive model.
We don't really know how ChatGPT works so well (as you always say), and if you fed the video feed of like every NBA game since 2000 in, do you think it could figure out some sort of model output that would be accurate?
Do you worry that doing the types of modeling you are doing may eventually be outsourced to AI? Why pay to find an Eli, or hire a Nate to come up with a model when I can just blast my GPUs at the problem for a long time?
Enjoyed this piece, and am really liking your book, but I was surprised that you could write so many words about your career but not mention the great FiveThirtyEight Burrito tourney. And maybe tease an upcoming mini-version for pastrami. (I doubt that even you would venture into deciding America's Best Pizza.)
I think you’ve mis-categorized HCR! She writes her Letters from an American every night (usually quite late) and only takes Saturdays off, which is just a recent shift from her 7-day-a-week production.
I'm possibly being to HCR, who is sort of the GOAT is Substack, but I'm looking at things like the fact that her newsletters are nearly always free (and yes, I'm a paid subscriber anyway) which is almost certainly not profit-maximizing.
On that, I agree 100%. As someone with only a light understanding of the economics of writing, I have been at least hoping her book sales have compensated. Traditional? Yes. On brand? Also yes.
The blue cheese funky Nate Silver Flavor is the main draw for me. 💚
I hope you continue to post about what’s piquing your interest, whether it’s data science, the political horse race, sports, AI, or whatever. The more unfiltered the better. I’m happy to “endure” copy errors, weird tangents, and Twitter screenshots - it’s simply more fun when you are less “edited” or otherwise constrained.
I am super excited that you recorded the audiobook yourself! I’ll reference the ebook for the charts and the parts that are easier to sight-read, and enjoy listening to the rest.
One thing I miss about the (old) 538 days is your banter with Harry. It’s interesting to hear you in conversation with other deep and/or quirky thinkers. The Ezra Klein interview for the book is a great example of that. Maybe having guests on the podcast could be a venue for such conversations, or even publishing chats in Substack between you and whoever.
Great post as usual, Nate. I have a slightly philosophical question that’s come up at my work. Do you have any thoughts on developing in-house forecasting models vs. purchasing vendor solutions? I’m at 10-25K employee B2B company. I’m just launching a model that prioritizes which accounts to target first and how. Leadership had done an evaluation last year at the start of project, but we’re getting renewed pressure to consider a vendor that’s unsurprisingly implemented and often promoted by many marketing teams within our PE firms secretive portfolio and it’s tentacles (don’t get me started on that). General pros and cons are somewhat obvious like recurring costs, integration and maintenance issues, and ability to customize and scale. Welcome any thoughts you or readers have.
Not Nate (obviously), but I do have an opinion (based on considerable experience in this space). This is a really tough question. I currently lean on “build in house,” but having built in house, built as consultant, and purchased models from consultants, all options have their pros and cons.
Internal models are great because the data is all at-hand, and it’s usually cheaper per-hour-of-work, but you need “permanent” FTEs that need something to do when they’re not modeling, and it can be hard to get leadership to buy into weird side projects by some nerd in Department X.
OTOH, you (possibly) only pay for external models once, and it can be easier to get leadership (and everyone else) to pay attention when you pay real money, but search/contracting/data sharing costs are real, and it’s totally possible (even likely?) that you’ll get sold a bill-of-goods by a slick talking sales person who doesn’t really know much about what they’re selling.
Who knows though. Hope this is (even a little) helpful!
Oh, and imho, if you’re in the market for models like this, regardless of if it’s internal or external, the quality of the modeling is often inversely proportional to how often the modeler (or sales person) uses the term “AI.”
Nate, write about the Rangers all that you want, but if I see a single post about the fucking Blackhawks, I swear to God…
Yes, I would enjoy a hockey discussion even if it was Rangers focused.
Would love to see a return of soccer/European football coverage but understand the priorities list.
All & all, thanks for this current content & look forward to how it evolves ... or doesn't.
The fact that Kyle Davidson seems like such a likeable guy kills me.
Yes. Hope Davidson eventually falls into the nice guy/smart guy matrix container. <sigh>.
Nate I want to echo your claim people read your work for your voice. When FiveThirtyEight had fewer Silver articles — and a metric ton of content I can’t imagine you’d ever approve — I lost all interest in the site. I stopped visiting it long before you left because it seemed like a totally different site. Also because everything was a fucking video or “podcast.”
I’ll read whatever you put out. I don’t even like poker but I read those too.
Seconded. I very slowly stopped reflexively typing fiverhirtyeight.com as fewer and fewer articles were Nate, and it’s been a breath of fresh air to have SB, with (almost) purely Nate’s perspective and writing style.
Nate—I think I’m a founding member. I’m here for what you think is interesting, and I simply don’t read the articles in which I have little interest. For instance, in terms of your election modeling, I just want to know the time, I don’t need to know how you are building (and upgrading) the watch. I love the March Madness model (I leveraged it to win my office pool) and would love anything that will give me an edge in my family’s Fantasy Football league (my younger sister annoyingly keeps winning). I like and appreciate your quirkiness: I read or skim most of the posts, I bought the e-book version of your new book, and I listen to Risky Business (when you and Maria have time to record). Thanks for the radical transparency and I look forward to an interesting fall.
I completely agree with this. I’m one of the small sub-group of readers that isn’t a gambler, although my profession involves a high degree of taking calculated risks.
But I’m probably also a founder and I came here to read Nate’s wild, random thoughts. One of my favorite things in life is when an author makes me suddenly realize there’s a whole new way to think about something. Nate has done this several times over the years. If he puts himself in a box, I think I’m less likely to get the same experience.
So please be random Nate! Post about whatever you want. That’s when things get most interesting for me.
I would be curious to know what fraction of the SB readership gambles. I don't either, never have.
Same. While my masters in Operations Research does touch on probability and forecasting, it's a pretty minor part of it. Personally I don't enjoy gambling and never have.
Same, I hate gambling, but have a slightly libertarian lean toward what people (like Nate) do for fun. Why am I then a subscriber? I like sports, I follow politics relatively intensely, and as a researcher and teacher, I actually love statistics! :) Would love to see the percentage of subscribers who gamble.
Never gamble.
Agree with you both here!
I actually don’t care for the basketball ones at all, but Nate’s general approach to breaking down, analyzing and articulating the data trends is always refreshing.
As a data professional, it’s something that’s very difficult to find and cultivate in others and Nate is generally one of my first examples of best-in-class.
I’ve loved the new book so far and there are some sections of signal and the noise that I quote and recommend often. The housing market chapter sticks out as one for all my twenty-something peers who think house prices (or stonks 😝) only go up.
Nate, my question is about book/author economics. You mentioned that your publisher “probably cares the most about print hardcover sales” which got me thinking about broader book economics. Is it better for you, the author, if I buy a hardcover print version than an ebook? Than an audiobook? Is it beneficial at all to you if I get your book at the public library? Would it hurt your chances for a next book (with a publisher) if the audiobook did better than the hardcover? In short, what format is most important to you as the author and what are the dynamics of the importance of all the different formats?
It's probably outmoded, but the NYT hardcover list and combined ebook/hardcover list just have a lot of prestige in the industry. But please order whichever version you like!
I think you should have a model for cricket that you only code while slightly drunk, that could be the occasional light entertainment newsletter.
There actually would probably be a great deal of interest in the IPL (Indian Premier League), it's a pretty massive market.
Or some other somewhat esoteric or niche sport. Rugby, curling, dinghy sailing (I’d suggest ILCA or sunfish), or if you get the NHL model redone you can add one for the PWHL’s second season.
Disc golf!
I may or may not be in the majority, but I'm mostly here for the political modeling. Initially, I didn't plan to sub very far past the election. However, a lot of the "tougher call" politics stuff here would pull me in. I'm a bit of a politics junkie, so the more coverage you gave it, the more likely you are to get my dollars.
How do you invest your money?
Great stuff, Nate! I love the Substack model. I encourage you to do whatever the hell you want to do. No need to try and please everybody.
+1 from me for making sure the newsletter doesn't lose your voice in the clutter of a huge expanded stable of contributors. I'm completely ignorant about the economics of scaling up an independent media business, so I can't speak to the trade-offs, but as a reader I think you need to be very careful about diluting your centrality here. Of course your models have excellent proprietary value, but I'm not convinced the value isn't directly tied to and/or subservient to your write-ups of them. Your actual underlying biggest comparative advantage in the market might be your writing style/skill/voice. You aren't boring or hard to read.
My experience with previous iterations of 538 was that it was totally hit or miss with the other contributors. I can't even really explain why that was true. One answer is that you are just plain more fun to read. Another is that I liked your beat/lane better than theirs. But that wasn't strictly it. If I came to the Times site looking for your content and got something from, say, Micah or a freelance guest, it just felt off in a weird way, even if the content was really interesting. Like, it could be an article that I would have been perfectly happy reading if I happened upon it somewhere else, but it just didn't fit into my expectation of what I'd get from you. A mismatch for the mood/reason/whatever for why I happened to show up. And when you moved from the Times to Disney that feeling multiplied. Some stuff felt too slate-pitchy; some stuff felt too newsy; later-on some stuff felt too Vox-y. But a lot of it just felt...off.
I also think---again, admittedly ignorant of the economics side here---that Substack is much more suited to blogger-style engagement, where voice and quirkiness are part of the value. I get the hesitation to do too many "WHERE THE'S BEST FOOD?!?!?!" and "WHAT IS UPSTATE NY?!??!?!" posts, but building a mostly-genuine persona to your writing is definitely a feature, not a bug. Calibrating that is obviously hard, but it's almost impossible for freelance voices or occasional contributors. And I say that as an occasional contributor, LOL.
H-E-double-hockey-sticks yeah for more NHL (or cawlidge hawkey!) content.
This article is why I read you, wherever you are. You are a good example of the advice I give kids, “Just be yourself”. Thanks, Poblano.
I'd be interested in your thoughts on using "AI" in modeling. Rather, using machine learning to come up with models and make predictions through brute force.
For example, there are quite a lot of statistics about basketball. Almost 3000 games per year, each player having all these stats. At some point (if not already) the upper River academic folks will find the time to aim the "machine learning AI laser" at a huge NBA dataset, and see if it can come up with a predictive model.
We don't really know how ChatGPT works so well (as you always say), and if you fed the video feed of like every NBA game since 2000 in, do you think it could figure out some sort of model output that would be accurate?
Do you worry that doing the types of modeling you are doing may eventually be outsourced to AI? Why pay to find an Eli, or hire a Nate to come up with a model when I can just blast my GPUs at the problem for a long time?
Enjoyed this piece, and am really liking your book, but I was surprised that you could write so many words about your career but not mention the great FiveThirtyEight Burrito tourney. And maybe tease an upcoming mini-version for pastrami. (I doubt that even you would venture into deciding America's Best Pizza.)
I think you’ve mis-categorized HCR! She writes her Letters from an American every night (usually quite late) and only takes Saturdays off, which is just a recent shift from her 7-day-a-week production.
I'm possibly being to HCR, who is sort of the GOAT is Substack, but I'm looking at things like the fact that her newsletters are nearly always free (and yes, I'm a paid subscriber anyway) which is almost certainly not profit-maximizing.
On that, I agree 100%. As someone with only a light understanding of the economics of writing, I have been at least hoping her book sales have compensated. Traditional? Yes. On brand? Also yes.
The blue cheese funky Nate Silver Flavor is the main draw for me. 💚
I hope you continue to post about what’s piquing your interest, whether it’s data science, the political horse race, sports, AI, or whatever. The more unfiltered the better. I’m happy to “endure” copy errors, weird tangents, and Twitter screenshots - it’s simply more fun when you are less “edited” or otherwise constrained.
I am super excited that you recorded the audiobook yourself! I’ll reference the ebook for the charts and the parts that are easier to sight-read, and enjoy listening to the rest.
One thing I miss about the (old) 538 days is your banter with Harry. It’s interesting to hear you in conversation with other deep and/or quirky thinkers. The Ezra Klein interview for the book is a great example of that. Maybe having guests on the podcast could be a venue for such conversations, or even publishing chats in Substack between you and whoever.
Glad this stack is a success. 🙂
Great post as usual, Nate. I have a slightly philosophical question that’s come up at my work. Do you have any thoughts on developing in-house forecasting models vs. purchasing vendor solutions? I’m at 10-25K employee B2B company. I’m just launching a model that prioritizes which accounts to target first and how. Leadership had done an evaluation last year at the start of project, but we’re getting renewed pressure to consider a vendor that’s unsurprisingly implemented and often promoted by many marketing teams within our PE firms secretive portfolio and it’s tentacles (don’t get me started on that). General pros and cons are somewhat obvious like recurring costs, integration and maintenance issues, and ability to customize and scale. Welcome any thoughts you or readers have.
Not Nate (obviously), but I do have an opinion (based on considerable experience in this space). This is a really tough question. I currently lean on “build in house,” but having built in house, built as consultant, and purchased models from consultants, all options have their pros and cons.
Internal models are great because the data is all at-hand, and it’s usually cheaper per-hour-of-work, but you need “permanent” FTEs that need something to do when they’re not modeling, and it can be hard to get leadership to buy into weird side projects by some nerd in Department X.
OTOH, you (possibly) only pay for external models once, and it can be easier to get leadership (and everyone else) to pay attention when you pay real money, but search/contracting/data sharing costs are real, and it’s totally possible (even likely?) that you’ll get sold a bill-of-goods by a slick talking sales person who doesn’t really know much about what they’re selling.
Who knows though. Hope this is (even a little) helpful!
Oh, and imho, if you’re in the market for models like this, regardless of if it’s internal or external, the quality of the modeling is often inversely proportional to how often the modeler (or sales person) uses the term “AI.”
Ha! I completely agree with the last part. Thanks for your thoughts and insights, Ian!
Thanks for the post and for linking to some older ones (I enjoyed your framing of how to evaluate food cities).
Just chiming in to say that I’d welcome your economic index model on an ongoing basis.