♠️ Poker • 📈 Daily Fantasy Sports •💰Player Props • 📕 Author of ‘The Theory of DFS’ at https://teapawt.net/t.co/eXBKzsaWEP • 📺 @RotoGrinders Analyst
Joined November 2009
- Tweets21K
- Following448
- Followers18K
- Likes10.4K
Jordan Cooper retweeted
On the prediction markets vs. models vs. legacy institutions debate (talking about elections, but most of what I'm saying also applies to sports):
- The markets obviously tailed the high-quality models, and Kalshi trying to act like they knew "more than the experts" is embarrassing and dumb. PMs would have done much worse in the lead-up to Nov. 3rd if Nate Silver, Lakshya Jain, etc. didn't publicize their work. And to whatever extent they'd have been close, it would be because of market participants using similar private models.
- That said, on election night / as new information is rapidly coming in for hundreds of races, I found the PMs to be a useful snapshot of what we knew from people incentivized to be correct about the specific race they were putting action on. They weren't always right (NE Senate for the independent got up to 70% at one point, etc.), but as long as you know what a prediction market is actually saying, I found them useful as a spectator, alongside Kornacki on my screen.
- It is true that PMs correctly hit 99% or more for most outcomes 1-2 hours before any legacy media organizations called races, but the incentives here are very different. A 1% error rate for NYT, FOX, NBC, etc. in calling a race's result would be catastrophic and unacceptable. I think it's good that we have institutions whose incentive is to be right rather than fast.
I continue to think that prediction markets have practical uses, but the push (in their marketing and business practices) to get recreational bettors/"dumb money" onto them to provide liquidity to the sharks (at best) or literal insiders (at worst) as much as possible so they can make money is caustic and bad.
Good idea, bad mass consumer product. And an insufferable communication strategy.
$2b in volume *by the end*, sure. But not in liquidity - unless I’m using the term wrong somehow - meaning “how much money can you get down/trade in-out on a position”.
Back in August, ample liquidity wasn’t even available for a vast majority of races. You want Ojeda 20c for NC-09? Well, there’s only $100 waiting for you. You want Hamilton 18c for KS-Sen? Ok, there’s $2k to scoop up.
A bunch of margin of victory markets had to create higher strike prices a week out from Election Day simply because there was zero action for the highest bands previously.
You’re correct that markets amplify smart money. It did. But you still need enough counterparties and liquidity to make it worth the time for larger smart money. Take a look at most graphs and you can even see when things started to converge with sharp models, often within the span of a week. The more liquid, the faster. The less liquid, the slower.
Same happens in sports betting hence why top-down steam chasing is a viable strategy. There’s a reason why books set limits on openers - you can’t just get $100k down on some random overnight MLB pitcher prop.
As money comes in, lines move, limits are raised, and lines move again. Many times this progression can happen in mere minutes. It’s just a matter of how much you can get down within limits before the inefficiency is gone.
$2b volume *sounds* like a lot. Never mind the fact much of it was within a week of the election. But it pales in comparison to other markets. I read (feel free to fact check) that even on Election Day, political markets still only represented 4% of the day’s volume while the bulk was NFL, CBB, NBA (when there weren’t even games occurring day-of).
To put things in perspective, one would regularly have the ability to get more down on a random NFL rushing prop on a Thursday morning than a vast majority of election markets in August when Nate’s FLIPR model was first published.
We’re really not in disagreement here. It’s just a matter of time horizon.
I’m nearly 100% in agreement here. You could have taken FLIPR in August and turned a profit on a predominance of markets. Same for the past several election cycles.
One reason why is that they’re nowhere near as liquid as sports betting markets on a random Thursday night football game. It’s ripe for finding inefficiencies.
Another is that it’s full of dumb money betting on vibes and/or counterparties that simply don’t believe data due to faulty narratives pushed by pollsters & pundits.
I’m not a sophisticated modeler by any stretch of the imagination. I barely get by in Excel. But just as I’ve done in DFS, I’ve been profitable the past 5 elections straight mostly from analyzing sources (and specific underlying data) that are way more predictive in certain areas than the betting public and sometimes even the forecaster him/herself believes.
As Nate has tracked, the PM prices eventually converged in agreement with what most sharp models already forecasted in their data. It was just a matter of getting down (and in the most inefficient markets) before the correction.
My only disagreement comes in the approach most forecasters take in extrapolating & utilizing their data. They’re in the political industry, so their pursuit is to appease a political audience, academically and/or commercially - as opposed to a betting audience seeking alpha more similar to the finance & investment industry.
Even so, Nate is one of the few who actually communicates more so like a bettor and it’s a big reason why he has often gotten so much shit the past decade from political audiences about his model and the analysis he derives from it.
The explanation that @lxeagle17 gives here in his election forecasting methodology is probably the best I’ve ever heard - and the most fruitful approach in my opinion.
The equivalent transformation in the DFS space was with contest sims.
youtu.be/bOIzrJa25eE?t=2333&…
On last night’s @PredictQmkt show, I said that I’d trust political modeling & analysis from those that profitably do so in DFS, sports betting, etc. over news outlets or academics.
But it’s not necessarily because “they actually put their money where their mouth is” so their opinions count more. It’s due to fundamentally different goals set from the very start.
Academics are in pursuit of the truth. Bettors (sharp ones at least) are in pursuit of edge.
This is the stark difference between sharp bettors and sharp political data analysts.
The latter doesn’t have counterparties. They’re approaching their work in a vacuum, which fundamentally shifts their focus to “how accurate was my model?” rather than “what mistakes is the market making?”
It’s often been my bugbear in DFS with GPP ownership projections when at times sites will promote their model’s r-squared.
It doesn’t matter. I don’t care about the overall accuracy. Whether you got a hundred 1% owned players right doesn’t affect me. Getting the most owned players right does.
For instance, having the split between Gibbs & Bijan be 38/14 and then when they come in at 22/24 (with CMC going from 6 to 12) - this matters heavily. Nailing that Pop Douglas was 3% and Brenton Strange was 5% really doesn’t much.
To me, this is treating DFS ownership projections as an academic pursuit rather than providing useful extrapolations of data to those seeking edge against counterparties in a market.
Fortunately, most sharp DFS players and analysts lean heavily toward the latter and therefore it’s why I’d trust their approach to betting on political markets over most others.
The races are so over that by 4am I wonder if Kornacki starts analyzing his DoorDash options on the big board just to pass the time.
Still can’t believe it’s before 11pm and I’m considering going to sleep. It’s probably the greatest election night of my lifetime besides 2008.
Live look at the Republicans getting ready for their commute to TV news spots tomorrow morning. Remember, only two aides each can fit in the boat!
I’m going LIVE now on the @PredictQmkt midterm election stream with @JustinHerzig @jbresl @the_dean221 following all the prediction markets and giving our analysis as votes are counted.
Watch here: youtube.com/live/PyPBcOMBm5Q…
“Dems are flooding money into PMs only to give the perception to others that they’re winning” - right wing grifters who convinced their audience to buy $1 Trump coins by the boatload for three bucks a piece.
So much fake shit on the fucking timeline today. It’s making the FYP almost useless.
I don’t understand the purpose of posting demonstrably false/misleading info & analysis that will ultimately disappoint any followers that ever thought the account was worth a fuck.
