The Devil's Advocate https://teapawt.net/t.co/x4MfBkQWTG
Naples, FL
Joined June 2009
- Tweets8.9K
- Following768
- Followers643
- Likes53.9K
YouGotHozed retweeted
You don't need to crash a plane to learn to fly ... and you don't need VC to build a business.
Everyone who learned the word EPA in 2021 wants the coach to go for 2 down 14. I pulled every play since 2016 to see what it's actually worth.
169 times a team was down 14 in the 4th and scored a TD. 124 kicked, 45 went for 2.
Went for 2: second TD 8 of 45 times, then 4-4.
Kicked: second TD 35 of 124 times, then 16-16-3.
Coin flip either way. And in 77% of these games the team never scored again, so the conversion choice was theater.
The 2-pt rate itself isn't the problem, 23/45 is right on the league number. The problem is the lever. The paper gain is 12.5 points of win probability, but only once you get a second TD, and you get that second TD 1 in 4 times. Net: 2-3% of games.
A win every couple of years, maybe. You'd need thousands of games to prove it, and the error on your own team's 2-pt rate is bigger than the edge.
I bet numbers, not narratives. This one is a narrative. It's not an edge, it doesn't price into a live line, and a coach who kicks isn't costing you a single unit over an entire career. If your model is leaning on it, your model is leaning on noise.
2016-2026, 169 games, nflverse pbp, run it yourself.
The NFL analytics models have no conditional inputs to the actual situation. They are all league averages over a decade plus of sample size.
Every team, every game, every situation is not league average, which is why conditional and in the moment inputs matter, which the model is NOT quantifying.
Therefore, the models have enormous model error, and are flat out wrong.
If the inputs are flawed the outputs are wrong. Period.
My NFL + NCAAF models at edgefndr.com are +116.3u on 539 plays this season.
Every one posted before kickoff with stake and book, graded as posted. The full ledger is on the site.
@EdgeFndr
Stop guessing, start finding an Edge.
The NFL analytics models have no conditional inputs to the actual situation. They are all league averages over a decade plus of sample size.
Every team, every game, every situation is not league average, which is why conditional and in the moment inputs matter, which the model is NOT quantifying.
Therefore, the models have enormous model error, and are flat out wrong.
If the inputs are flawed the outputs are wrong. Period.
Guys my models are +106u on the year:
edgefndr.com
If you want to stop guessing and start finding an Edge, every pick, every book, posted days before kickoff with expected edge, sizing, and graded accordingly.
Fade or follow up to you.
edgefndr.com
+106.18u on the year NFL + NCAAF
Stop guessing, start finding an Edge.
"The analytics say going for 2 down 14 is +EV in all situations"
Meanwhile the models are only taking into account decades long league averages and nothing situational or conditional in the moment.
If the goal of going for 2 when down 14 is it increases your win % - why would you not only go for 2 on the second TD when you have a chance to win or tie?
Going for 2 on the first TD is abhorrently bad. 50% of the time (roughly league average) you are now playing a 50/50 JUST TO TIE THE GAME rather than win.
Add in model error and these are not 50/50s in the moment team by team, game by game... it is a massive massive blunder to go for 2 down 14. Kick the extra pt and you go for it TO WIN on the 2nd TD if the coach wants to get aggressive.
Really every team should be playing for OT. They are raising their risk of ruin (in game) every time they go for 2 down 14 unless they are wildly above league average on 2pt conversions, which only about 4 teams are.
edgefndr.com +103u on the year if you guys want to stop guessing and start finding an edge
The models are just assuming extremely favorable and tight conditions to show a +EV win % when going for 2, when in reality these inputs are not even remotely correct.
The models that say go for 2 down 14 have error and are incorrect.
The models that say go for 2 down 14 are assuming 50/50 win share in Overtime.
Just start and stop right there. That's a wildly optimistic assumption and creates error in everything downstream of it.
So many fake sharps expose themselves over and over and over again when it comes to NFL "analytics"
You guys REALLY need to take a statistics course before you speak publicly
Cant understand conditional inputs or sample size so goes straight to blocking. LOL standard. Clowns dont like to be exposed.
Maybe there's a reason why so many fake sharps disagree with my takes yet my models are +103u on the year and all they have is their opinions.
No. Depends on efficiencies and frequencies this is not just cold league averages which is the only thing the models are accounting for. There is massive model error and conditional inputs that are not accounted for.
This is an extremely fake sharp take to just blanket go for 2 down 14.
I know this owner personally and walked the property many times. They're very wealthy and this home among others is just a fun project for them. They have no need or real intent to sell but if someone is willing to pay their price they will.
I remember last time I was there she left me with:
"The price is going up $5m per year cause we dont care to sell"
$LPTh goes from saying they wont disclose any deals under $10m to putting out a BS presser for 50k units of optical assemblies.
Why? Cause they dont have any contracts over $10m and their g5 qual is still not nearly finished if it ever will be so they have to find something to try to pump.
This company is deeply, deeply unserious. It will be going lower.
My NFL model is +70.8u on the season and +21.99u this week with Seattle ML live..
and I still think the NFL is manipulating outcomes with this atrocious officiating.
@EdgeFndr
