Which CFB team is your NFL team most similar to?
🚀 Dolphins 🚀 Washington
🤠 Chiefs 🤠 Texas
Houston 🤝 Houston
Check out all the teams below… had fun with this one!
With the CFB rule allowing the clock to run after 1st downs, the affect on the number of plays per game is yet to be seen...
Each week I'll update this plot evaluating the number of plays per game through the 2023 season compared to 2022
Here are the results after Week 0:
Looking at breakdowns of CFB team talent by year (Fr, So, etc) and
@ESPN_BillC
SP+ ratings showed Junior year talent as most predictive of success.
Here are the 2023 top 25 most talented rising Junior classes and current preseason SP+ ratings:
Here's the week 1 update!
The median play per game difference dropped from 8 (after week 0) to 5 plays... nowhere near the "40-60" play range that's been floating around
Just wrapped up my first submission to the
#BigDataBowl
. I've used the data in the past and I'm excited to put together a project for this year.
My project aims to quantify yards saved by successful tackles.
Check out the full submission here:
Just wanted to share that my shiny app is live and running at:
This app allows users to see the likelihood of a team choosing to run or pass based on game state and personnel.
Go check it out!
Are NFL combine attendees better athletes than in previous years or are inflated combine results due to selection bias?
Over the last 3 years, players have been more selective on the events they participate in (the ones they’re good at) and produced better scores as a result.
Remember the talk about whether Nick Saban smiling more means anything for Alabama?
I watched every Saban camp press conference since 2017 with a stopwatch, counting smiles to see if they correlate with the year’s success.
Here’s my lab report ⬇️⬇️⬇️
I've been tracking the affect of the CFB clock rule change on plays per game ( ~ 5 play drop from 22 to 23)
Coaches like Chip Kelly and Lane Kiffin seem to have the most complaints, so I wanted to dive a little further into how this rule might affect teams differently...
Week 13 Update:
Movement in the last week!
Median play difference ends at 5 plays. Overall, this rule change didn’t affect the game as much as anticipated. It did cap the upper end of plays per game and reduce the between game variance in 2023. This concludes the thread 🧵
A little late but here’s a plot that explains what happened on the Clemson vs Duke game: Clemson was expected to score 41 points based on their drive effectiveness. They scored 7.
@PFF_Moo
put out a great article on a drive level scoring metric for the NFL, this is something
Awesome insight from
@DataWithBliss
. I looked at something similar, with CFB data (18% success). Using EPA of success/failure I calculated the value of the strategy at each down and distance.
Here's how the strategy's value changes at 7% vs 18% success. No value at 7%...
A common 4th down strategy is to have offense attempt to draw the defense offside.
Here is the breakdown of 4th & ≤5 plays with:
- An offensive timeout and then punt/FG
- A delay of game/false start and then punt/FG
- A defensive offside
Since '16, 7% have led to first downs.
Florida State's offensive struggles since Travis's injury (worse than everyone but Iowa) have to have played a big part in the CFP decision to keep them out
Would've been interesting to see the top 4 if Oregon beat Washington... I imagine they would be in over FSU as well
It's 2024 and we live in the 12 team CFP world now...
Using a simple WP model based on
@ESPN_BillC
SP+ ratings, I ran 10,000 simulations of a 12 team playoff for each season.
Here are the resulting national champion probabilities for the top 12 teams over the last 10 years
Apparently there's a big game in Tuscaloosa this weekend...
Not too much difference between these two stacked rosters ⭐️⭐️⭐️⭐️⭐️
Let me know if there are other matchups you want to see!
CFP Signing Day Class Breakdowns
Clemson, Northwestern, Air Force and Navy with 0 transfers!
Colorado, Louisville, Texas A&M, and South Carolina with some portal heavy classes
Sorry for the small bars… had to squeeze a lot of teams on here
Introducing momentum plots!
These plots look at changes in win probability over the previous 30 plays to highlight game flow
🧵
49ers took control in the 2nd quarter and didn't look back. Super impressive from them
@bburkeESPN
Just trained a random forest model to predict QBR based on a team’s previous QBR, QB’s previous QBR, starting experience, and recruiting ranking. Happy to share code if interested.
NFC/AFC Championship Game momentum plots
Chiefs were able to shut the door on the Ravens every time it looked to be heading their way
The 49ers Lions plot turned out about how I expected…
Ohio State might have two first round WRs on their roster this year with Marvin Harrison Jr.
@MarvHarrisonJr
& Emeka Egbuka
@EgbukaEmeka
. Both had monster seasons in 2022, but here's how they found success differently...
Thanks to Statsbomb for the sample data to use
Been thinking about quantifying offensive style after listening to
@statsowar
talk about rush rate over expected...
I ran PCA on: pbp, season, RROE, and opponent adjusted EPA metrics.
The first two PCs clearly came through performance and style.
Here is P5 from 2022:
After simulating CFP in 4 team and 12 team formats, here were the teams from each of the last 10 seasons with the most to gain/lose by switching to the 12 team format
Winners: 2nd best SEC/Big Ten team who just missed playoffs
Losers: The 1 seed (9/10 years) or title favorite
It's 2024 and we live in the 12 team CFP world now...
Using a simple WP model based on
@ESPN_BillC
SP+ ratings, I ran 10,000 simulations of a 12 team playoff for each season.
Here are the resulting national champion probabilities for the top 12 teams over the last 10 years
The new MLB Statcast bat speed data is fun.
Wanted to see which factors maybe be influencing swing length and bat speed (besides the hitter).
One prominent factor in swing length and swing speed is strikes. Batters are swinging shorter and less hard with strikes in the count.
Team changes in median plays per game from 2022 to 2023
On average, teams are only running 2 fewer plays per game in 2023.
However, it’s very clear that high tempo teams last year (Texas Tech, Ole Miss, Washington, etc.) were affected more by the new clock rule change.
Thought this was great insight from
@benbbaldwin
and wanted to compare to how college coaches are doing.
NCAA coaches may be improving on 4th down decisions but still firmly behind NFL coaches
2023 Top 25 Most Talented CFB Teams Ranked by Junior Class Talent (most predictive of success)
Updated with current 2023 roster information from
@On3sports
@On3Recruits
Which programs are going to benefit the most from the 12 team playoff?
This table highlights programs with the biggest increases in odds to have won at least one title if we'd had a 12 team CFP since 2014.
Talented teams who had consistently fallen just short of 4 seed had the
It's 2024 and we live in the 12 team CFP world now...
Using a simple WP model based on
@ESPN_BillC
SP+ ratings, I ran 10,000 simulations of a 12 team playoff for each season.
Here are the resulting national champion probabilities for the top 12 teams over the last 10 years
Week 9 Update:
Difference still at 4 plays (assuming we will stay here the rest of the season)
One thing standing out to me is the true peak in the distribution around ~126 plays that 2022 doesn’t show.
Hold Out Fantasy Points!
This metric takes out the highest fantasy scoring play each game, penalizing inconsistent scorers whose weekly points mostly come from one big play.
Depend on big plays: Hall, Addison, Davis
Stable usage: Jacobs, Chase, Brown
Players above the line
Seeing a few of these float around, and I tried to fix some of the shortcomings I've seen with other methods
☑️ Measuring performance against the spread to account for general (NCAA avg) home field advantage
☑️Looking only at games with spreads less than 14, assuming peak home
Great opportunity to work in sports data analytics on the performance side! Offers significant experience working across multiple sports with great people on the Sports Science staff.
Which teams are running/passing more than expected?
Notes on a few leaders:
⬆️ Nebraska - Sims with more carries than completions
⬆️ Mississippi State - No more air raid :(
⬇️ Washington - 74% pass rate! (most in P5)
⬇️ Georgia - Throwing a lot to build Beck's confidence?
Wanted to look at a team's RB1 carry rate and rushing efficiency.
Separating teams into above or below avg. RB1 usage, high usage teams (bell cows) are no more efficient than low usage teams (RBBC).
Rushing success has more to do with team situation than the guy with the ball.
Looking at teams' postgame win probabilities in wins & losses reveals some insights on teams whose performance may not have been reflected by their records
Notable (+) Regression Candidates (High Win Prob. in Losses): Illinois, Alabama, Texas, K State, Clemson
Notable (-)
@bballbreakdown
Awesome video!
Worked through some play by play data today to see how Wemby's length may disrupt opposing offenses using ON/OFF shooting splits:
ON / OFF
FG%: 37% / 47%
Avg shot dist: 14.4ft / 11.4ft
Super small sample but these changes seem pretty significant!
🪄THE WEMBY EFFECT🪄
Looking at how Victor Wembanyama's length may disrupt opposing offenses using ON/OFF shooting splits:
ON / OFF
FG%: 37% / 47%
Avg shot dist: 14.4ft / 11.4ft
Limited samples, but a 10% change in Opp. FG% and a 3ft change in avg shot distance are worth noting
Happy 4th of July! 🇺🇸
With the Nathan's Hot Dog Eating Contest going on today, here are predictions for the winning number of hot dogs eaten today (by Joey Chestnut):
Logistic function (1975-present) - 75 dawgs
Yield Loss Function (2007-present) - 73 dawgs
CBB Net Turnover Margin (conference games only)
Been exploring the new cbbdata package by
@andreweatherman
✅ ISU, NMSU, Marquette
🚩Alabama, Creighton, Oklahoma
See the rest of the Torvik
@totally_t_bomb
below 👇
The most impressive thing about Vladimir Guerrero Jr.'s Home Run Derby win last night?
His ability to maintain his strong HR exit velo throughout the competition.
With
@MLB
statcast data, here are his HR exit velos plotted throughout the derby...
#HomeRunDerby
#AllStarGame
Kickoff Stats?!?
The recent change in the NFL kickoff format made me want to investigate teams' kickoff strategy.
The Carlson bros and Nick Folk had a high % of their kicks returned, and had an above average return team starting field position on these kicks (good coverage).
Is (QB EPA - PFF Grade) an OC/play caller metric?
Love & Ridder at 1&2 in very QB friendly systems
Lawrence & Mahomes performing well despite offensive struggles
2022 full season results included for comparison
Updated QB efficiency and grading through Week 2, including MNF
- Pickett and Watson didn't do themselves any favors last night
- Young still has the lowest PFF grade
Does drafting a WR and RB from the same NFL team limit your fantasy team's overall potential?
With many fantasy football drafts going on around this time, I wanted to investigate this question. At surface level, there is seemingly no correlation between points scored by WR1 and
NFL Draft QB Prospect Player Comps
Not as informative from a physical comparison side because QBs don't do the combine tests (sigh)
Joe Milton ➡️ Josh Allen 👀👀👀
Continuing on with some kickoff analysis here's a breakdown of some of the best returners in the NFL
Excited to see guys like Deejay Dallas hopefully get some more recognition with the kickoff changes allowing for many more kickoff returns
If the NCCA decides to cut the transfer portal window in half, which CFB teams may be affected the most going forward?
Using
@cfbfastr
player usage metrics, these schools relied on 1st year offensive transfers (QB, RB, WR, & TE) the most in 2022:
Is the SEC losing it's grip as the best conference in college football?
Here's how each conference has performed against non-conference P5 opponents over the last ~10 years
Biggest plays from some of the top matchups this past weekend (based on cfbfastR win probability model)
🧵🧵🧵
Starting with Haynes King's game winner against Miami - 94% win probability added
Saw
@MrCondeMx
graph on production and volatility. More productive WRs have inherently higher variance (R=0.68). The cube root of weekly fantasy points is more normally distributed, so the variance of the cube root more accurately depicts volatility.
More championship weekend plots
Trying to visualize game flow by looking at win probability change for each drive.
Overlooked the fact that KC punted every drive in the 2nd half. Expecting SF to try to copy paste Macdonald’s game plan in the SB.
Also SF “won” 8 straight
Week 7 Update:
Still at a 4 play difference…
Interesting to note that 19 of the top 20 games in terms of plays per game came in 2022, with the one exception being WKU v USF in week 1 (167 plays).
How different are kickoff returns going to look in the NFL in 2024?
✅ better starting field position
✅ more fireworks
Comparing kickoff return distributions highlights how fun XFL kickoff return rules are. It’ll be awesome to see these rules in action next season.
I used a LSTM model to predict relative carrier to defender distances post-tackle and the frame at which the next hypothetical tackle would occur.
Yards saved is then calculated by taking the difference in original tackle position and the position of the projected next tackle.