A Fool's Library of Alexandria: Serious Answers to Unserious Questions

The history and legacy of rule changes and game mechanics through an unconventional lens.

What's a doink, how'd we get here, and who are you?

A "doink" isn’t a statistic. It has no league definition, no official API endpoint, and no named column in a standard box score. For fifty years, these hits have lived in the cracks of football history, recorded and digitized really only when the commentators, coaches, and culture immortalized the event. Truthfully, caroms and ricochets were probably even more common 50 years ago than what they are today.

If you haven't read about the history of the doink or the story of how modern goalposts came to be, you should stop reading and check out Genealogy of American Football: Evolution of goal posts by Barry Shuck (an actual professional football historian). It's a fascinating walkthrough of the history of the upright, and I cannot stress enough the importance of the work done to preserve the legacy and primary source documents of the NFL. Mr. Shuck is one of many of the unsung heroes of a storied sport full of stars and legends.

That's as good of a segue as I can hope for to share both a disclaimer and a spoiler:
I am a passionate data nerd and avowed lover of the NFL, but the dramatic, absurd, and often meaningless stories that I want to share with you deserve a better voice than I can give them. There are doinks that define careers, lucky breaks that fell dynasties, and untold tales of what ifs and what might have beens.

I have never been a kicker, I am not a professional sports historian or journalist, nor have I ever even coached football. I am a lifelong Bengals fan, and I am a database professional and pretend project manager (I play one on TV Microsoft Teams). This is my love letter to the beautiful, stupid, chaos of the NFL, and my contribution to help catalog the absurd stories buried in the footnotes and play-by-play. In that way, doinks aren't the story - they're a lens to find the story.

Do made doinks matter?

Well, yes. Statisticians much smarter than me have calculated expected points added (EPA) for extra point and field goal attempts. Even more directly, made doinks are scoring plays. But this project started with that simple question, an email to a podcast, and a list of nine kicks that struck the uprights and still went in. The original hypothesis of this project at its inception in late 2025 was a half-baked take, and it still lives on the site today:

Making a doink has a small but significant psychological impact on the kicking team.

I haven't done the math to see if the current made doink win rate of 61% is solely the product of correlation with EPA, but the goalposts here are shifting; the NFL's literally did in 1974. Along the way, I've found more interesting answers to new questions:

"Has a non-kicker ever hit the upright?"
Yes (Ndamukong Suh).

Also, yes (Jeff Heath with the make).

"What kicker has the most career doinks?"
Mason Crosby**.

**We'll come back to Mason later.

"If you could design a play to lead to a doink, what would it look like?"
This.

The Visual Archeology of YouTube Compilations

If you do a web search for doinked field goals, YouTube is sure to be among the first results. This makes perfect sense - the clang of a Seabass kick against the upright or the high-profile dismay of Parkey's problems at the post have always been appointment television.

Besides, in a day and age of clickbait, doinked kicks are an oasis of consistency - they are as advertised in every way.

I started with compilations of Satisfying Doinks and Year End Doink compilations from a handful of helpful YouTube channels (Baseballer24, Savage224 Gaming, and Savage Brick Sports). Their compilations were rich sources of doinked kicks, and being doink compilations, validation of each occurrence was easy, but finding the right game was not.

Early on, I was a digital scavenger, reverse-engineering games from the breadcrumbs available: broadcast scorebugs. Sometimes, a specific font or a clock layout was the only way to differentiate a random week 7 2015 miss from the same matchup in the same stadium a year or two later. Video-by-video, play-by-play, in the blink of an eye, 9 turned to 43 turned to 90 turned to 430. During this process, I built out a proper references framework to further "show my work" (as in, I'm not an expert, but you don't have to trust me - you can check the process) and to provide proper attributions for the YouTubers who helped fill part of the role of new age historians.

But these wells eventually ran dry, and I had exhausted the "easy" archives (author's note - these were not easy, nor was this effective. There's a bit of this process I look at in hindsight like watching someone run up the down escalator).

From Play-by-Play to Gamebooks + Dope Sheets

To find the doinks that didn’t make the highlights, I went to the primary sources: play-by-play reports and official NFL gamebooks. Initially, this was manually trawling through pages of search results and hundreds of PDFs looking for specific flavor text like "hit crossbar" or "off left upright." If you're thinking for the second time in as many paragraphs that there was probably a better way to accomplish this, you'd be right.

This push yielded around 210 more entries here, but with more time and energy for each doink, the effort turned to a slog: Since I was scanning play-by-play to find doinks, then sourcing video and validating game specifics, the effort for every new entry was increasing exponentially. I still found interesting nuggets and the dataset continued to evolve, but the process was taxing.

During a late-night review session, human error crept in. In my excitement to find new doinks, I incorrectly logged a kick and duplicated another. Different players, similar context, wrong conclusion.

The short of it - I couldn't outwork (my own) deficient process.

Process over Effort

My original plan was to build a Wikipedia of doinks, democratizing the data and the process. Thankfully, this intention translated well for the human-side (read - fallible) of this project. I figured if there was one thing I was equipped to bring to this project, it's a healthy level of pedantry and pragmatism to curate this weird dataset and grow it into a public resource.

When a community member identified the mistakes, I took the opportunity to build out an errata framework to help enforce data standards and preserve the integrity - if I can't always be perfect, I can always be honest. It's the same audit-trail pattern you'd see in any serious dataset, applied to something monumentally unserious.

nflfastR + the Parable of Working Smarter

Every single milestone and record tracked in OffTheUpright's database builds from the work of dedicated historians, both formal and informal. This is especially true today: Thanks to the incredibly rigorous and thorough efforts of the maintainers at nflverse, play-by-play data is readily searchable and makes it easier to quickly find likely doinks, both made and missed. This data, plus continued tooling built into my data entry processes reduced errors and helped finish building the database of verified doinks in the 33 yard extra point era. Instead of trying to find games that match the profile of drives and plays, I could focus on finding video evidence of each play from the known game.

The process is still ever evolving, and the challenges in data collection are the story before 1999.

Even so, completing more of the dataset allowed me to re-shape my analysis; it unlocked era-comparison and a more contextual picture of the data.

Current Status: The Verification Backlog

I am currently managing a verification debt of roughly 160 kicks from the 1999–2005 era. These exist in the logs, but the film is elusive. There are a few games in particular (the 1987 NFC Championship featuring the Redskins and Vikings) where I have confirmation of the doinks via the broadcast team or local newspapers, but video is harder to find.

I am working newest-to-oldest, ensuring the 2000s expansion era (2000–2006) remains fully video-verified while we hunt for rare, poorly documented edge cases from the early 2000s.

Based on doink volume from 1999-2005 and absent rule changes that would have altered the outcomes and data, we are likely missing 20-25 doinks on average yearly from 1995-1998, as well as closer to 30 doinks missing yearly from 1974-1994. If this estimate is correct, the total number of eligible doinks from the game the goalposts moved until the advent of OffTheUpright is between 1,500 and 1,600.

Lots of ifs and assumptions, but that would mean that what we have is just over half of the historical data.

On the Importance of Asking the Wrong Questions

Or: "With Apologies to Cody Parkey"
Back when there were only 400-and-some doinks, Mason Crosby sat atop the theoretical leaderboard of off-the-upright kicks, neck and neck with Chris Boswell and Harrison Butker.

At that second, I formed a narrative and a hypothesis - long-tenured kickers' doinks said more about the culture of the franchises and their willingness to be patient with struggling kickers.

• Boswell famously considered retirement in the middle of a 2018 slump, but was supported by the team and was able to return to form.
• Parkey's double-doink kick was only one infamous attempt, but it was the end of his rope in Chicago.

It made sense. Or at least, I thought it did: Pause with me for a moment—

If I told you that the kickers who were the most doink-prone kickers per game in the league were Crosby, Boswell, and Butker, you would probably call bullshit - and you'd be right to do so.

The thing is, Mason Crosby has the most career doinks with a staggering 22, but he doesn't have the highest count of doinks per season:

"What kicker has the most career doinks per season?"
Butker and Parkey.

Confirmation bias is a cruel mistress.

The problem is that Parkey wasn't a full-time kicker, so comparing his 7 seasons of doinks with the same average per season as Butker is like using a ruler to measure first downs - the wrong tool for the job gives unsteady results.

Professional statisticians still reading at this point are likely screaming out the obvious. A proper rate statistic, normalized by games played, clears the situation up like new eyeglasses:

"What kicker has the highest doink rate?"
Cody Parkey.

I am sad to report that it's not even close.

• Parkey's doink rate of once every 5.3 games (or roughly 3 per season) is nearly double that of Chris Boswell (one doink every 9.9 games - 1-2 times every season).
• The second highest kicker all-time behind Parkey? Matt Gay, with a measly doink rate of once every 8 games (2 times every season).

I'm sorry, Cody.

Odds'n'Ends

• This project is for fun - I claim no ownership of the data or video footage.
• I don't collect your data because I don't want it (unless you're a kicker in the NFL and you hit the uprights).
• Got video footage available online for one of the games? Use the Submit page to help complete the data.
• See something wrong? Feedback is love.

So, do doinks matter? I'd like to think so. But if not, at least I've got some cool stories to share.