Introduction

Officially, football games are won and lost by the points a team scores. Unofficially, football is a battle for distance, measured in yards from the opponent’s endzone. While a defender's primary responsibility is to prevent points, their main objective after an offensive player gets the ball is to minimize the yards gained by the ball carrier.

Before getting into the data preparation, let me introduce the main metric used to evaluate the yards given up by the tackler after initial contact: "Yards After Contact." Given that the endzones in our tracking data are positioned at the far ends of the X-values, "Yards After Contact" focuses solely on the X-direction and is calculated as the difference in the ball carrier's X-value between the "first_contact" event and the "tackle" event. A higher X-distance between these events indicates a poorer performance by the tackler, as it means they allowed more yards after initial contact before completing the tackle. This metric is added to the tracking data and labeled as "x_totaldis_fc_t" or sometimes "x_dist."

This post investigates which attack angles by a tackler (referred to as Attack Zones) lead to the fewest yards gained after the tackler has made first contact with the ball carrier. The analysis focuses specifically on solo tackles, standardizing the directions of motion for both the ball carrier and the tackler. Although this falls under the metric category, the insights gained are invaluable for coaches as well.

Data Preparation

The primary data sources for this notebook are from Big Data Bowl (BDB) and the R stats package "nflverse." The BDB data includes game, play, player, tackle, and tracking data for weeks 1-9 of the 2022 NFL season. Unique "gameId" and "playId" columns were merged into a single "game_play_Id" column, which was then matched with the corresponding “nflverse” game and play data. This integration added additional context, such as solo tackles, penalties, and expected points added, for each "game_play_Id," covering over 50,000 plays in total.

To filter the data, I isolated every "game_play_Id" that met the following criteria:

• Involved a single solo tackler
• Did not involve a penalty
• Did not involve a turnover
• Did not end out of bounds
• Was not a special teams play
• Was not a sack

This filtering process left approximately 5,450 plays within the Weeks 1-9 tracking data. I then used the BDB data to identify the unique ball carrier and tackler "nflId" for each "game_play_Id" and isolated these two "nflId"s (along with the football) in the tracking data for all solo tackles.

A big part of the data preparation involved standardizing the ball carrier’s direction of motion (and play direction) to the right and adjusting the tackler's angle of attack based on changes in the ball carrier's direction. To do this, I modified all direction angles from the given 0 degrees (increasing clockwise) to the more standard 0 degrees (increasing counterclockwise). I then rotated both the ball carrier's and tackler’s angles so that the ball carrier’s final direction was always 0 degrees to the right.

It's important to note that the angle between the ball carrier and tackler (angle of attack) remained unchanged, only their direction of motion was standardized. While this adjustment altered the position data, once I had the "Yards After Contact" X-distance for each solo tackle, I no longer needed the position data, as my focus shifted to the angle of attack. This standardization enabled a consistent comparison of each tackle's angle of attack.

Exploratory Analysis

For context, the chart below illustrates the correlations between expected points added (EPA) per play versus total yards gained per play (left) and "Yards After Contact" (right). As anticipated, both total yards gained and "Yards After Contact" show a positive correlation with a high EPA per play. This indicates that a defender should aim to limit both metrics as much as possible.

In my exploratory analysis, I first examined the time duration between the "first_contact" and "tackle" events in the tracking data. As shown in the chart below, the vast majority of solo tackles occur within 2 seconds of first contact, with 80% happening in under 1.8 seconds.

To analyze the plays and tackles themselves, I isolated the tracking data for just the ball carrier and tackler for each solo tackle. I began by creating basic animated GIFs of each play. A sample of these animations is shown below, displaying the players' movements along the field using their X and Y position data, with a black arrow indicating their orientation (the direction they were facing) and a green arrow showing their direction of movement. This visualization provided important context on the players' movements that the raw data alone couldn’t.

In-Depth Analysis

After investigating whether a player's momentum and force (calculated by multiplying their weight by their speed and acceleration, respectively) influenced the "Yards After Contact" given up on a play, I decided to focus on the player's attack angle toward the ball carrier.

Tacklers are traditionally taught to square up to the ball carrier and meet them head-on (moving in opposite directions) to drive the ball carrier backward and minimize "Yards After Contact." While this makes sense theoretically, in practice, most ball carriers have a significant momentum advantage when they meet a tackler, often driving the tackler backward from the point of first contact.

By standardizing the ball carrier’s direction of motion to 0 degrees right and rotating the tackler’s direction accordingly, I derived a single angle measurement (tackler’s angle of attack) instead of dealing with multiple directions. Due to the variety of angles, I grouped them into 12 groups of 30 degrees each, forming 12 Attack Zones.

Zone 1 ranges from 0 degrees right counterclockwise to 30 degrees, continuing in this manner until Zone 12, which covers 330 degrees to 360 degrees (0) right. This grouping allowed for a clearer understanding of which of the 12 Attack Zones were more successful in limiting the "Yards After Contact" compared to examining 360 individual degrees. Below are the descriptive statistics for the Attack Zones, along with an overall chart showing the median values and counts for each Attack Zone.

Given the directional nature of the data, I created the radar charts at the top of the post. These charts display the direction and median "Yards After Contact" values, as well as a similar radar chart illustrating the EPA/play for each Attack Zone. Lower values in both metrics are better for a defender, as they indicate giving up fewer "Yards After Contact" and expected points when attacking from a smaller angle rather than a larger one.

Results and Discussion

Contrary to conventional tackling wisdom, meeting the ball carrier head-on (Attack Zones 1 and 12) does not result in the lowest median "Yards After Contact" or lowest EPA/play. Surprisingly, these zones exhibit some of the highest values. This may be because defenders might be stationary, waiting for the ball carrier, making it easier for them to be overpowered.

Interestingly, the best Attack Zones from which a tackler should engage a ball carrier are from the side and at an angle from behind. Tackling from the side makes sense as the ball carrier may overcome a head-on momentum transfer and continue moving, whereas momentum transferred from the side is harder for them to counter. Even if a tackler has less momentum than the ball carrier, a smaller force from the side can effectively bring the ball carrier down compared to a larger force head-on.

The same principle applies to tackling from behind. The additional momentum from a different angle is more challenging for the ball carrier to overcome than a direct head-on confrontation. While it may seem counterintuitive to tackle from behind, as it might push the ball carrier forward, this forward push is often less than the "Yards After Contact" gained by meeting the ball carrier head-on with the same momentum.

For limitations about these results, it's important to note that this data only considers solo tackles that include both the "first_contact" and "tackle" events. Tackles lacking a "first_contact" event are not included in the "Yards After Contact" calculations, as the final metric cannot be determined without the initial contact point. Further research is needed to understand the distinction between these events fully.

Additionally, this analysis did not explore how Attack Zone knowledge could assist tacklers in collaborative efforts, such as helping from the side or back to complete the tackle faster. Nor did it examine "missed tackles" and which Attack Zones result in the most missed tackles. Both assisted tackles and missed tackles should be investigated further, as this notebook focuses solely on completed solo tackles that have both "first_contact" and "tackle" events.

Recently on Twitter, Anna Gat asked the question, “What was the last Big Idea?” and there were many great responses, many of which have been world-changing.

For me, two ideas immediately came to mind: "Attention Is All You Need" a paper from Google researchers in 2017 that has transformed the world of AI, and Dr. Iain McGilchrist’s Divided Brain theory, introduced in his 2009 book “The Master and His Emissary.”

A brief intro to the two ideas (errors in summarizing are mine of course):

1. Attention Is All You Need: This paper introduced a new way of understanding and generating language via the transformer. Previously, models focused on word-by-word prediction and had trouble with longer inputs and understanding the context of the text. However, the new model doesn’t work sequentially word-by-word but instead reads the entire sentence or input as a whole and then uses its understanding of the context to focus its attention on the most important parts. This not only greatly increases text comprehension and generation but also enhances speed and efficiency.

2. McGilchrist’s “Divided Brain” theory: McGilchrist wasn’t the first to introduce the concept of the two hemispheres of the brain being better at different tasks, but he popularized the theory that they focus attention in different ways. The divided brain has a left hemisphere (LH) that is more focused on details and good at tasks like language, math, and logic, processing information step-by-step. The right hemisphere (RH), however, is better at understanding the big picture, appreciating art and music, and processing information all at once.

Considering these revolutionary ideas together for the first time, I realized that they both involve the same basic function: the role of attention. With LLMs, attention means the algorithmic weights given to words or parts of words – a machine version of attention. With the divided brain, it means the more traditional human understanding of attention. This post is an attempt to connect the two (with the help of ChatGPT in italics).

The main similarity between the two is that although the ideas handle attention in different ways, they both involve it simultaneously to achieve the best results:

• For the transformer, the two types of attention involved are called self-attention and multi-head attention. Self-attention weighs the importance of different words in a sentence relative to other words and to the text as a whole. Multi-head attention is similar but has multiple variations of the self-attention weights computed in separate layers and then merged, making it better at concentrating on separate details. Self-attention runs sequentially, and multi-head attention runs in parallel, so there are multiple ways the attention is calculated, and they run simultaneously.
  

• For the Divided Brain theory, the LH focuses on the details of something, and the RH focuses on how it all fits together. The LH is analytical and logical, focused on representing things in our mind (thus making virtual representations and removing the liveliness from objects), whereas the RH is holistic and looks at the whole picture, focused on presenting something as a whole (thus invoking the liveliness of objects). Both types of attention are crucial for understanding intricacies and the whole picture, and both happen simultaneously.

• Self-attention is similar to the RH of the brain since it takes into account the context of the integrative whole, whereas multi-head attention is similar to the LH since it’s better at focusing on details and specific parts.

For the 5th time in the past 10 years (previously 2023, 2022, 2019, and 2014), above is the NFL Draft Visit list from Walter Football organized by team and position. The numbers shaded in blue are conditionally formatted to show what position the team is concentrating on compared to other positions, those shaded in red are formatted to show what position each team is focusing on compared to other teams in the NFL.

This year I kept the same draft visit points system as last year:

• Visits worth 1 point = visits where the teams were already there and just talked to a player = Senior Bowl visit, NFL Combine visit, virtual visit
• Visits worth 2 points = visits where the team sought out the player, but in a group setting = Pro Day visit, Local visits (players that grew up or went to college in the area)
• Visits worth 3 points = individual visits where the player was brought in or worked out, often the most important = Private visits (teams have 30 of these to use), Workout visits

One difference this year is instead of displaying the sum total of the amount of draft visit points per team per position, I chose to display the percentage proportion of that team's total draft visit points used (what I call NFL Draft Visit Resources). The reasoning is simple: even if not every visit is reported, what IS reported should be consistent within the team. 

For example, if a team reports their visits often, it's likely that the reported visits are consistent across positions within that team, meaning we should be able to see what positions they're spending more resources on (resources in this case effectively mean time). Every team only has a certain amount of time pre-draft, and it's my theory that they allocate this time as efficiently and effectively as they believe they can. 

Please keep in mind that this list is in no way complete, as these are only the visits that teams/agents/players have reported that Walter Football thankfully organizes every year. There are likely many prospect visits that are not reported (for various team reasons) or not collected in the list. Also, some teams definitely report their visits A LOT more, there are some local newspapers/blogs that report a lot more, and there's certain players whose agents report more, etc. (For example: the data on Los Angeles Rams visits is pretty sparse, whereas the data on Pittsburgh Steelers visits is pretty robust. See the linked spreadsheet at the bottom for visit info)

Even though the data isn't 100% accurate, I believe it's still useful to look for trends on maybe what teams are concentrating on what positions. And I believe it'll be useful to similarly-minded mock draft nerds like myself. Happy mock drafting!

Link: Spreadsheet with data (keep in mind it was put together in 3 hours so only basic data verification/cleaning was done)

[The following was adapted from GPT-4 after I fed it a transcript of a 2 minute ramble I had recorded while walking a couple of weeks ago and asked it to turn it into a 1000 word essay]

I've come to understand — it has taken a while — that the impact of my words matters as much as their intention.

I want to help.

I want to optimize.

I want to make things better.

Yet, the art of feedback is delicate. It's about choosing the right moment, delivering thoughts with kindness and tact. It's a skill I'm still honing, balancing honesty with sensitivity.

Despite this awareness, there's a part of me that fiercely holds onto this trait; it's a part of who I am.

It drives me to question.

It pushes me to innovate.

It motivates me to seek excellence. 

But with this drive comes a responsibility – to be mindful, to be gentle.

So, here's my heartfelt revelation:

I am learning.

I am growing.

I am striving to find that sweet spot between sharing my insights and respecting the moment.

It's a journey of self-improvement, of understanding when my voice is needed and when silence is more powerful. 

In closing, this is my ongoing quest – to be thoughtful, to be considerate, yet to stay true to my instinct. To be the voice of change, yet the listener of the moment.

And amidst this complex dance of words and silence, I embrace this flaw, for it is also my strength. I'm working on being softer, more empathetic.

But in the depths of my soul, where truth whispers, I can't deny it:

I'm sorry... but I'm right.

[The following was copied and pasted from GPT-4 after I fed it a transcript of a 5 minute ramble I had recorded while walking just now and asked it to turn it into a 3000 word essay]

In today's digital age, the allure of becoming an influencer or content creator has captivated the imaginations of many, particularly the younger generation. This aspiration is not unfounded; the internet has democratized fame, allowing anyone with a camera and an internet connection to share their life, thoughts, and entertainment with the world. Platforms like Twitch, YouTube, and social media sites have become stages for individuals to broadcast themselves, playing video games, living their daily lives, or simply hanging out with friends. The concept of being paid to simply exist, to be oneself, has become a modern-day aspiration, symbolizing the ultimate dream: earning a living with minimal effort.

The phenomenon of influencers and streamers capitalizing on their daily lives for income taps into a long-standing human desire: to achieve success and recognition with as little exertion as possible. Platforms such as Twitch.com have popularized this trend, showcasing individuals who garner attention and income by sharing aspects of their lives, from gaming sessions to mundane daily activities. This trend is underscored by the importance of authenticity, especially for Generation Z, who value genuine self-expression and relatability in the content they consume.

However, the apparent ease with which influencers and content creators seem to earn their living belies the hard work and dedication required behind the scenes. Building an audience large enough to generate significant income through streaming or posting content online is a formidable challenge. It involves not only consistent content creation but also engaging with audiences, marketing oneself, and often innovating to stand out in a crowded space. The seemingly effortless lifestyle of successful influencers is the tip of the iceberg; beneath the surface lies a vast amount of unseen effort in content planning, production, and audience engagement.

Take, for example, the world of Twitch streaming. Some of the most popular streamers may spend hours daily broadcasting their gameplay or daily activities. Yet, what viewers do not see is the extensive preparation that goes into each session: the technical setup, the content planning, the interaction with fans off-camera, and the constant need to stay relevant in a rapidly changing digital landscape. Similarly, the trend of documenting one's life on social media might appear to be a simple task, yet it demands a constant stream of content creation, a knack for storytelling, and the ability to connect with followers on a personal level.

The dream of being paid to "do nothing" – to simply exist and share one's existence with an audience – is a misnomer. Successful content creators and influencers are storytellers, entertainers, and entrepreneurs who work tirelessly to build their personal brand and engage with their audience. They must navigate the complexities of the digital world, from understanding the algorithms that drive visibility on various platforms to responding to the shifting tastes of their viewers.

Moreover, the aspiration to become an influencer reflects a broader cultural shift towards valuing personal branding and individuality as commodities. In the past, careers were often defined by roles within organizations or specific professions. Today, the rise of the influencer economy suggests a move towards a more individualistic career path, where personal identity and the ability to captivate an audience can be as valuable as traditional professional skills.

This shift is not without its challenges. The saturation of the market means that for every successful influencer, there are thousands struggling to gain visibility. The pressure to constantly produce content that is both authentic and engaging can be mentally and emotionally taxing. Furthermore, the reliance on platforms and algorithms for visibility introduces a level of unpredictability and instability to these career paths.

Despite these challenges, the appeal of becoming an influencer persists. It speaks to a universal desire for recognition and to make a living by sharing one's passions and life with others. However, as with any career, success in this space requires hard work, persistence, and a willingness to adapt and grow. The dream of being paid for doing nothing is, in reality, a dream of being valued for who one is and what one can share with the world – a task that is far from effortless but potentially rewarding for those willing to undertake it.

As we look to the future, it is clear that the influencer economy will continue to evolve, shaped by technological advances and changing cultural attitudes towards work and entertainment. The journey to becoming a successful influencer or content creator is complex, demanding a blend of creativity, entrepreneurship, and resilience. Those who navigate this path successfully will not only achieve personal fulfillment but also redefine the boundaries of what it means to work and succeed in the digital age.