What is a field goal? In some cases, a field goal is a failure: few things deflate a fan’s excitement more than a big play into a scoring opportunity followed by a stalled drive and a three point kick. In other cases, a field goal is a life-raft: many a fan has bargained with the football gods before halftime, “If we can just get three here, then we can make a push to close the gap in the second half.” In all cases, a field goal is better than nothing, although less preferable than a touchdown.

The next installment of the Offseason Special Teams Series centers around those field goals. In a previous installment, I talked about punts and positioning, looking at how a team’s punting game affected their outcomes. This week, I turn that attention to field goals, diving deep into the general landscape of kicking in college football.

I. An Empirical Survey of Field Goals in 2018

I.A Summary Statistics

This section begins with a short note regarding data and reliability. The data I have comes from collegefootballdata.com, a website set up by a dedicated fan and community-oriented reddit user. It is a great place to look if you are searching for play-by-play, game-by-game, or any other kind of in-the-weeds raw data for college football. The data is comprehensive, and on average, well-above accurate. In the process of conducting this project, I noticed 4 field goals I cannot account for: one from New Mexico, and three from South Alabama. I have put no game filters, no garbage time restrictions on the data, and so somewhere in the process of cleaning or scraping, these four were lost. Given the size of the data, I a not too concerned, but, in the interest of full disclosure, I felt it prudent to document the omission somewhere in my analysis. Now, on to the numbers.

What do these numbers tell us? Well, not much, without context, as with many numbers. But, we do see some trends: in the top five longest teams, we have four pretty-not-great offenses (Pitt at 60th is the highest) and one pretty great offense (WVU at 24th). So, it seems that two things can generate long field goal distances. Either 1) you’re a bad team who struggles to move the ball, indicating that even when you do get lucky with a big play, you can’t capitalize and are forced to kick to try for points or 2) you’re a good team who gets lucky often with big plays, so when you kick, it’s only when you didn’t get a big play, which is rare.

As for the shortest distances, I have three hypotheses. First, your coach is a stodgy decision maker who calls plays in the red zone in a weird way, and never wants to go for it on fourth down (hello Stanford, NC State, BC, MS). Second, you’re a bad team with a sparky offense and you can get the ball into opponent territory and not do much else (UMass). Third, you’re a flying death machine and sometimes you’re up by so much you don’t care to go for it on fourth down inside the red zone (OU).

The two stand-outs in this group for me are Oklahoma and Oregon. Let me put fandom aside for a moment - Austin Seibert is an incredible kicker. He’s a special teams gem. I could write an entire Austin Seibert appreciation post. He added so much value to an already amazing Sooner team. Secondly, Oregon had a rough year. I’m not here to put struggling kickers on blast, but 54% from a college kicker, in the Power Five? Rough. The Ducks left some points on the board in trusting their kicker too much.

I.B Calculating Expected Points

To help contextualize the raw stats, I created a measure of expected points for every field goal kicked this season. I grouped field goals by yard line, placing each field goal into a five yard bucket, and then took the average amount of points teams scored in that situation. Then, I added that average points number up for each of a team’s field goals and compared it to their total actual points. The result is below:

On this graph, the red line is a predicted value for each team of expected points on actual points. It’s not quite a 45 degree line, so the rule of thumb doesn’t hold exactly, but generally, above that line is good, below that line is bad. We see the overachievers: Illinois, ECU, Utah, Syracuse, and we see some underachievers: Texas A&M, Cincinnati, UNC, Mississippi.

Top Five Expected Points Overachievers

1. Utah (+14.0795 points)
2. LSU (+14.0789 points)
3. Georgia Southern (+13.29 points)
4. Syracuse (+13.169 points)
5. Illinois (+11.6115 points)

Top Five Expected Points Underachievers

1. Texas State (-12.68 points)
2. Tulsa (-11.64 points)
3. Arkansas State (-9.589 points)
4. Cincinnati (-9.586 points)
5. Wisconsin (-8.99 points)

Let’s just check the S&P+ Special Teams ratings real quick. Utah, Syracuse, LSU are the top 3 teams in ST S&P+ this season, and all five teams in the bottom five finished outside the top 100. The top 4 teams all secured more than four FGs of added value this season, points that came in handy, for example, for LSU and Syracuse teams in divisions with college football powerhouses. On the flip side, the bottom four teams all missed out on more than three field goals. That may not seem like much, averaged out per game, but ask any coach of those five teams, and they’d gladly take those three scores back.

I.C Modeling Marginal Effects of Field Position

The next step in our analysis of field goals is to estimate the probability of those field goals given a situation. I carried out a simple probit regression of field goal success on field goal distance, controlling for each team’s kicking unit. A one yard increase in field goal distance is associated with a 1.7% decrease in probability of making a field goal, all else constant. The average predicted probability for each field goal distance is plotted below.

The red line represents the average probability of a made field goal from any given distance. The scattered dots are the team effects. As the field goal distance increases, we see a broader spread of dots, highlighting the relative strength and weakness of each team. TCU, for instance, is predicted as, on average, being 16% less likely to make a field goal from any given distance than the average team. Those team effects tell some interesting stories. OU, as noted above, got a lot of value from the kicking position: 25.9% more likely to make a FG. Those numbers can also obscure reality - Pitt, for instance, is 39.7% more likely to make a FG, but they also had one of the shortest average FG distances. What that means is that, in the model, Pitt was given credit for a lot of high probability short field goals they never actually made. In reality, the Panthers were -7 expected points from FGs on the season. That example helps keep our inference on these numbers in check, but all in all, these expected probabilities provide a nice benchmark for team rankings.

II. TCU Field Goals by the Numbers

II.A Summary Statistics

TCU kicked 21 field goals this season, excluding PATs (which require little analysis other than “teams should make these just about every time”), missing 8 (61.9 completion %). 61.9% is ~12 percentage points below the aforementioned national average of 73.47%. TCU kicked field goals of an average length of 34.7 yards (80th longest in the nation, or, a little bit below average for the nation). For those of you keeping score at home, you add the 10 yards for the end zone and about 4 yards for the holder, so TCU kicked from the 20 yard line, on average, a mean indicative of a fairly conservative kicking strategy. 50% of TCU’s FG attempts landed in the 29-41 yard range (15-26 yard line). The longest make was 46 yards, and the shortest miss was 31 yards.

II.B Play By Play (Note: this is detailed, skip to the graph without feeling bad)

• 1st Qtr vs. Southern - TCU 3, Southern 0: 26 yard FG (Expected points: 2.618, Actual Points: 3, Aggregate difference: +.382)
• 3rd Q vs. Southern - TCU 41, Southern 7: 30 yard FG (EP: 2.618, AP: 3, ADiff: +.764)
• 1st Q vs. SMU - TCU 0, SMU 9: 48 yard FG MISSED (EP: 1.515, AP:0, ADiff: -0.751)
• 1st Q vs. Ohio State - TCU 0, OSU 3: 31 yard FG MISSED (EP: 2.415 , AP: 0, ADiff: -3.166)
• 1st Q vs. Texas - TCU 3, Texas 0: 46 yard FG (EP:1.515, AP: 3, ADiff: -1.651)
• 1st Q vs. Texas - TCU 6, Texas 7: 23 yard FG (EP: 2.828, AP: 3, ADiff: -1.479)
• 3rd Q vs. Texas - TCU 16, Texas 10: 29 yard FG (EP: 2.618, AP: 3, ADiff: -1.097)
• 4th Q vs. Texas - TCU 16, Texas 24: 41 yard FG MISSED (EP: 1.808, AP: 0, ADiff: -2.905)
• 4th Q vs. Iowa State - TCU 17, ISU 14: 28 yard FG (EP: 2.618, AP: 3, ADiff: -2.523)
• 2nd Q vs. Texas Tech - TCU 7, TTU 3: 47 yard FG MISSED (EP: 1.515, AP: 0, ADiff: -4.038)
• 2nd Q vs. Oklahoma - TCU 24, OU 28: 41 yard FG (EP: 1.808, AP: 3, ADiff: -2.846)
• 3rd Q vs. Oklahoma - TCU 27, OU 31: 41 yard FG (EP: 1.808, AP: 3, ADiff: -1.654)
• 3rd Q vs. Oklahoma - TCU 27, OU 38: 38 yard FG MISSED (EP: 1.963, AP: 0, ADiff: -3.617)
• 2nd Q vs Kansas - TCU 3, KU 7: 31 yard FG (EP: 2.415, AP: 3, ADiff: -3.025)
• 1st Q vs WVU - TCU 3, WVU 0: 30 yard FG (EP: 2.618, AP: 3, ADiff: -2.643)
• 1st Q vs Baylor - TCU 3, BU 0: 29 yard FG (EP: 2.618, AP: 3, ADiff: -2.261)
• 1st Q vs Ok State - TCU 0, OSU 3: 40 yard FG MISSED (EP: 1.963, AP: 0 , ADiff: -4.224)
• 3rd Q vs Ok State - TCU 24, OSU 3: 26 yard FG (EP: 2.618, AP: 3, ADiff: -3.842)
• 4th Q vs Ok State - TCU 31, OSU 24: 34 yard FG MISSED (EP: 2.415, AP: 0, ADiff: -6.257)
• 4th Q vs Cal - TCU 7, Cal 7: 44 yard FG MISSED (EP: 1.808, AP: 0, ADiff: -8.065)
• OT vs Cal - TCU 10, Cal 7: 27 yard FG (EP: 2.618, AP: 3, ADiff: -7.683)

TOTAL:

Expected Points: 46.732 (38th in the nation)
Actual Points: 39 (73rd in the nation)
Difference: -7.683 pts (11th worst nationally)

In games where TCU attempted a FG, they averaged a loss of 0.64 expected points per game due to kicking. That doesn’t sound like much, and on its own, it isn’t much. .64 expected points a game would’ve only affected the final outcome in one game - so maybe it would’ve helped to avoid the worst loss of the Patterson era, but not much else.

What we have to consider with field goals is the path-dependency of the game of football. Football is non-ergodic; that is, the game is not a series of independent events that you could alter and expect nothing else to change. No, each changed result (in this case, a missed FG changed to a made FG), would change the field position of the opposing team, the intangible confidence of the team, and perhaps even the strategy (due to game states).

For instance, let us consider field goal misses in games TCU lost. The Frogs missed 4 field goals in games they lost. Again, when we aggregate, that number sounds benign (and it is, for the most part, until you consider the fact that lack of trust in TCU kickers played a big role in TCU kicking decisions, a post in itself for another day). We can examine how the missed field goal sealed the Frogs’ fate in each of the four games. (And for fun, I’ve included in parenthesis the percentage change in make expectancy associated with one extra yard and five extra yards gained on the play before)

The drive after those four misses, for the opposing team:

Texas - Frogs miss a 41 yard FG, could’ve brought the game to 24-19, within one score. Starting at the 24 yard line, Texas drives 76 yards for a touchdown. (+1 yard = +3%pts on avg, + 5 yards = +11%pts on avg)

Ohio State - A missed 31 yard FG early on stymied the hype of the Frogs, depriving them of an opportunity to set the tone. Instead, the Buckeyes flipped the field and TCU gave up a turnover touchdown, digging themselves a 10-0 hole. (+ 1 yard = +.03%pts, + 5 yards = +9%pts)

Texas Tech - With a scoring opportunity in hand to exentend the lead, the Frogs miss a 47 yard FG after three failed pass attempts (1 scramble, 2 incompletions), giving the Raiders a chance to run out the clock and stop the bleeding. Who knows how this game turns out if the Raiders go into the half down 10-3. (+1 yard = +6%pts , +5 yards = +13%pts)

Oklahoma - In the 3rd quarter, TCU could’ve kept pace with OU, pulling again within a touchdown of the lead. Instead, TCU missed a 38 yard FG and the dam broke, as OU and TCU traded punts before the Sooners turned on the jets and never looked back. (+1 yard = +1%pts, +5 yards = +8%pts)

III. Conclusions

In summary, I’ve begun a line of research into contextualizing college football stats. This essay is just the tip of the iceberg, but it lays a nice foundation for analysis: isolating situations, describing context, and then modeling outcomes based on situation. My study here reveals that teams suffered and benefitted up to 9 points (positive or negative) over the course of a season as a direct result of field goal performance. Further investigation into the effects of game state (up or down in scoring, time of clock), sequencing (multiple kicks in a game, success after a make or a miss), and more historical data (a project for a future day, certainly) are required to validate and perfect this research. As it stands, we have a baseline picture of the landscape and impact of field goals on college football.