xFpts: Defenders

Today we are continuing our xFpts analysis by looking specifically at defenders. If you haven’t already read our introductory article, take a look here! We cover the basics of our new metric, xFpts, and analyze forwards there. Our midfielder analysis can be found here.

xFpts modeling for defenders

As a reminder, we use xG and xA to scale our estimates of expected fantasy points in order to reduce the amount of luck reflected in a player’s point totals. Unlike forwards and midfielders, however, defenders come with an additional challenge – the clean sheet. While goals and assists still represent a large share of defender fantasy points, clean sheets and goals against are two additional point categories that strongly affect a defender’s ultimate point total (6 and -2 points, respectively). Since both are directly linked to the number of goals the opposition scores, the resulting points from these categories are then similarly at the mercy of luck. If a player’s team concedes 2.0 total xG but somehow still ekes out a goal-less draw, we would want to take this under consideration when evaluating his value. More likely than not, he was fortunate not to concede.

Therefore, in addition to goals and assists, clean sheets and goals against are scaled proportional to the total xG the opposing team accumulates against the defender’s team (Fig 1).

Figure 1. Calculation of xFpts for each position group, assuming standard togga scoring rules.

Just like with the forwards from our first post, I wanted to assess how my xFpts estimates compared to the actual observed points players scored as starters. To visualize how Premier League defenders have performed this season thus far, I plotted the average fantasy point and xFpt output of each player with at least 5 starts for each position group (Fig 2).

Figure 2. Scatterplot of average xFpts versus average actual fantasy points scored for defenders.

The black line indicates where the xFpt average equals the actual fantasy points average. The dotted red line represents the threshold where a player is out-performing their xFpt average by more than 1.5 points while the dotted blue line represents under-performance by at least 1.5 xFpts per week. Therefore, players above the red line have been getting noticeably more fantasy points than expected from their xFpts thus far (may have been lucky), while those players beneath the blue line were expected to earn more fantasy points than they have thus far this season (may have been unlucky).

Unlike the forwards and midfielders (post coming tomorrow!), defenders on average have a greater discrepancy between their xFpts and their observed fantasy point totals. This is because clean sheets and goals against are a lot less predictable. A prime example of this is Manchester City’s 0-2 loss to Tottenham, where the xG would have implied the opposite result, namely 3.1 – 0.3 in Manchester City’s favor. Obviously, that example is a tremendous outlier, but also points to this being a noisier statistic to predict.

Marcos Alonso and Emerson are…the same player?

The most striking example of the utility of using our xFpts model is when it is applied to the curious case of Emerson and Marcos Alonso. Both play the same position at Chelsea, have started the same number of games (10), and often when one is chosen, it is at the expense of the other. If you only look at the observed points per start, Marcos Alonso is in first place with 18.75 points per start! This surpasses even Trent Alexander-Arnold, widely considered the best fantasy defender this year. Emerson, on the other hand, merely averages 8.7 points per start, making him the 36th best defender by this metric. Clearly Marcos Alonso is the superior player, perhaps even the most superior player you can start on defense, right? Think again! When considering xFpts, a new picture emerges. I’ve plotted a boxplot of both player’s actual fantasy points per start and their xFpts per start (Fig 3).

Figure 3. Comparison of the distributions of actual fantasy points and xFpts between Emerson Palmieri and Marcos Alonso.

This plot allows you to visualize the distribution of points scored under each measure. The huge gap between their performances in the actual fantasy point data disappears completely when converted to xFpts. It becomes clear that these players have performed remarkably similarly despite the massive difference in actual fantasy points. Where Marcos Alonso has been very fortunate with his attacking and defensive actions, Emerson has had an equal and opposite dose of very bad luck.

Since clean sheets are so important to defender points, it could be argued that this difference is a result of scheduling. A cursory glance at the schedule refutes this, as both players faced relatively equally difficult opponents. When starting, Emerson’s Chelsea has conceded only 0.93 xG/game compared to Marcos Alonso’s 0.97. With similar defensive outputs, you’d also expect similar goals conceded. This is not the case, as Chelsea concedes a full 0.6 more actual goals/game when Emerson starts. Assuming xG is doing what it should be doing, with sample sizes of just 10 games each, this is almost assuredly down to luck/sampling noise. Given more opportunities, Emerson and Marcos Alonso should put up similar numbers.

Important point to consider here: due in some part to his poor luck, Emerson has fallen out of favor with Frank Lampard. Sometimes players who are under-performing their true production see playing time fall simply as a product of perceived incompetence. That being said, with the upcoming pandemic-induced congested fixture list and increased rotation, I think it’s likely Emerson gets a second chance. If he does, he is a must-start.

Comparing your potential starting defenders

For your managerial starting decisions, we’ve created this defender xFpts ranges plot for quick player comparisons (Fig 4). Again, the error bars represent +/- 1 standard deviation.

Figure 4. xFpt ranges for eligible defenders, ranked in order of mean xFpt performance. 1 standard deviation from the mean is plotted in both directions.

Finally, I’ve again embedded the top 25 defenders ranked by mean xFpts. Quick observations:

1. All Chelsea wingbacks score in the top 8. Presumably a product of their wingback system, anybody who starts at outside back for Chelsea is a must-start.

2. Unsurprisingly there is a general bias towards wingbacks across the ranking, simply due to a greater tendency to accumulate offensive points in addition to defensive production.

3. For anyone still hoping Hector Bellerin turns the corner, I’m not optimistic. He ranks second to last across his 7 starts, outperforming his paltry 1.4 xFpts total by 4.2 actual fantasy points.

4. Early returns from Aymeric Laporte, however, are quite promising. He lands just outside the top 25 with a respectable 9.65 xFpts/start. Just gotta hope he doesn’t get hooked early!

Please don’t hesitate to get in touch with any questions or comments!

Below you can find a link to download the full table with our xFpt statistics for all 131 defenders. All code and data can be found on our github. xG and xA data pulled from http://www.fbref.com and fantrax statistics from fantrax.com.

NOTE: Data is only included through week 29 of the 2019/2020 season. For up-to-date ranks, please refer to our xFpts Ranking page.

defenders_xpts_table-1Download

Special thanks to Andy Gear for his help again.