What is Goals Saved Above Replacement?

cover photo belongs to Jeff Swinger/USA Today Sports

Goals Saved Above Replacement was first created back in 2013. Originally a stat that was solely eye-test related, it has since grown to a series of formulas building off the idea of comparing one goalkeeper action against a standard average. For example, if a goalkeeper is scored on in a certain situation, GSAR gauges how difficult the shot was and the percent chance an average MLS goalkeeper would save the attempt on goal. Across a variety of other situations a goalkeeper faces, GSAR finds the goals saved in comparison to a replacement player. If the number is large, the goalkeeper saved several goals and did an outstanding job. The farther a goalkeeper’s GSAR is below zero, the worse they have performed. An even “0” represents a score a replacement player would earn.

What makes GSAR any different than another goalkeeping metric?

There are two main differences. First, GSAR is an all-encompassing statistic. While most goalkeeping stats are centered solely around shot-stopping, GSAR takes into account crossing, handling abilities, distribution, slotted balls, and other situations that don’t fit in a save percentage model. In each situation, the goalkeeper’s actions are based off expectations from an average of season performances. For example, completed and incompleted passes are weighted according to the chance of goal creation, whether for the goalkeeper’s team or the opposing one.

The second main difference is found in the shot-stopping element. When considering the few advanced shot-stopping statistics out there, they are still handcuffed by the problem of looking at where the shot enters the goal, as opposed to where the ball passes a goalkeeper’s line of attack. If a goalkeeper gets scored on from atop the 18 into an upper corner but the goalkeeper is at the penalty spot (only six yards from the shooter), then the ball passed the goalkeeper at a much closer distance than where the ball entered the goalmouth. Along with tracking the speed of the shot, GSAR helps understand the difficulty of a shot more accurately through reaction time and the distance from ball-to-goalkeeper.

What categories does GSAR track?

GSAR is built to value every touch a goalkeeper makes. To make it easy to digest for readers, we’ll use the 2018 goals saved as a starting point.

Each MLS goalkeeper has had their season broken down into seven categories.

1. Shots <10 - The first two columns are shots from inside and outside ten yards. The distance measured is from the shooter to the goalkeeper, not the shooter to the goal.

While traditional expected goal models focus on a shooter’s location on the field, GSAR focuses on different criteria: where the shot passes the goalkeeper and how long the goalkeeper had to react.

A shot’s difficulty is not deemed by where it enters the goalmouth but how far the ball was from the goalkeeper when it intersected the goalkeeper’s dive line. This angle is affected by a goalkeeper’s starting position as well as where the shooter is located on the field. If a shot is taken near the end line, the ball will pass the goalkeeper within a few feet even if it is hit the upper corner. Similarly, if a goalkeeper is closer to the shooter, they “cut down the angle” and cover more of the goalmouth, putting the ball’s path closer to their body.

2. Shots >10 - While the first category is largely impacted by a goalkeeper’s reaction abilities, the second has more emphasis on a goalkeeper’s ability to move his feet and general angle play. Typically older goalkeepers perform well with farther shots and struggle on close-range ones, while younger goalkeepers are the reverse.

3. Penalties - Penalties aren’t a large part of the MLS season - only occurring once every six or seven games - but they do carry a heavy weight. On average a penalty has a success rate of around 80% and can severely boost or tank a goalkeeper’s GSAR.

4. Crossing - This category takes into account if a goalkeeper punched, claimed, or (for a negative value) let a cross drop in a position they should have challenged for. The position has recently seen a swing towards favoring passive goalkeeping when it comes to crosses, which explains the relatively low ratings.

5. Error - Covering many different areas, only negative numbers will be found here. This can include gifting a poor rebound to the opposition, giving away a penalty, or other actions that result in creating another chance on goal.

6. Misc. - The miscellaneous tallies cover any non-tradition goalkeeping action. Slotted balls back to the center of the box are the most common, as well as any actions that don’t fit a proper formula, which are entered in by hand. Unique shot deflections and 1v1 situations (amongst other actions) can be found here. While hand-adjusted values aren’t ideal, they help cover bizarre situations that a formula doesn’t work for. These situations make up less than .1% of all goalkeeping actions.

7. Passing - Passing stats consider how often and where a goalkeeper completes a pass as well as where turnovers occur. A turnover at the other side of the field is negligible, while a turnover in front of one’s own goal returns a larger negative value.

Minutes and average GSAR/90 minutes are tacked on at the end.

Projected Salaries Based Off of GSAR

There are many hurdles to tacking a dollar amount to a goalkeeper’s GSAR. For starters, identifying a baseline or replacement-level for MLS goalkeepers is tricky with salaries and talent levels constantly swelling over the past twenty years. Matching a “0 GSAR” goalkeeper with the median salary of MLS goalkeepers in 2018 ($132,625.00) proved to be the easiest route. Using this standard, we’ll take a look at seven different categories to obtain an overall GSAR rating, as well as put a dollar amount on each MLS goalkeeper’s performance from 2018.

Howard’s multi-million DP salary and GSAR were excluded for normalcy’s sake.

Howard’s multi-million DP salary and GSAR were excluded for normalcy’s sake.

Another challenge when considering this method is recognizing one team’s willingness to spend high on goalkeeping doesn’t necessarily mean the rest of the league will. To find a fair expected payment, the salaries and GSARs were listed in descending order to find a trend between the two. This brings up certain issues but overall it puts every goalkeeper on an even playing field when it comes to receiving payment for their services.

As some goalkeepers didn’t play the whole season - whether due to injury or a coach’s decision - finding a projected dollar amount would either have to extrapolate a goalkeeper’s stats for a full 34 game season or shrink down the corresponding payment. For example, Attinella only played two-thirds of the season but compiled a 4.74 GSAR. Should his projected GSAR-based salary be off what he could have done over 34 games or should it account for only the games he played? With a goalkeeper’s true impact being dependent on what they can bring to the field every game, I opted to extrapolate the dollar amount out to 34 games.

Categories are explained in more detail below. Goalkeepers are sorted by the difference ($$.diff) in their projected payment minus their actual. “$$.diff” is not what goalkeepers deserved to be paid, simply just the difference between actual and deserved.

1. m.GSAR/gm - Simply dividing a goalkeeper’s GSAR over the minutes they played, unless the goalkeeper played less than 900 minutes in the season. If this was the case, a goalkeeper was given either a positive or negative .03, depending on their GSAR. It’s not a great siphoning method, but +/- .03 keeps backup goalkeepers’ GSARs from getting out of hand with such a small sample size.

2. adj.GSAR - What a goalkeeper’s GSAR would have been had they played all 34 games (3060 minutes, excluding stoppage time).

3. gsar.$$ - How much a goalkeeper deserves to be paid, converted from a goalkeepers’ adj.GSAR. The conversation formula is based on the previous orange and white graph.

4. $$.diff - Goalkeepers are sorted by this column, which simply subtracts real.$$ from gsar.$$. Tyler Miller was underpaid by $333,503 while Andre Blake was overpaid by $410,805.

Have any questions? Head over to the contact page for any specific inquiries.

2019 MLS Goalkeeper of the Year Race

cover photo from US Sports Guru

This season Everybody Soccer will be tracking the top MLS goalkeepers with Goals Saved Above Replacement. Check out last year’s ratings with an explanation at how we arrived at a single, compact number. For more detail statistics on each goalkeeper, click here to view the web page which breaks down each goalkeeper’s contributions into seven categories, week-by-week performances, and how much they deserve to be paid.

Last Updated October 11, 2019

RankGSARKeeperTeam Mins GSAR/90
1 8.45 Steve Clark POR 2160 0.35
2 6.82 Matt Turner NE 1766 0.35
3 6.27 Maxime Crepeau VAN 2340 0.24
4 5.55 David Bingham LAG 2970 0.17
5 4.36 Luis Robles RBNY 2970 0.13
6 3.44 Sean Johnson NYC 2520 0.12
7 3.13 Tim Howard CLR 2205 0.13
8 2.83 Tyler Miller LAFC 2520 0.10
9 2.70 Tim Melia SKC 2880 0.08
10 1.99 Evan Bush MON 2880 0.06
11 1.86 Brian Rowe OCSC 2880 0.06
12 1.36 Bill Hamid DC 2970 0.04
13 1.15 Vito Mannone MIN 3060 0.03
14 0.59 Brad Guzan ATL 3060 0.02
15 0.02 Quentin Westberg TOR 2520 0.00
16 -0.26 Kenneth Kronholm CHC 1800 -0.01
17 -0.74 Stefan Frei SEA 3060 -0.02
18 -0.99 Daniel Vega SJ 3060 -0.03
19 -1.29 Jose Luis Gonzalez FCD 2880 -0.04
20 -1.43 Spencer Richey CIN 1710 -0.08
21 -1.91 Nick Rimando RSL 2610 -0.07
22 -2.86 Andre Blake PHI 2286 -0.11
23 -3.95 Joe Willis HOU 2430 -0.15
- 2.77 Zac MacMath VAN 720 0.35
- 2.08 Clint Irwin CLR 945 0.20
- 1.35 David Ousted CHC 1260 0.10
- 1.34 Carlos PHI 319 0.38
- 1.04 Brad Stuver NYC 450 0.21
- 0.77 Eloy Room CLB 1080 0.06
- 0.32 Andrew Putna RSL 450 0.06
- 0.22 Adrian Zendejas SKC 90 0.22
- 0.19 Jimmy Maurer FCD 180 0.10
- 0.10 Jeff Attinella POR 900 0.01
- -0.06 Pablo Sisniega LAFC 540 -0.01
- -0.14 Chris Seitz DCU 90 -0.14
- -0.15 Matt Lampson LAG 90 -0.15
- -0.22 Greg Ranjitsingh OCSC 180 -0.11
- -0.38 Eric Dick SKC 90 -0.38
- -0.44 Clement Diop MON 180 -0.22
- -0.43 Jon Kempin CLB 270 -0.14
- -0.48 Ryan Meara RBNY 90 -0.48
- -0.58 Matt Freese PHI 455 -0.11
- -1.16 Zack Steffen CLB 1170 -0.09
- -1.16 Brad Knighton NE 664 -0.16
- -1.38 Alex Bono TOR 630 -0.20
- -1.72 Tyler Deric HOU 630 -0.25
- -1.88 Cody Cropper NE 630 -0.27
- -3.24 Joe Bendik CLB 540 -0.54
- -4.83 Przemyslaw Tyton CIN 1350 -0.32