Goalkeepers are the hardest players to judge with numbers. Clean sheets depend on the whole defence. Save percentage treats a tap-in and a 35-yard shot as equal. A keeper behind a great defence may face few shots and look brilliant, while one behind a poor defence faces a storm and looks shaky. Modern analysis solves much of this with one idea: measure how hard the shots were to save.
Why the traditional numbers mislead
| Statistic | What it measures | The problem |
|---|---|---|
| Clean sheets | Matches without conceding | Mostly about the defence in front |
| Goals conceded | Goals let in | Depends on how many good chances were allowed |
| Save percentage | Saves divided by shots on target | Ignores how difficult each shot was |
A goalkeeper who faces mostly long-range efforts will have a high save percentage almost automatically. One who faces many close-range shots will have a lower one, however good they are.
Expected goals on target (xGOT)
Expected goals (xG) measures the quality of a chance before the shot is taken. Expected goals on target, xGOT, goes one step further. Calculated only for shots on target, it combines the original chance quality with where the shot ended up in the goal. A shot into the top corner gets a higher value than the same shot struck straight at the goalkeeper.
Like xG, xGOT runs from 0 to 1. A shot worth 0.3 xGOT would be scored about 30% of the time, which also means a typical goalkeeper would stop it about 70% of the time.
Goals prevented
Add up the xGOT of all the shots on target a goalkeeper faced and you get the number of goals an average keeper would have conceded. Compare that with the goals actually conceded:
goals prevented = xGOT faced − goals conceded
Example: Invented figures for illustration. Over a season, a goalkeeper faces shots on target worth 42.0 xGOT and concedes 36 goals. Their goals prevented is 42.0 − 36 = +6.0: the keeper stopped six more goals than an average goalkeeper would have, given the shots faced. A keeper with −4.0 let in four more than expected.
Making the comparison fair
Goalkeepers who face more shots have more chances to prevent goals, or to concede them. To compare keepers with very different workloads, analysts also use a rate: the goals a keeper was expected to concede divided by the goals actually conceded. In the example, 42.0 ÷ 36 = 1.17, meaning the keeper conceded noticeably fewer than expected.
Key point: Goals prevented separates the goalkeeper from the defence. A poor defence increases the xGOT a keeper faces, but it does not count against the keeper unless they let in more than an average keeper would.
What goals prevented does not measure
- Claiming crosses and sweeping. A keeper who comes out to stop a chance before a shot is taken gets no credit in xGOT.
- Distribution. Passing and long kicks are a big part of the modern role.
- Organisation. Communicating with defenders prevents chances that never appear in the data.
- Small samples. Over a few matches, goals prevented swings a lot. A full season is a much better guide.
A clean sheet example from 2025-26
Arsenal's David Raya kept 19 clean sheets in the 2025-26 Premier League, the most in the division, behind a defence that conceded only 27 goals. Clean sheets that high reflect both the keeper and the defence; goals prevented is the number that would separate the two. The season is summarised in The 2025-26 Premier League Season in Numbers.
To understand the expected goals model that xGOT builds on, read How to Read Football Stats: xG, xA and PPDA.
- Clean sheets and save percentage depend heavily on the defence and the shots faced.
- xGOT measures how likely an on-target shot was to go in, using where it ended up in the goal.
- Goals prevented (xGOT faced minus goals conceded) credits goalkeepers fairly over a full season.
Sources
- What are expected goals on target (xGOT)? — Opta Analyst
- Opta football stats definitions — Opta Analyst
Premier League and Champions League by the numbers.