- Joined
- Apr 14, 2013
- Messages
- 9,709
- Reactions
- 5,045
- Points
- 113
For those not familiar with it, Elo is a statistic that was created for chess, to compare "relative skill levels" of players, and has been adapted for tennis by people like the folks at Ultimate Tennis Statistics and Jack Sackmann of Tennis Abstract (the two sites use slightly different formulas, with Sackmann's numbers being somewhat lower, but they are mostly consistent as far as relative values...I use UTS because their records are more thorough, and because it adjusts for tournament level). I'm not going to get bogged down here with a deep explanation of Elo, but here's Sackmann for a deeper discussion.
Elo is a tool, no more, no less. It is not meant to be fully definitive of how good players are, though Sackmann used it as the basis of his "Tennis 128" rankings. In truth, as Sackmann put it, it is meant as an alternative to the ATP rankings.
What do the Elo values mean? In other words, how good is, say, a "2173" Elo? Well, here are some guidelines, using UTS's version. These values represent the average player at different Elo ratings in the year-end rankings across the Open Era (1968-2022).
1830 Elo: #100
2000 Elo: #30
2125 Elo: #10
2223 Elo: #5
2348 Elo: #2
2415 Elo: #1
Meaning, if you average out the Elo ratings of every #1 (in the year-end Elo rankings), it is 2415. This gives us a basic comparison: When a player has a 2415 Elo (or so), they're as good as the typical #1 player. Or we can look at 2000 Elo as the rough level of a player good enough to be seeded at Slams, and 1800ish as the level of a top player.
Players enter the tour at 1500 and then go up from there (1500 is the lowest possible). The highest ever Elo was 2629 by Novak Djokovic in early 2016 (according to Sackmann it was Borg at 2473; UTS has Borg's highest being 2622).
According to UTS, a 400 point difference equates with a roughly 90% chance of winning. Meaning, your typical #2-3 player (2300 Elo) has a 90% chance of beating a 1900 Elo player (roughly #50ish). Now of course this is only in terms of on-paper probabilities, and is only a rough estimate; certainly, the actual real-life--flesh and wood and pressurized rubber match doesn't always follow probabilities. But it gives us an estimate, based upon player performance.
Furthermore, Elo ratings are meant to be constant over time. Meaning, a 2300 Elo in 1975 is of equal value as 2300 in 2015...or at least its supposed to be. This allows us to compare across eras, or at least within the Open Era (evidently Sackmann has Elo ratings for the pre-Open Era, but has of yet not shared them publicly).
So where's the fun?
It is in the charts, of course! So let's start with this one:
Elo Rating Ranges of the Big Four (2000-22):
El Dude, all I see are a bunch of colors...what am I looking at here? Well, as you can see, this charts the careers of the Big Four (except for Roger's two first partial seasons in 1998-99, when his Elo ratings were too low to be on the chart...you can see the downward arrow on the graphs of Roger in 2000, Rafa in 2005, Novak and Andy in 2006).
The bar represents the player's Elo range within a given year. For instance, we can see that in 2004, Roger ranged from 2261 (his year-end in 2003) to 2420, where he ended the year - which is denoted by the star.
There are some details that this chart picks up. For instance, notice Rafa's first year when he rose quickly. And notice also where the star is in each bar...with Rafa, you can see how it is usually not at the top, because his Elo tends to peak after clay season, and then go down over the second half of the season (with a few exceptions, most notably 2008, 2010, and 2018-19...the first and last two being the three seasons that he was the YE #1 in Elo).
Most importantly, this chart gives us a sense of the relative strength of these players year to year, outside of their actual titles. We can see Roger's early dominance and peak, Rafa joining him at a lower level, Novak and Andy coming along, and then Roger and Rafa being neck-and-neck from 2008-12, with Novak joining them in 2011 and then steadily rising to his peak in 2015-16, which was higher even than Roger's in 2005-07.
We can also see Rafa's steep decline from 2013-16, then his rise up again; and Novak's similar (though less extreme) decline from 2015-19, and then how both stabilized at a high--but lower level than their respective peaks.
And then there's Andy, who hovered below the other three until he rose in 2016, then plummeted every year since. What is most striking about his decline is the stars: he ended each year at the bottom of his range, which means he didn't plateau, and only steadily declined.
There are also some subtleties that are picked up. For instance, in 2013, Rafa peaked at a higher level than Novak, but Novak finished at a higher Elo because of a stronger second half. We can see similar dynamics in 2010-11, when Rafa peaked higher than Roger, but Roger finished higher because he was stronger in the later season. This also means that if the year ended after Roland Garros, Rafa would have more "year-end" #1 finishes.
On one hand, this chart doesn't really tell us anything new, but illustrates it visually to highlight some of the subtleties, and allows us to see how the Big Four have compared to each other over the years - as well as to themselves. But there are also some somewhat new revelations. For instance, it really brings forward the fact that there were not one but two periods in which the Big Four were all relatively close: I like to call 2012 the "Year of the Big Four" because they all won a Slam and were at their closest, but it really started in 2009. And then again, in 2017-18, they were somewhat close, although it was more of a "passing ships by night" as Andy fell, and Rafa and Roger rose back.
One final note. The chart highlights a point of concern for Rafa fans: his Elo trajectory over the last few years. Notice how not only has it gradually lowered, but each year his year-end Elo is the lowest. So while 2016-19 saw a clear upward trend, as he "Agassi-ed" his later career, 2019-22 has a clear downward trend.
This also illustrates some similarities: What 2013 was to Roger, 2016 was to Rafa and 2018 was to Novak...a low point in their later career (for Novak, just to be clear, it was early 2018 - as he grew stronger in the latter half of the year). All three bounced back to elite level, though at a lower level than their peaks. And herein lies the difference with Andy: he didn't bounce back, didn't have that second "Agassi-style" wind.
Elo is a tool, no more, no less. It is not meant to be fully definitive of how good players are, though Sackmann used it as the basis of his "Tennis 128" rankings. In truth, as Sackmann put it, it is meant as an alternative to the ATP rankings.
What do the Elo values mean? In other words, how good is, say, a "2173" Elo? Well, here are some guidelines, using UTS's version. These values represent the average player at different Elo ratings in the year-end rankings across the Open Era (1968-2022).
1830 Elo: #100
2000 Elo: #30
2125 Elo: #10
2223 Elo: #5
2348 Elo: #2
2415 Elo: #1
Meaning, if you average out the Elo ratings of every #1 (in the year-end Elo rankings), it is 2415. This gives us a basic comparison: When a player has a 2415 Elo (or so), they're as good as the typical #1 player. Or we can look at 2000 Elo as the rough level of a player good enough to be seeded at Slams, and 1800ish as the level of a top player.
Players enter the tour at 1500 and then go up from there (1500 is the lowest possible). The highest ever Elo was 2629 by Novak Djokovic in early 2016 (according to Sackmann it was Borg at 2473; UTS has Borg's highest being 2622).
According to UTS, a 400 point difference equates with a roughly 90% chance of winning. Meaning, your typical #2-3 player (2300 Elo) has a 90% chance of beating a 1900 Elo player (roughly #50ish). Now of course this is only in terms of on-paper probabilities, and is only a rough estimate; certainly, the actual real-life--flesh and wood and pressurized rubber match doesn't always follow probabilities. But it gives us an estimate, based upon player performance.
Furthermore, Elo ratings are meant to be constant over time. Meaning, a 2300 Elo in 1975 is of equal value as 2300 in 2015...or at least its supposed to be. This allows us to compare across eras, or at least within the Open Era (evidently Sackmann has Elo ratings for the pre-Open Era, but has of yet not shared them publicly).
So where's the fun?
It is in the charts, of course! So let's start with this one:
Elo Rating Ranges of the Big Four (2000-22):
El Dude, all I see are a bunch of colors...what am I looking at here? Well, as you can see, this charts the careers of the Big Four (except for Roger's two first partial seasons in 1998-99, when his Elo ratings were too low to be on the chart...you can see the downward arrow on the graphs of Roger in 2000, Rafa in 2005, Novak and Andy in 2006).
The bar represents the player's Elo range within a given year. For instance, we can see that in 2004, Roger ranged from 2261 (his year-end in 2003) to 2420, where he ended the year - which is denoted by the star.
There are some details that this chart picks up. For instance, notice Rafa's first year when he rose quickly. And notice also where the star is in each bar...with Rafa, you can see how it is usually not at the top, because his Elo tends to peak after clay season, and then go down over the second half of the season (with a few exceptions, most notably 2008, 2010, and 2018-19...the first and last two being the three seasons that he was the YE #1 in Elo).
Most importantly, this chart gives us a sense of the relative strength of these players year to year, outside of their actual titles. We can see Roger's early dominance and peak, Rafa joining him at a lower level, Novak and Andy coming along, and then Roger and Rafa being neck-and-neck from 2008-12, with Novak joining them in 2011 and then steadily rising to his peak in 2015-16, which was higher even than Roger's in 2005-07.
We can also see Rafa's steep decline from 2013-16, then his rise up again; and Novak's similar (though less extreme) decline from 2015-19, and then how both stabilized at a high--but lower level than their respective peaks.
And then there's Andy, who hovered below the other three until he rose in 2016, then plummeted every year since. What is most striking about his decline is the stars: he ended each year at the bottom of his range, which means he didn't plateau, and only steadily declined.
There are also some subtleties that are picked up. For instance, in 2013, Rafa peaked at a higher level than Novak, but Novak finished at a higher Elo because of a stronger second half. We can see similar dynamics in 2010-11, when Rafa peaked higher than Roger, but Roger finished higher because he was stronger in the later season. This also means that if the year ended after Roland Garros, Rafa would have more "year-end" #1 finishes.
On one hand, this chart doesn't really tell us anything new, but illustrates it visually to highlight some of the subtleties, and allows us to see how the Big Four have compared to each other over the years - as well as to themselves. But there are also some somewhat new revelations. For instance, it really brings forward the fact that there were not one but two periods in which the Big Four were all relatively close: I like to call 2012 the "Year of the Big Four" because they all won a Slam and were at their closest, but it really started in 2009. And then again, in 2017-18, they were somewhat close, although it was more of a "passing ships by night" as Andy fell, and Rafa and Roger rose back.
One final note. The chart highlights a point of concern for Rafa fans: his Elo trajectory over the last few years. Notice how not only has it gradually lowered, but each year his year-end Elo is the lowest. So while 2016-19 saw a clear upward trend, as he "Agassi-ed" his later career, 2019-22 has a clear downward trend.
This also illustrates some similarities: What 2013 was to Roger, 2016 was to Rafa and 2018 was to Novak...a low point in their later career (for Novak, just to be clear, it was early 2018 - as he grew stronger in the latter half of the year). All three bounced back to elite level, though at a lower level than their peaks. And herein lies the difference with Andy: he didn't bounce back, didn't have that second "Agassi-style" wind.
Last edited:
