Much like Steph Kovalchik did a few days ago, now that there is no live tennis I found myself wondering which historical matches to watch. We both developed approaches to pick the most interesting ones, and I think the results are different enough that I thought I’d put my selection up too.

The way I decided to go about it was to rank matches by players’ combined Elo rating. I fit an Elo model that takes into account surface skill differences as well as the margin of victory as measured by the game spread (perhaps more on that in a future post). I then summarise each match by the total rating. One slight technical detail: because the spread of ratings is different across the different surfaces in my version of Elo, a rating of 2400 on clay would be less impressive than the same value on hard courts. I therefore rank matches based on the z-score of their total Elo, which takes the spread of the ratings into account. For example, if one player has a rating of 2400 and the other has a rating of 2200, and the standard deviation of ratings is 100, I would compute the z-score as $\frac{(2400 - 1500) + (2200 - 1500)}{\sqrt{100^2 + 100^2}}=11.3$, whereas if the standard deviation is 200, this would only produce a z-score of 5.7. I keep only matches that went to five sets and only consider matches from 1980 onwards, since finding matches from before that can be difficult (or, if they exist, their quality is often quite bad).

What kind of matches should this approach pick out? The total Elo skill should tell us something about the quality of the match. If both players have a very high Elo rating, they should both be considerably better than average players, giving the sum of their ratings a high z-score. Keeping only five-setters should go some way towards ensuring that the matches were competitive. With hindsight, I’m not sure that goes far enough: for some matches, perhaps the outcome seemed fairly clear throughout, despite a match going to five. I’ve been playing with limiting the list to matches where the pre-match gap in ratings was small, but I’m still experimenting with that.

Anyway, without further ado, here’s the list (ranked by z-score):

	Winner	Loser	Tournament	Year	Round	Score	z	Winner Elo	Loser Elo
1	Bjorn Borg	Ivan Lendl	Roland Garros	1981	F	6-1 4-6 6-2 3-6 6-1	15.1	2901	2402
2	John McEnroe	Bjorn Borg	US Open	1980	F	7-6 6-1 6-7 5-7 6-4	14.5	2251	2551
3	John McEnroe	Jimmy Connors	US Open	1984	SF	6-4 4-6 7-5 4-6 6-3	14.1	2462	2288
4	Novak Djokovic	Roger Federer	US Open	2011	SF	6-7(7) 4-6 6-3 6-2 7-5	13.9	2488	2244
5	John McEnroe	Jimmy Connors	US Open	1980	SF	6-4 5-7 0-6 6-3 7-6	13	2289	2328
6	Bjorn Borg	Jimmy Connors	Wimbledon	1981	SF	0-6 4-6 6-3 6-0 6-4	13	2643	2583
7	Rafael Nadal	Novak Djokovic	Roland Garros	2013	SF	6-4 3-6 6-1 6-7(3) 9-7	12.9	2534	2438
8	Novak Djokovic	Andy Murray	Australian Open	2012	SF	6-3 3-6 6-7(4) 6-1 7-5	12.8	2362	2228
9	Andy Murray	Novak Djokovic	US Open	2012	F	7-6(10) 7-5 2-6 3-6 6-2	12.7	2203	2374
10	Bjorn Borg	John McEnroe	Wimbledon	1980	F	1-6 7-5 6-3 6-7 8-6	12.6	2784	2382
11	Ivan Lendl	John McEnroe	Roland Garros	1984	F	3-6 2-6 6-4 7-5 7-5	12.5	2414	2502
12	John McEnroe	Mats Wilander	US Open	1985	SF	3-6 6-4 4-6 6-3 6-3	12.5	2420	2133
13	Bjorn Borg	Roscoe Tanner	US Open	1980	QF	6-4 3-6 4-6 7-5 6-3	12.4	2527	2011
14	Jimmy Connors	John McEnroe	Wimbledon	1982	F	3-6 6-3 6-7 7-6 6-4	12.3	2601	2509
15	Stan Wawrinka	Novak Djokovic	Australian Open	2014	QF	2-6 6-4 6-2 3-6 9-7	12.2	2110	2410

I found the list pretty striking. Here are some thoughts:

• I hope you’re a McEnroe fan because his name comes up a lot. He’s in 7 of the top 15 matches! The famous Borg - McEnroe 1980 Wimbledon final is on the list, but interestingly, the Borg - McEnroe US Open final of the same year which McEnroe won comes up slightly higher at number 2.
• Djokovic is popular too, appearing five times. I’m glad the Nadal - Djokovic 2013 French Open made it on the list; I remember watching that and just being in awe at the quality of tennis.
• This method really loves the early 1980s and 2010s: all matches are from one or the other of these eras.
• This model thinks Borg was a beast on clay in 1981. His Elo rating coming into the match against Lendl was 2901! Compare that against Nadal in 2013 against Djokovic, who “only” had a rating of 2534. It actually doesn’t seem so crazy when you look at Borg’s record on clay. This model takes the margin of victory into account and Borg was rarely dropping sets and regularly winning them 6-0. Coming into that match he must have seemed unbeatable, even though he ended up having a tough time against Lendl. That match was to be his last at a French Open.
• I think most people would agree that these are probably good matches. It’s perhaps more surprising to note which matches aren’t in the list, such as the 2008 Nadal - Federer Wimbledon final. I’ll note that this is just one approach, and many subjective choices determine which matches end up at the top.

Personally, I’m looking forward to checking out some of those McEnroe - Connors matches. I hope you get some ideas from reading the list too!

Links

1. Borg - Lendl 1981 RG: link
2. McEnroe - Borg 1980 US Open: link
3. McEnroe - Connors 1984 US Open: link
4. Djokovic - Federer 2011 US Open: I couldn’t find the full match sadly but here’s a chunk of it: link
5. McEnroe - Connors 1980 US Open: link
6. Borg - Connors 1981 Wimbledon (sadly fifth set only): link
7. Nadal - Djokovic 2013 RG (sadly highlights only): link
8. Djokovic - Murray 2012 AO: link
9. Murray - Djokovic 2012 US Open (highlights only): link
10. Borg - McEnroe 1980 Wimbledon (only a part of it): link
11. Lendl - McEnroe 1984 RG: link
12. McEnroe - Wilander 1985 US Open: link
13. Borg - Tanner 1980 US Open: I couldn’t find this, sadly!
14. Connors - McEnroe 1982 Wimbledon (highlights only): link
15. Wawrinka - Djokovic 2014 AO (highlights only): link