The Clutch Factor

[Johnsteinbeck]

So Fed's Istanbul performance made think of something that i've been considering, and about which always wonder why noone points it out, since it's pretty simple and - if you ask me - quite telling (with the correct sample size) :

Assuming that in men's tennis, break points are usually the decisive points, on which the pressure is biggest, most viewers and commentators will turn to break point conversion/defense rates and wonder about the mental toughness of a player. however, it's only natural that better servers will have more success in defending break points.

For 2014, Isner, Karlovic and Fed lead the break point saved percentage for the tour. were they the most clutch?

so the real question is: on a break point, how much better/worse do you fare than you usually do on serve/return?
therefore:
(BrkPointsSaved% - SrvPointsWon%) + (BrkPointsWon% - RetPointsWon%) = ClutchFactor (CF)

(minor factors skewing those stats: i guess that significantly more than half of all break points are played serving to the ad side, which plays an important role, especially in *ahem* some match-ups *ahem*. Also, TB are of course ignored, and it'd be interesting to work those in somehow).


Looking at Fed for 2014, he saved 71% of break points he faced. He also won 71% of all his points on serve. hence, his servingCF is 0. on return, he converted 39% of his break points, while making 40% of all points in return - a returnCF of -1%, for a total of -1%

Isner, 2014: CFsrv=75%-72%=+3%; CFrtn=24%-30%=-6%. Total CF=-3%

which makes sense: on serve, Isner not only has the advantage of a great serve, he also uses it very good - probably because he knows how reliable a weapon he has. whereas on return, it's the negative that dominates. all in all, i think it's not surprising to say that he's not as 'clutch' as Fed.

now, where i get thrown off a little is Novak:
2014: CFsrv=-6; CFrtn=+2, total CF=-4.

alright, so 2014 wasn't Novak's best year. and what was said about Isner seems true here as well: with Novak, it's the return that is the most reliable weapon, so this is where he gets his boost, whereas the serve isn't as much. but he actually fares worse (-6%) in the (so far) stellar season of 2015. what could be said is that currently, Novak does not need to be "clutch" - he's just plain dominant on the court, so it doesn't matter that on break points, he doesn't do all that well.

now for Rafa (usually an example of clutch, right?):
2014: CFsrv=-1, CFrtn=+5, total CF=+4
so he's the first with a positive factor. not surprising, right? however, i'm not sure how much of that is due to the Ad-side-Factor working in favor of the lefty, which would seem to be an issue even on the return (just think about running around the bh for a fh).

for comparison:
Novak2011: 0 (-3/+3)
Rafa2012: total CF +7 (+2/+5)!
Fed2013: total CF -5 (-4/-1)!


of course, true clutch means winning important matches and tournaments. however, i think it's a good number to distinguish playing well on pressure points from simply dominating all the way, and maybe to determine "effectiveness". and i think it would be much more interesting to look at the Factor than just mere BP conversion rates, especially for a longer match or the course of a tournament. for example, what Fed displayed at Istanbul was anything but effective, and the Factor would've been very clearly showing why it took him three long matches in a row to get to that title even though he should have easily been the better player. or we could use it to see much more effective a player is in big matches - semis, finals, slams, to have another way of gauging their "clutchness" (as is visible here: Fed2014 is head and shoulders above Fed2013, and that was not just his overall level of play, but how he fared on the pressure points, especially on his own serve).

 

[Federberg]

This is an awesome thread. You'll out dude El Dude with this!

I've often wondered the same things actually. You've neatly encapsulated some of the flaws, but at a first step to trying to understand what it is to be clutch it's a good attempt. Probably better to think of this as conversion efficiency rather than clutch though. Purely because 'clutch' situations don't just occur around the serve - whether returning or serving. For sure this metric you've come up with helps, but there's more to it than that. I'll have a think and try to identify some of the other factors. At first glance, it's clear that this is not the key to success. You can absolutely dominate and not need to worry about clutch situations for example

 

[Murat Baslamisli]

JS, another thing to consider and throw in the mix regarding clutchness is the Tie Break record, where every point played on the serve is a break point faced . That would help the overall data, no?

 

[Johnsteinbeck]

thanks federberg, and yes, that's where i arrived in the end: it's more about efficiency than it is about clutch - but the two are somewhat linked of course

and yes Murat, Tie Break record is important, too; but it's also something where the better server usually has an advantage (strong servers usually top the list*; they also get to/have to rely on TBs more regularly, of course) and i wouldn't know how to get rid of that factor; plus, the sample size is somewhat small, with just 20something TBs for a full year for a lot of players. it'd be somewhat drastic (but not quite incorrect) to simply include Every point on TBs (they're not called mini-breaks for nothing); plus, they'd actually even out at least a little bit of that ad-side dilemma, and by that, you also could count points equally and wouldn't have to resort to any kind of "weighing", which is always bound to lead to some slant one way or another. however, i reckon that there's no usable stats with ALL TB points won/lost for each player.

also - Monfils leads the 52 week TB stat right now! who would've thought? and what does this say about the usefulness of the TB stat? his sample size is somewhat small, though.

btw, the match that rekindled my interest in this factor, Fed's Semi vs Schwartzmann: Fed was -22! (-8/-14; with 16 BPs (5/11), the sample was even big enough for this to be more than just an oddity).


*the FedExATP-PerformanceStatsThingy shows a few more indicators regarding "pressure" situations. if one would want to combine it to one big decisive "Clutch" number, they'd have to include some of those factors, like results in finals, of course. but that's where the weighing comes in again. anyway, i think the "clutch/efficiency factor" is just one more stat, but one that would be easily calculated and which therefore would be somewhat interesting to see a little more often.

 

[Puppet Master]

Excellent thread JS, you gave us some intriguing stats, but I think I have an answer for Novak's unexpectedly bad CF. On his own serve he can be very dangerous and troublesome to deal with, but he is just the type of guy that loses focus without any warining signs. I do admit he is the most consistent player on tour as of now, but think of how many games he just got broken out of nowhere. For example his match with Diego Schwartzman at the US Open 2014. He was dictating from the baseline, serving great, and out of the blue, Diego broke him, and managed to pressure him in a few other service games even though the match wasn't even close. But hey, if you are his fan you can't really complain about his focus now

 

[GameSetAndMath]

Actually, there is one more way to look at this thing. If a player A generates 10 break points
in a game and eventually converts in that game, his BP conversion ratio will be just 20%, but he
got the mission accomplished. In a recent interview (in Istanbul) when asked about his low
BP conversion ratio, Roger himself said that there is more than one way to look at these stats.
A low conversion rate can be considered negative, because you converted very few times.
On the other hand, it could be considered positive, because you generated so many chances
for yourself.


Similarly, even if a player has to defend a lot of break points,
his mission is accomplished as long as he eventually wins the game (not to mention that this
will psychologically demoralize the opponent and will typically yield a quick break to player
B in the very next game when A serves).

So, what is probably more important than the break point conversion rate would be the
break game conversion rate. Here the break game conversion rate is defined as the percentage
of games in which you had a break point (one or more) and actually won that game. Unfortunately,
nobody is maintaining that stat to crunch out numbers. But, it is this one which has more effect
on the outcome of the match.

 

[mrzz]

^Good point GSM, but as you said it is another way to look at it. The simplicity and effectiveness of JS's formula is its power. One of the most telling tennis statistics I've ever seen.

But if I had such a powerfull moniker I would have taught of it too...

 

[Johnsteinbeck]

^ thanks

GSSM - yes, i agree that it's one way to look at it (and btw, taking in at generated BPs would be a way to even out the Ad-dilemma; i actually think the problem might even be worse for a right-hander with a 'weak' bh return - he'll create a lot of BPs on the deuce side, only to have his weakness a little more exposed when returning on the ad side. but that's probably just my fed-bias talking here).

however, the break game conversion rate is, as you said, not as simple to find.
and still, needing 3, 4 or 5 BPs for one break, is at the least, a sign that you're not very effective in getting to that finish line. so really, i think if you finally break after wasting 7 BPs, especially if you make like 45% of all the return points, i do think that it might a good sign for the state of your game (you're good enough), but not a compliment on your calm going into these points.

 

[GameSetAndMath]

^ thanks

GSSM - yes, i agree that it's one way to look at it (and btw, taking in at generated BPs would be a way to even out the Ad-dilemma; i actually think the problem might even be worse for a right-hander with a 'weak' bh return - he'll create a lot of BPs on the deuce side, only to have his weakness a little more exposed when returning on the ad side. but that's probably just my fed-bias talking here).

however, the break game conversion rate is, as you said, not as simple to find.
and still, needing 3, 4 or 5 BPs for one break, is at the least, a sign that you're not very effective in getting to that finish line. so really, i think if you finally break after wasting 7 BPs, especially if you make like 45% of all the return points, i do think that it might a good sign for the state of your game (you're good enough), but not a compliment on your calm going into these points.

Your last sentence is indeed a valid one. What you are trying to measure is whether a
player can do clutch play at crucial moments.

However, people tend to think that all break points are the same and all are clutch situations.
IMHO, that is not really the case. There is a fundamental asymmetry here. Say, player A is serving and digs himself into a 0-40 hole. All of the next three points are clutch points for player A
as if they miss one of them, they lose the game. On the other hand, the same cannot be said
for Player B who is returning. At that score, there is no reason why player B should indulge
in clutch play when they can well afford to waste a couple of break points. Also, typically under
such circumstances it is often best to put the ball in play (rather than force the issue when
there is not a good opportunity in the play) and the let the opponent make a mistake (which
they will be prone to due to pressure).

 

[Johnsteinbeck]

^ agreed with most of that. except for "can well afford to waste a couple of break points". that might be true when Novak is up 6-1 5-0 against, i dunno, Rui Machado. but in the reality of most ATP matches, you better take your moments when they come. more than enough times i've seen people squander opportunities at 0-40 only to lose the game and get broken right in the next game.

your idea of a "break game conversion rate" would actually adress that, it was just your wording that i thought needed a little context . oh, and neither the break points nor the break game of my 6-1, 5-0 scenario would be "clutch" for either player; but that's the nature of statistics, of course: for the most part, break opportunities are pressure situations, so on the whole, they give us an indicator. of course, numbers can never tell the whole story.

 

[mrzz]

^Well, remember that you broke the factor in an CFsrv (serving) and CFrcv (receiving). The one who's serving can never afford nothing, so only the receiving one would be affected by the above discussion.

But, repeating what have already been saind, that's statistics: you put a lot of information down to one or a few numbers. You are bound to lose granularity. If you want detail, watch the match and take notes...