For this article, my plan is to detail how to create and use
an effective ‘script’ to trade a tennis match.
There has been a lot of interest in me creating a trading video, but I’m
not convinced live trading a tennis match totally suits a video format, due to
there being some wait for action on a regular basis. Therefore I thought it would be much more
useful to write an article on creating a script, which would form the basis of
what a trading video would cover.
It’s my belief
that having a script is highly useful for several reasons:-
1 – It ensures you have all statistical information to
hand prior to and during the match.
2 – It ensures that you don’t make impulsive decisions
without any logical reason.
When creating a script my first job is to assess the
projected holds of both players.
For this article, I’m going to use a hypothetical match
between Victoria Azarenka and Li Na on hard court at the Australian Open. It’s quite apt because they were the two
finalists at the upcoming Grand Slam, the Australian Open, in 2013. It’s also useful because they are similarly
ranked and there isn’t a huge ability difference. For the purpose of simplicity, I’m going to
assume that both players are equally fit and there are no match-up issues.
Projected holds for this match up were as follows:-
Azarenka 56.6%
Li 51.6%
These projected holds have also taken into account a
surface adjustment. In the 2013, the
average surface hold of the Australian Open was 61.5%, and when including 2012
data, was also 61.5%. This is below the
current WTA hard court average of 63.1% so it can be assumed conditions play a
little slow.
Both players have much lower projected holds than the WTA
hard court average, so from this data we can draw the conclusion that there
will be more breaks than average in this match.
Interestingly, in the 2013 final, that was indeed the case – there were
16 breaks in 29 service games (just 44.83% of service games were held).
It’s also worth noting that Li has much better break
point stats. She saved 60.4% to
Azarenka’s 58.4%, and converted 52.6% to Azarenka’s 50.5% in 2013 (across all
surfaces). This is borne out by her superior
break point ‘clutch score’ of 7.1, compared to Azarenka’s 1.5. A break point clutch score of over 3.0 is
considered pretty strong, and is definitely worth noting and taking into
account when compiling a trading script.
This data would give a starting price of around 1.80 on
Azarenka, which isn’t too dissimilar to her starting price of 1.71 in both the
Australian Open Final, and at the recent WTA Tour Championships in
Istanbul.
It’s also reasonable to assume that as Azarenka lost the
match in Istanbul, her price would lengthen a little for a match-up in the near
future, so the 1.80 on her looks pretty accurate. For the purposes of this article, I’m going
to assume that’s the case.
So as it stands, we have the following information:-
·
Starting prices are correct.
·
Both players are expected to hold serve much
less than average, especially Li.
·
Li should save break points more than average.
All this information is available for every match in the daily ATP/WTA spreadsheets.
On that basis, we can start to formulate a trading plan…
With both players expected to hold serve much less than average,
a good starting point would be to lay both servers for their individual service
games when the first set is on serve. At
the end of each service game we hedge our position, taking a profit if there is
a service break, or a loss if there is a service hold. Essentially this is a short term trade.
In normal circumstances, because her projected hold is
very low, it would be worth looking at laying Li’s serve for a higher stake
than Azarenka’s, but in this match-up her break point clutch score needs to be
considered. It would also make me
consider taking a more conservative approach to her service games. One way of doing this would be to take some
liability out at a scoreline such as 0-30 or 15-40, and definitely 0-40.
Several approaches can be considered here:-
·
You could clear all liability at one of these
scorelines, leaving all potential profit on Azarenka and a scratch position on
Li, guaranteeing profit (higher profit if Azarenka still breaks).
·
You could leave some liability on Li and more
potential profit on Azarenka, knowing that if Li does save break points to
hold, you could guarantee a pretty much scratch trade.
Because the markets on the exchanges generally base
themselves on the tour surface hold average when moving to a service hold or
break, finding players – especially those that the market may not necessarily
expect – who should struggle to hold serve is a valuable asset.
A further, more detailed approach, would be to use the ‘Rolling
Projected Holds’ newly available via the Tier Two Daily Spreadsheets.
My research found that there was a very strong
relationship between a player’s previous service game scoreline (e.g. hold to
0) and their next service game scoreline.
I don’t want to go too much into the percentages as a) they vary from
match to match according to a player’s base projected hold and b) I feel that I
need to keep that information for my subscribers, but I will say that projected
holds can vary by as much as 10% either way based on previous service game
scorelines.
Once there is a break of serve, prices will start to
deviate strongly from the starting price.
This has a more pronounced effect in ATP matches due to the men holding
serve more (hence a break is a rarer commodity) but still has a strong
influence in WTA matches. It’s very
difficult to give a price guide to how much this will change because time decay
(games elapsed) is a strong factor – for example the price will decrease
sharper if the break comes to give a player a *5-3 lead as opposed to a *2-1
lead, for example.
The next step at this point would be to assess whether
there is any viability in laying the player a break up in the set. At this point, unless the player who is a
break up was a fairly strong underdog to win the match before it started, the
player a break up will be trading odds-on (1.xx price). Laying players at 1.xx means that we generate
more potential profit than our potential liability, so it’s generally preferred to
laying players at prices odds against, although I don’t mind laying players
between 2.00 and 3.00 (occasionally up to 3.50) for individual service games.
The way I feel is best to assess whether the
player a break up can be laid is to see the break lead defence and break
recovery stats available via both the Ultimate In-Play Spreadsheets and also
via the new Tier Two Daily Spreadsheets.
By this I mean the set went back on serve from those points (it does not
take into account what happened after the set went back on serve).
In this match-up, the following break-back stats apply:-
·
Azarenka gave up a break lead 38.46% in 2013,
and recovered a break deficit 62.50%.
·
Li gave up a break lead 42.86% in 2013, and
recovered a break deficit 66.67%.
From this we can see that there isn’t much between the
two players in this area. Li gave up a
break lead slightly more, but compensated that by recovering more break
deficits.
If Azarenka was a break up, my approach would be to
assess the combined score for the scenario (Azarenka break lead loss % + Li
break deficit recovery %), which would be 105.13.
If Li was a break up, my approach would be to assess the
combined score for the scenario (Li break lead loss % + Azarenka break deficit
recovery %), which would be 105.36.
These stats are almost exactly the same, and come
marginally above the required 105 combined score to lay a player a break up as
detailed in the WTA Break Back Percentages article, so it’s viable to lay either player a break up in
this match, as a medium term trade. By
this I mean that I will look at keeping my position until either the player a
break up gets broken (I can then hedge for profit) or the end of the set is
reached (I can then hedge for a loss).
All these stats are available in the new Tier Two Daily Spreadsheets.
So far in the article we’ve looked at how you can use
statistics to gain a very workable edge in the tennis trading markets, using
basic projected holds when the match is on serve, and break lead/deficit stats when
a player is a break up in the set.
My next job is to illustrate how the Ultimate In-Play
Spreadsheet can be used to alter projected holds from early and late service
game stats, as well as taking positions at the end of sets based on player
tendencies.
At this point it’s probably worth clarifying the
scenarios covered by early and late games.
An ‘early service game’ is either of the first two service games of a
set for each player. A ‘late service
game’ is any service game from the point that at least one player has reached
four games in the set.
It goes without saying that early and late service game
stats are of great use – knowing which players start slowly or quickly, or
thrive or wilt under the pressure of the end of sets, is hugely valuable in the trading markets.
With trading the player a break up already covered, the
main way of using these early and late service game stats is to use adjustments
to the projected hold figures when the set is on serve.
In the hypothetical match-up between Azarenka and Li, we
have the following figures for the two players:-
Victoria Azarenka:-
12 month average service hold – 67.3%
12 month early service game hold – 67.5%
12 month late service game hold – 67.1%
Li Na:-
12 month average service hold – 71.5%
12 month early service game hold – 70.9%
12 month late service game hold – 70.4%
From those stats we can see that Azarenka’s stats barely
deviate from the start to finish of sets.
Li’s stats have a little more deviation, showing that she holds serve
1.1% less in late service games than average.
Subscribers of the Ultimate In-Play Spreadsheet will know that this
percentage isn’t a huge drop-off, with many players having far bigger issues at
the business end of sets – with some players holding over 10% less than their average
in this scenario!
Due to this slight deviation, the adjusted projected hold
stats for this match-up barely alter based on the early/late service game set
data.
Victoria Azarenka’s projected hold would move from 56.6%
to 56.8% in early games of sets, and 56.4% in late games of sets.
Li Na’s projected hold would move from 51.6% to 51.0% in
early games of sets, and 50.5% in late games of sets.
I don’t keep stats for the middle part of sets, but based
on these stats I probably wouldn’t be far wrong in assuming that Victoria
Azarenka holds 67.3% on average in middle parts of sets, and Li Na is pretty
strong in middle parts of sets, holding on average about 73.2%, which would be
worked out by: 71.5% + (71.5-70.9) + (71.5-70.4).
This deviation is definitely not enough to want to alter
our trading plan in any way for this match-up, but there are many cases where
it would. A very basic example would be
a player that has a projected hold around average for the match (hence it is
not viable to lay their serve on that basis) but is one of the players
previously mentioned that holds over 10% less than their average at the end of
sets. In that case it would make laying
their serve extremely viable at the end of sets, when it previously wasn’t at
other points in the set.
The final area I want to assess in this article is how to
use the ‘set percentages’ data in the Ultimate In-Play Spreadsheet to look at
how we can take a longer term position in a match at the end of a set.
For those that are new to trading, the end of the set is
a key point in the match – after the first set, a player has won 50% of the
sets they need to win the match, and their opponent needs to win both to win
the match (unless it’s a 5-set men’s Grand Slam match).
This situation means that the starting price on a player
will significantly shorten if they win the first set, somewhat depending on how
dominant they are in the process of winning that set. There is a more detailed explanation of this
in the TennisRatings Trading Handbook. The
WTA tends to have a slightly bigger drop in a player’s starting price when they
win the set, compared to the ATP, but the market is correct in this as WTA players
have a marginally higher win percentage when they win the first set than ATP
players. This may come as a surprise to
some readers, who may assume that the WTA is very unpredictable – and it often
is. However, much of this
unpredictability takes the form of in-set movements as opposed to anything
else.
Because a player’s price will significantly shorten when
they win the first set, a potentially viable position could be to lay that
player at the end of the first set.
A basic way of assessing that viability is to look at
whether the player a set up has a lower win percentage in the second set than
their average, and whether the player a set down has a higher win percentage in
the second set than their average. There
are far more advanced ways of doing it, such as using triggers based on first
set stats (which is something I have spent a lot of time researching), but for
this article I want to concentrate on the set percentages.
These percentages can hugely fluctuate from player to
player. My recent article on Petra
Kvitova showed that the Czech world number six won 68.6% of first sets, 52.2%
of second sets and 65.7% of third sets in 2013 (average set win percentage of
61.5%). These stats illustrated why so
many of her matches went to three sets (especially ones where she won the first
set) and she’d clearly be a player who it would be a good idea to look to lay when
she won the first set.
In this match-up between Azarenka and Li, the stats were
as follows:-
Victoria Azarenka:-
Won 75.6% of sets in 2013.
Won 71.2% of first sets in 2013.
Won 80.8% of second sets in 2013.
Won 73.3% of third sets in 2013.
Li Na:-
Won 72.1% of sets in 2013.
Won 75.9% of first sets in 2013.
Won 70.7% of second sets in 2013.
Won 61.5% of third sets in 2013.
From these stats we can see that Azarenka was a slightly slow
starter in matches in 2013, with a superior second and third set win percentage
than her first set win percentage. Her
second set win percentage was especially impressive, with it being 5.2% above
her average set win percentage, and 9.6% above her first set win percentage.
Li had the opposite problem, starting fast – winning 75.9%
of first sets. This was 3.8% above her
average set win percentage and significantly eclipsed her other set win
percentages.
With Azarenka winning more second sets than other sets,
and Li’s second set win percentage being lower than her overall mean, it’s
clear that laying Azarenka when she wins the first set is not at all viable,
based on those stats.
However, should Li take the first set, laying her would
be much more viable, based on the above stats.
Azarenka has an excellent second set record, and Li’s drops from her
strong first set win percentage. With Li
starting the hypothetical match-up at around 2.25, her price after the first
set would probably be 1.3x or 1.4x, so if Azarenka did take the second set, we’d
have created a very nice position.
If that did happen, the prices wouldn’t be hugely
dissimilar to the starting prices. The
market tends to make the player that wins the second set (to equalise the match
at one set all) a little shorter than their starting price, which is something I
don’t necessarily always agree with, as my research shows that momentum isn’t a
huge factor in this regard (the TennisRatings Trading Handbook has more
information and stats on this).
Roughly speaking, if we had laid Li at 1.3x or 1.4x at
the end of the first set, we could then hedge our position by backing her at around
2.3x or 2.4x generally at the end of the second set. This would be a huge tick gain.
There are other ways of approaching the trade. For example, those who like adopting an
approach with a little less risk may want to hedge or remove some or all
liability should Azarenka lead by a break in the second set (this would be an
especially good idea in this match-up based on the fact that Azarenka’s
combined score when a break up was over 105).
As mentioned above, at the end of the second set, prices
are generally not hugely dissimilar to starting prices. Unless a player is a very short price,
opening a new position here isn’t an approach I like, and again I’d need stats
to justify that move.
Whilst Li’s stats aren’t especially strong in deciding
sets, I’d want her deciding set win percentage to be below 50% to look to
oppose her at the start of a deciding set.
It’s worth mentioning at this point that there are plenty of players
that do have atrocious deciding set records compared to their records in other
sets or declining set win percentages as the match progresses (Benneteau and
Llodra are several in the men’s game, Date-Krumm and Pironkova are examples in
the women’s game).
However, I’ll be aware that Azarenka has a better record
than Li in deciding sets and might ease off laying Azarenka’s serve, or when
she is a break up, in a deciding set on that basis. Opposing Li’s serve, or laying her when she
is a break up, on a slightly heavier basis in a deciding set could also be
considered.
Hopefully by now you have a good idea of how to create a
tennis trading script based on statistics, and also a better idea of how the
stats in the Ultimate In-Play Spreadsheet can help you in your trading.
