Wednesday, 22 January 2014

An Eye Opening Insight Into The Type Of Gambler You Are

It’s unlikely that you have ever wondered whether you are more like Fox or a Hedgehog. Important scholars stretching back to the Grecians have, however, used this animal analogy to characterise important traits that illustrate the way we think and impact our ability to successfully predict. One man’s crusade to measure predictive ability and relate to the Fox-Hedgehog dichotomy led him to insights that gamblers should find eye-opening.

Honing your betting skills is essentially the pursuit of more accurate predictions. This is a challenge that transcends gambling, and has huge implications for spheres such as finance, civic planning and politics, and essentially relates to how we think.

The Greek poet Archilochus suggested that ‘foxes know many things and the hedgehog one big thing’. Several important thinkers since have expanded on this concept to suggest that the way people think can generally be characterised as being Fox or Hedgehog like.

“Who the hell wants to hear actors talk?” — H. M. Warner, Warner Brothers, 1927.

One of the great difficulties in measuring predictive ability is that those same fields where the implications of the accuracy of predictions are so profound – think poor Intelligence and the Iraq war – accountability is rare, or very difficult to pin down. One man however, has tracked predictions for over two decades, exploring what constitutes good judgement and utilising the Hedgehog vs. Fox distinction. His insights make fascinating and hugely pertinent reading for gamblers.

Philip Tetlock spent 20 years recording the predictions of government officials, professors, journalists and politicians, and discovered that from over 28,000 predictions they were only slightly more accurate than chance. His work and approach are summarised in his 2005 book: ‘Expert Political Judgement? How good is it? How can we know?’

Tetlock developed an array of calibrations and adjustments in order to be fair to those making predictions, and his results were shocking, effectively suggesting that (as a whole) experts were only marginally better than chance.
Rather than writing off all forecasters, Tetlock was able to distinguish characteristics that identify someone as being better suited to making more accurate predictions, and these are equally valuable whether you are trying to make a complex policy decision or trying to consistently predict sporting outcomes.

Tetlock’s approach was to avoid looking at specific successes: how many times have you seen tipsters and talking-heads trying to live off the glory of rare headline grabbing predictions? Instead, he gave more credit to consistent predictive success over time and in differing contexts.

Success wasn’t reduced to a yes or no assessment, because prediction is as much about correctly predicting future events as it is the speed with which you recognise you have got things wrong and subsequently adjust your belief.

It does Tetlock a disservice to summarise his work in a few words, but for the purposes of aspiring gamblers the key take away is to focus on thinking the right way.

Nate Silver provided a useful summary table that outlines the important character traits that Tetlock’s work discovered:
Fox-like characteristics
Hedgehog-like characteristics
Multidisciplinary – Incorporates ideas from a range of disciplines
Specialised – Often dedicated themselves to one or two big problems & are sceptical of outsiders
Adaptable – Try several approaches in parallel, or find a new one if things aren’t working
Unshakable – New data is used to refine an original model
Self-critical - Willing to accept mistakes and adapt or even replace a model based on new data
Stubborn – Mistakes are blamed on poor luck
Tolerant of complexity – Accept the world is complex, and that certain things cannot be reduced to a null hypothesis
Order seeking – Once patterns are detected, assume relationships are relatively uniform
Cautious – Predictions are probabilistic, and qualified
Confident – Rarely change or hedge their position
Empirical – Observable data is always preferred over theory or anecdote
Ideological – Approach to predictive problems fits within a similar view of the wider world

The Fox-like approach is an agile one, incorporating changing circumstances to refine and adjust your predictions. Anyone who applies Bayesian analysis will quickly recognise the connection.

Bayesian theorem uses an iterative process of assessing what you know about the probability of a future event, then tests the impact of new evidence as it becomes available. Bayes was an 18th century English Presbyterian Minister, but almost certainly a fox.

Of course a Fox-like approach doesn’t imply infallibility. Getting things wrong is inevitable, the key thing to use an approach that maximises your chances of getting things right

Which animal characterises your way of thinking about uncertainty? It may help you improve your betting.



Monday, 13 January 2014

Are Big Servers Handicapped In Grand Slam Tennis Tournaments?

This article looks at the lack of success for ‘big servers’ in the men’s five-set Grand Slam format by focusing on their recent records and why this may be. We also analyse the players ranked 11-25 with a view to using statistics to create viable betting opportunities in the forthcoming Grand Slam.

It’s worth mentioning that the table below illustrates that only one player currently ranked 11-25 – Jerzy Janowicz – has reached the 6th round (semi-final) of a Grand Slam and this gives further weight to the general consensus that Grand Slams strongly favour the top players.

The ATP service hold average for the previous 12 months is currently 78.5% at the time of writing, and in the table below showing the data for the players ranked 11-25, we can see there is a large discrepancy between the players, with six players holding serve 79.5% or below, and nine players holding serve 82.7% or above (no player held serve between 79.5% and 82.7%).

When taking into account service breaks of opponent, clearly the average is 21.5% (100-78.5) and there are five players who break their opponent’s serve less than 20% and hold over 82.7% – Milos Raonic (15.5%), John Isner (12.2%), Kevin Anderson (17.7%), Jerzy Janowicz (17.4%) and Philipp Kohlschreiber (19.8%). Bearing this in mind, it is clear that these five players are the big servers ranked 11-25.

What the stats suggest on big servers
Between 2011 & 2013, these five players had the following Grand Slam records (when ranked within the top 40):
For clarity, only matches where at least one set is completed are included, and all prices used are Pinnacle Sports closing prices.

Player
Matches
Wins
Best Win/Odds
Worst Defeat/Odds
Profit/Loss (based on level £100)
Milos Raonic
27
18
P. Kohlschreiber (1.677)
M. Berrer (1.181)
-£534 (-20.54% ROI)
John Isner
25
16
R. Stepanek (2.462)
A. Falla (1.135)
-£161 (-6.44% ROI)
Kevin Anderson
28
17
M. Raonic (2.337)
M. Baghdatis (1.404)
-£486 (-17.36% ROI)
Jerzy Janowicz
13
9
R. Stepanek (1.837)
M. Gonzalez (1.084)
£2 (0.08% ROI)
Philipp Kohlschreiber
25
16
J. Isner (4.603)
D. Istomin (1.274)
£215 (8.60% ROI)*
* Kohlschreiber’s two biggest wins were both against Isner (4.603 and 3.600) and without these matches against a fellow ‘big server’ his record would have showed a loss. Furthermore, it can be seen that Raonic, Anderson and Kohlschreiber’s best wins were against one of these five players.

These figures illustrate how poorly big servers perform in Grand Slam events. Only Phillip Kohlschreiber (notably the player with the best break of opponent percentage) recorded a win when priced over 2.50 between 2011 and 2013, and the combined records are as follows:

Player
Matches
Wins
Profit/Loss (based on level £100)
Combined Player Stats
118
76
-£964 (-8.17% ROI)

Why is this the case? The table below provides some answers…

Player
Service Hold %
Opponent Break %
Match Win %
Sets per match*
Games per set*
Games per match*
7-5+ set per match %
5 Set Match %
Best Finish (Round)
Round Eliminated*
Raonic
91.3
15.5
18
3.52
10.17
35.78
30.53
10.71
4
2.80
Haas
83.4
23.7
9
3.57
10.36
37.00
36.00
21.43
5
2.80
Almagro
84.9
24.7
26
3.76
10.29
37.63
28.78
21.05
5
3.17
Isner
90
12.2
16
4.08
10.89
42.73
40.20
34.62
5
2.60
Youzhny
79.5
24.9
24
3.78
9.59
36.22
23.53
25.00
5
3.00
Fognini
72.6
30.5
3
3.43
9.71
33.39
33.33
0.00
3
1.75
Nishikori
78.7
29.2
16
3.48
9.10
31.65
17.50
13.04
5
3.29
Robredo
78.7
26.8
10
3.57
9.44
33.71
22.00
21.43
5
3.50
Simon
73.9
27
20
4.10
9.77
40.10
27.73
36.67
4
3.00
Anderson
85.6
17.7
17
3.64
9.90
36.07
27.45
17.86
4
2.55
Janowicz
82.9
17.4
9
3.58
10.14
36.33
27.91
15.38
6
3.25
Kohlschreiber
84.4
19.8
16
3.71
10.04
37.25
33.71
20.83
5
2.89
Dimitrov
83.7
21.4
3
3.80
9.84
37.40
21.05
33.33
3
2.00
Gulbis
82.7
25.8
3
3.50
10.24
35.83
33.33
14.29
3
1.75
Seppi
76.7
23.1
14
4.14
10.14
41.95
36.26
47.83
4
2.56
11-25 Mean
81.93
22.65
64.15
3.65
9.99
36.45
29.31
23.58
4.4
2.73
Big Servers Mean
86.84
16.52
63.33
3.72
10.25
37.75
32.48
20.00
4.8
2.82
* Stats based on the player average

The first area to look at is the amount of sets per match. The data from this was inconclusive, with Isner playing more sets per match than all but two players (Gilles Simon and Andreas Seppi) but only Fabio Fognini, Kei Nishikori and Ernests Gulbis playing less sets per match than Milos Raonic. When the five players were grouped together, they averaged 3.72 sets per match, marginally above the 3.65 set average for this 11-25 ranked bracket of players.

However, it can be seen that there is a very clear bias for big servers in the games per set area. Only Kevin Anderson (9.90) was below the 11-25 mean of 9.99, with Isner’s games per set the highest at an incredible 10.89. The five ‘big servers’ averaged 10.25 games per set on average, which is a fair bit above the 9.99 mean.

The effect of this is fairly obvious – as ‘big servers’ play more games per set, their sets are tighter. Tight sets mean that even smaller margins than usual decide the winner of a set (such as a tiebreak) and require more mental strength and concentration than straightforward sets.

We can see that Raonic, Kohlschreiber and especially Isner all play tighter sets (7-5 or bigger) more often than the 11-25 rank average of 29.31%. Isner’s 40.20% is considerably higher than any other player. The five ‘big servers’ grouped together have an average of 32.48%, 3.17% higher than the 11-25 rank average.

Looking at sets per match, the data somewhat surprisingly highlights that the ‘big servers’ do not play significantly more games per match than the 11-25 rank average. The five players combined played 37.75 games per match, compared to the 11-25 rank average of 36.45 games per match. Indeed, Raonic, Anderson and Janowicz actually played fewer games per match than the average.

Evidence shows big servers struggle in Grand Slams
On that basis it can be concluded that the tight nature of a ‘big servers’ match may hinder them from succeeding in Grand Slams, and because these players require more intense concentration and mental strength for longer than other types of player, the fatigue that this generates may affect them in the rounds following a victory.

There can be no doubt that Isner is an extreme example – he played the most games per set and the most games per match by a considerable distance in the 11-25 ranking bracket, as well as the tightest sets by some margin. The accumulated fatigue for Isner as well as the tight nature of his matches, explains why the giant American has failed to go further than a Grand Slam quarter-final.

Finally some other players are worth considering – it is clear from the stats that both Tommy Haas and Andreas Seppi have tendencies to play tight, and in the case of Seppi especially, long Grand Slam matches.

The opposite can definitely be said for Kei Nishikori (9.10 games per set and 31.65 games per match), Tommy Robredo and Fabio Fognini. It can be no coincidence that these three players broke their opponents more than any other analysed.

As with any form of betting, detailed analysis of a player and their historical tendencies are vital to achieve success. Men’s Grand Slam tennis tournaments are a much different betting proposition to normal ATP events and must be treated as such.