Over/Under 2.5 Goals Betting Based on Managers

When it comes to managers, it is pretty well-known that different managers have different styles. There are those managers that like open attacking football and aim to score one more than their opponents, regardless of how many their opponents score, while others prefer to keep it tight at the back and grind out low-scoring wins. Obviously, bookmakers know this just as well as the punters do, but I thought it would be interesting to look at whether it would be possible to look at particular managers and back either over or under 2.5 goals on a regular basis and make a consistent profit.

In this article, I am going to focus on La Liga in Spain. Using data from football-data.co.uk, I have taken every match since the start of the 2012/13 season up until the end of the 2015/16 season, so a total of four seasons of data. There are 34 managers that have managed for at least a season (38 games) in this period, so let us focus initially on those.

Just two managers have managed for the entirety of the four seasons or 152 matches - Atletico Madrid's Diego Simeone and the former Rayo Vallecano and now Granada manager, Paco Jemez. As a contrast in managerial playing styles, there could not be a more divergent pair. In those 152 matches, Diego Simeone's matches have averaged 2.47 goals per game compared with 3.16 for Paco Jemez. 84 of the 152 matches that Diego Simeone has managed in this period went under 2.5 goals, while 95 of the 152 matches for Paco Jemez went over 2.5 goals. We would expect the bookmakers to know this though and adjust the odds. However, did they adjust the odds enough?

It would seem that the answer is no. Had you backed Under 2.5 goals in every match that Diego Simeone had managed in La Liga since the start of the 2012/13 season, you would have made a 9.0% profit over the four seasons. Similarly, had you backed Over 2.5 goals in every Paco Jemez match, you would have made an impressive 15.0% profit in the same period.

Looking in slightly more detail, the vast majority of the return for Diego Simeone actually came in away matches. Simeone has actually seen over 50% of home matches go over 2.5 goals during this period and backing Under 2.5 goals would only have returned 2.9% in home matches. However, away from home, it is a different story. In these matches, backing Under 2.5 goals would have returned a huge 15.1% profit, which suggests that the bookmakers have consistently underestimated Atletico Madrid's tendency to grind out low scoring results away from the Vicente Calderon.

In contrast, backing Over 2.5 goals in matches involving Paco Jemez seems to show little difference whether the team is playing at home or playing away. At home, it would have returned an 18.2% profit, while away from home, it is slightly lower, but still an impressive 11.8%.

Further down the list, a couple of names stand out. The former Villarreal manager, Marcelino, would have returned 15.1% from Under 2.5 goals, driven by low-scoring games both at home and away, while Fernando Vasquez, Joaquin Caparros and Eduardo Berizzo have also shown profits of above 10% from the Under 2.5 goals selection.

In contrast, despite regular short-odds on Real Madrid scoring more than 2.5 goals, backing Over 2.5 during Carlo Ancelotti's reign would actually have returned 10.6% profit with 59 of 76 matches going over the line, while Unai Emery would have returned 7.7% profit from Over 2.5, driven particularly by matches at home.

Fran Escriba is a particularly interesting case. The former Elche and Getafe manager, who is currently at Villarreal appears to be nothing notable when you look at the overall figures. Indeed, a very small loss on both Over and Under 2.5 goals almost suggests that his matches are priced up accurately. However, if you split it down to home and away matches, there is a big difference. Backing Under 2.5 goals in matches than Escriba has managed at home would have returned a massive 31.0%, but away from home, backing Over 2.5 goals would have returned a similarly huge 27.2%. It seems that he oversees high-scoring away matches, but keeps things very tight at home. It might be one to watch at Villarreal this season.

This is obviously pretty basic analysis and it does not take into account anything apart from the manager. However, it throws up a few interesting angles that it might be worth thinking about when it comes to looking at the Over/Under 2.5 goals market in Spain.

The full table for all managers with 38+ matches is available here. If you want information on any other managers from Spain during this period, I have the stats, so just let me know...

How do players approach break points on their own serve?

Tennis is a sport that is often decided by the finest of margin and many times, we will see a match where there are barely a handful of points that separate the two players. However, some points are clearly more important than others and break points are some of the most important points of all. I thought it would be interesting to look in more detail at how players deal with break points on their own serve and how they go about trying to save them.

The data for this comes from the Match Charting Project at TennisAbstract. Now, while this is not a complete record of all matches for all players, for a growing number of players, there are enough matches charted to be able to start to look in more detail and begin to draw some conclusions. In total, there are 15,885 break points across 409 players and 2,475 matches, which is certainly enough of a sample to get going with.

The first thing to focus on is the first serve. Obviously, all players tend to win a higher percentage of points behind their first serve compared to the second serve. On a non-break point, the average first percentage of first serves in is 61.5%, which is higher than the 59.6% on break points, which is not really that surprising. We would expect the added pressure on break points to mean that players miss slightly more first serve. However, let us look in slightly more detail on a player-by-player basis:

The table shows the change between a player's 1st serve percentage on a break point compared to a non-break point. We can see that David Goffin has the biggest difference, increasing his 1st serve percentage by 5.9% on a break point compared to a normal point, followed by Juan Martin Del Potro, John Isner, Thomaz Bellucci and Bernard Tomic rounding out the top 5.

At the other end of the scale, we see Tomas Berdych at the bottom, with his first serve percentage dropping by 11.1% on break points. He is joined at the bottom by Viktor Troicki, Lleyton Hewitt, Fabio Fognini and Richard Gasquet.

What is interesting to look at in conjunction with the first serve percentage is the change in those points actually won behind the first serve. Here, we see a few interesting differences. We can see that David Goffin, despite hitting significantly more first serves on break point, actually wins far fewer points behind that first serve. Based on this, one might conclude that he looks to take a bit off his first serve and ensure that he gets the ball into the court and hopes that his superior ability in rallies will make up for the decrease in cheap points on serve. Indeed, we see that the percentage of aces and unreturned serves falls for Goffin, but so does the percentage of doubles faults that he serves. Basically, he goes for a very risk averse approach to serving on break point.

At the other end of the scale, we can look at someone like Viktor Troicki. We see that his first serve percentage drops by 7.3%, but he actually significantly increases the percentage of those points that he actually win. It seems as though Troicki is willing to risk going for a big serve on break point at the expense of missing a few and we do see that his ace and unreturned serve percentage does indeed increase on break points by 1.5%.

So, let us now look at how players win break points on their own serve once they are in a rally:

While we hypothesised earlier that David Goffin was very risk adverse when it came to getting the ball in-play, we can see that he balances this out with a slightly more dominant strategy once he is in a rally. The percentage of points won by Goffin hitting a winner or forcing an error increases by 6.3% on a break point compared to a normal point, suggesting that he looks to dominate the rallies slightly more and look to decide the point himself.

We also see Lleyton Hewitt, who appeared to look to go for a bit more on his first serve, presumably with the confidence that he would fancy himself in a rally if necessary on his second serve. We can see that Hewitt really boosts his percentage of winners on break point, suggesting that he prefers to decide break points himself, rather than rely on his opponent.

At the other end of the scale, we see Roberto Bautista-Agut and Gilles Simon, who appear to prefer to simply get the ball into play and wait for the error from their opponent, rather than risk going for the big shot themselves. However, it appears that the two players have very different success rates here. Gilles Simon sees a big increase in the percentage of points won via opponent unforced error, while Roberto Bautista-Agut actually sees a decrease. One wonders then whether Bautista-Agut takes this risk averse strategy too far and actually makes it easier for his opponent to hit plenty of winners on the big points, rather than giving them time to make mistakes.

We also see Roger Federer and Novak Djokovic at the bottom here - it appears that they both fancy themselves to outlast most players in a rally situation and are happy to cut down on any mistakes that they might make in exchange for waiting for their opponent to crack.

Finally, let us look at how players tend to lose points when we are in a rally on break point:

It is interesting to see that John Isner topping the list here. Isner actually increases his points won by aces and unreturned serves by 5.8%, but we also see here that he is able to reduce the percentage of points that he loses via unforced errors as well. It seems as though he goes big on the first serve to try and win the point early, but if that does not succeed, then he is happy to go more risk averse and make sure that he does not throw the point away himself with an unforced error.

The presence of David Goffin toward the bottom backs up our earlier theory that he plays conservative on the serve to ensure that he gets into a rally, then plays far more aggressive during the rally itself, increasing both his percentage of winners and unforced errors, which could easily increase as a result of looking to hit bigger shots.

One name that it is worth noting is Bernard Tomic. He appears near the top of all three tables - he hits more first serves, wins more of those first serve points, hits far more winners and hits fewer unforced errors on break points. That is a remarkable combination of statistics for a player that is not exactly renowned for his mental strength. Perhaps the pressure of break points concentrates his mind more?

On the flip side, Roberto Bautista-Agut does not come out of this looking great. He hits slightly more first serves, but a 10.2% decrease in first serve points won suggest that he takes a lot off the first serve. He hits fewer winners himself, but allies that with a big increase in unforced errors and also loses more points to opponent winners. It seems as though he simply plays far too passively and allows his opponent too many chances to hit winners and force the error.

Going forward, there are two areas that I intend to look at. Firstly, it would be interesting to look at serve placement in association with the first serve stats that we have seen already. One would imagine that those players that hit more first serves would hit more into the middle of the box as safer serves, but it would be interesting to see. The other thing is to look at how a player's risk profile translates to when they create break points on their opponent's serve. Do the risk averse players on their own serve also play conservatively on their own break point chances or are there difference?

Note: you can see the summary data for all players with 100+ break points faced here...

Looking at T20 Batsman-Bowler Combinations

Having looked previously at some of the top ranked T20 batsmen and bowlers in cricket based on my new ranking system, I thought it would be interesting to delve slightly deeper into a few of the players and take a more detailed look at how certain batsmen perform against certain bowlers, and vice versa.

Obviously, when we are looking at specific pairings of batsmen and bowlers, we are looking at relatively limited sample sizes, but it should still give us an idea of whether certain batsmen enjoy facing certain bowlers or whether they have particular bowlers that they struggle against.

Let us start off with the #1 rated batsman in T20 cricket - Chris Gayle. Being a player that has played a huge amount of T20 cricket over the past five years, there are actually no fewer than 38 bowlers in this period that have bowled 30+ legal deliveries at Gayle (as an aside, there are a grand total of 329 different bowlers that have bowled at least one delivery to the legendary West Indian in my database). The table below shows the highest strike rates for Gayle against particular bowlers:

Chris Gayle is a fearsome hitter of spin bowling, so it is no surprise to see a host of spin bowlers with some awful figures against him. Glenn Maxwell has conceded no fewer than 70 runs in 30 deliveries against Gayle, while Samuel Badree, the #1 rated bowler in the ICC ratings and #13 in my ratings, has also struggled, conceding a boundary every 2.33 deliveries. Dwayne Bravo has bowled a huge 79 balls to Chris Gayle, conceding 21 boundaries, but dismissing him on six occasions.

At the other end of the scale, Lasith Malinga has conceded just 29 runs in 53 balls at Gayle, which is really very impressive. Admittedly, he does benefit from bowling to Gayle early in his innings before he really gets going, but it is still hugely admirable.

Three spinners in particular here really stand out - Mohammad Hafeez, Sunil Narine and Ravichandran Ashwin. Given the way that Gayle can destroy spin bowling, the fact that the three of them combined have conceded just 102 runs in 144 balls at Gayle with just 9 fours and one six in those deliveries is very impressive. Whether Gayle has a real problem against those bowlers or whether he has the confidence in himself to just see off the star bowlers of the opposition, the data cannot tell us, but it is interesting.

So, if Chris Gayle struggles to score off Lasith Malinga, let us look at the batsmen that can. The table below shows the top 10 SRs of players to have faced at least 24 balls from Malinga:

In AB de Villiers, MS Dhoni and Shahid Afridi, we have a trio of three very destructive hitters at the end of the innings that have been able to score runs off Malinga, often mentioned in the debate about the greatest death bowlers in the history of T20 cricket. However, interestingly it is Marlon Samuels that tops the list, having hit a huge five sixes off Malinga in just 28 deliveries (all the other batsmen to have faced 24+ balls have 16 sixes combined off 956 balls). It was his memorable innings of 78(56) in the World Cup T20 Final in 2012 that did the damage here, when he took 39 runs off 11 balls from Lasith Malinga in what would be a match-winning innings.

So, we saw earlier that another bowler that Gayle struggled against is Sunil Narine, the #2 bowler in my ratings. However, he is far from the only batsman to struggle against Narine. Indeed, of the 29 batsmen to have faced at least 30+ deliveries from Narine, just 9 of them have a strike-rate of greater than 100.0. There are only four batsmen to have a strike rate of greater than 130.0 against Narine - JP Duminy and David Warner (both 133.3), AB de Villiers (140.0) and Suresh Raina (147.4).

There are actually no fewer than 7 of the 29 batsmen that Narine has over 50% dot balls against as well - Martin Guptill (51.5%), Dwayne Bravo (51.6%), Chris Gayle (51.9%), Marlon Samuels (53.1%), Naman Ojha (55.9%), Darren Sammy (57.6%) and Yuvraj Singh (64.4%). Interestingly, we find four of Narine's West Indian teammates in here, which suggests that facing him regularly in the nets does not seem to help in terms of being able to score runs off of him.

In Malinga and Narine, we have looked at two of the top T20 bowlers in world cricket. One name that has appeared in being able to score quickly off both of them is AB de Villiers, the #5 rated batsman in my rankings. Let us look at bowlers that he has scored particularly well against:

It is safe to say that there are more than a few bowlers that AB de Villiers has obliterated over the past five years. His South African teammate, Dale Steyn, is an interesting one - for a bowler than scored reasonably well in the economy part of the bowling ratings, he has been destroyed by AB de Villiers, conceding an extraordinary five sixes from just 19 balls.

It turns out that there are just six bowlers with 18+ balls at AB de Villiers to restrict him to a SR of less than 100.0. Indeed, Piyush Chawla has conceded just 27 runs from 30 balls against de Villiers, while dismissing him three times (joint-highest with Ashwin, Mathews and Balaji).

However, when it comes to destructive batsmen, there are few more devastating than Andre Russell. The table below shows the 11 bowlers that have bowled 20+ balls to Russell in the past five years:

With the exception of Sunil Narine, Andre Russell has a SR of upward of 150.0 against every bowler to have bowled regularly to him. Indeed, against five of the eleven, he has a SR of above 200.0, which is quite simply incredible. That Sunil Narine has only conceded a single boundary in 30 balls to Andre Russell just goes to show why he is undoubtedly one of the all-time great T20 bowlers.

Which T20 Batsmen Are Fast Starters?

In T20 cricket, getting off to a good start can be very important when it comes to setting a formidable total, but at the same time, different players need different amounts of time to really settle in before they can really start to play their shots.

In this article, I want to look at a selection of 20 different opening batsmen from seven different countries, all of whom have played plenty of T20 cricket across multiple different competitions over the past five years. In particular, I want to focus on the first 30 deliveries that they face. Obviously, the majority of T20 innings from opening batsmen are likely to last less than 30 deliveries and once a batsmen has reached his 31st ball, you would hope that he is in a position to really go on the attack, but we will look at that in more detail in a future article.

To begin with, let us get an overview of all 20 batsmen and their overall strike rate after each delivery in this 30-ball spell:

As we can see, virtually all the batsmen rapidly increase their overall strike rate over the first 10 balls that they face, at which point it begin to flatten out for the majority of batsmen. What this chart effectively shows is the score that we would expect a batsman to be on having faced a set number of deliveries. For example, with a strike-rate of 104.2 after 10 balls, we would expect Ahmed Shehzad to be on 10.42 runs, whereas with a strike-rate of 123.1, Aaron Finch would expect to be on 12.31 runs.

Obviously, the most aggressive batsmen also tend to have a higher risk of being dismissed, so in terms of an opening partnership, you might be looking to pair an aggressive batsmen that will get the scoreboard moving immediately with a slightly more gradual batsman, who may need a few balls to really settle before steadily increasing his strike rate.

If you have two slow starters, you potentially run the risk of finding yourself in a concerning position if they both fall after around 10-15 balls each at a run-a-ball and it increases the pressure on the players coming in afterwards.

So that initial chart is quite difficult to pick out individual players, so let us take a look at a few groups of players at a time. Firstly, we shall look at the four English players in the group - the current England opening pair of Alex Hales and Jason Roy, plus two former England internationals in Luke Wright and Michael Lumb:

We can see that the four players split into two pairings in this chart. The pair of Jason Roy and Alex Hales are both very fast starters, not only compared to the other English players, but compared to all of the opening batsmen in the sample. After five balls, Roy and Hales are 3rd and 4th respectively, but we can see that Luke Wright and Michael Lumb take a couple more deliveries to get moving.

Having said that, we can see that both Michael Lumb and, particularly, Luke Wright, increase their strike rate rapidly after the first couple of deliveries and it is also interesting to note that all three of Roy, Lumb and Wright are among some of the faster scorers through 20 deliveries. One concern for Alex Hales might be that his eventual strike rate is only around the middle of the pack, but at least he does reach that strike rate rapidly rather than eating up deliveries.

Next, let us look at the Indian batsmen. The immediate concern for India is that all their openers in this sample - Virat Kohli, Shikhar Dhawan, Rohit Sharma and Ajinkya Rahane - are all close to the bottom of the group in terms of strike rate. None of them are particularly rapid starters and none have a particularly high top gear in terms of making big scores in general. Admittedly, Rohit Sharma has hit a couple of incredible innings over the years, but those are not all that common.

The fact that none of the openers are able to constantly get India off to a flying start and with Rohit Sharma having the highest strike rate of the quartet after 30 balls at just 124.7, there is a lack of real dramatic acceleration, it puts a lot of pressure on the middle-order batsmen to score quickly from the start.

Chris Gayle have long been known as a relatively slow starter, but once he gets going, he accelerates well and importantly, continues to accelerate throughout the innings. Beyond the 30 deliveries in the chart, he continues to speed up reaching a final strike rate of an incredible 153.0. The concern here for the West Indies is finding an opening partner that can get the innings moving to give Gayle the time that he seemingly wants to get his eye in.

Neither Dwayne Smith or Lendl Simmons would appear to be the ideal foil for Gayle. In particular, Lendl Simmons is a very slow T20 opening batsmen and does not even break the 100.0 strike-rate mark until his 15th delivery. When partnered with Chris Gayle, this can lead to a very slow start to the innings, which then puts pressure on Gayle to convert his innings. With the big hitting further down the order, they can often get away with this, but finding a fast starting opening batsmen could improve them further. Time will tell whether Evin Lewis or Johnson Charles can be that player.

When it comes to fast-starters, there are very few in T20 cricket that are better than Aaron Finch. He is regularly able to get the innings off to a flying start, which gives his partner the opportunity, if needed, to build into his innings. The fact that he seemingly plateaus very quickly might raise the odd question about his stamina, but he is still a valuable opening batsman. He also makes a good foil for David Warner, who we can see starts slightly slower, but builds to a very good strike rate.

Another fast-starting opener, which will come as a surprise to very few, is New Zealand's Brendon McCullum. Now retired from international cricket, he is a very effective and fast-scoring opening batsman, who will undoubtedly be in great demand from franchises around the world, particularly given his ability either behind the stumps or in the field.

While he is pipped slightly over the first five deliveries, there is no opening batsman that can live with AB de Villiers. The South African is one of the greatest T20 batsman in history and the strike rate that he peaks at after 30 deliveries is streets ahead of his closest contender, Jason Roy. His rapid start is a perfect foil for his South African partner, Quinton de Kock, who is one of the slower starters in this sample, but he steadily builds to a very solid strike rate as his innings progresses.

The Effect of Different Grounds in T20 Cricket

Given all of the many T20 competitions that take place around the globe in the modern day, there are a huge number of matches taking places each year at many different grounds. While it sounds obvious, there are very few grounds that play similarly with pitches varying in preparation, boundaries being of different lengths and even different weather conditions affecting the score that teams should be aiming for.

In recent articles on both batting and bowling ratings, I have mentioned that I adjust expected runs by the ground at which the match is being played and in this article, I intend on looking in a little more depth at how different grounds vary in a number of ways. Even grounds at which the par score are the same can look quite different once you look in detail at how that par score might be expected to come about.

Firstly, a quick mention of the data used in this article. In my database of ball-by-ball information, there are just over 150 different cricket grounds that have hosted matches in the past five years - of these 150+ grounds, there are 96 that have seen at least 600 first innings deliveries (i.e. five full innings of deliveries), so we shall narrow the data down to these grounds for the analysis here.

So, let us first look at the average first innings scores at these grounds. The table below shows both the top 10 and the bottom 10 highest-scoring T20 grounds and the average first innings scores:

Now, most of the grounds at the top of this list are not likely to be too much of a surprise. The M Chinnaswarmy Stadium, home to the Royal Challengers Bangalore in the IPL, is renowned as a batting paradise, while Seddon Park was the venue for Richard Levi's record-setting T20 century back in 2012. Similarly, the likes of St Lawrence Ground, Eden Park and Trent Bridge are ground that are well-known to suit batsmen.

At the other end of the scale, we see an interesting number of grounds located in the West Indies. Providence Stadium in Guyana has an incredibly low average first innings score, while the Beausejour Stadium in St Lucia and Sabina Park in Jamaica also feature in the bottom 10.

Now, let us look at which grounds see the most sixes. Logically, we might imagine that there is a pretty strong relationship between high-scoring grounds and six-hitting. To an extent, this is true. The top five grounds in terms of balls/6 are Seddon Park (12.9 balls/6), Central Broward Regional Park Stadium (14.6), Eden Park (14.6), M Chinnaswarmy (15.0) and Wanderers Cricket Ground (15.0), all of which appear in the top 10 highest-scoring grounds.

However, there are a couple of interesting ones to look at. St Lawrence Ground in Canterbury actually only ranks 41st out of the 96 grounds in terms of balls per six, only seeing a six in the first innings every 24.0 deliveries, implying that there will be an average of five 6s in each first innings.

In contrast, we find the Beausejour Stadium, ranked 92nd out of 96 in terms of average score appearing at number 22 in terms of balls per six, seeing a six every 21.5 deliveries, while Sabina Park follows at 23rd in balls per six.

The table below shows the top 10 and bottom 10 grounds in terms of deliveries per six:

So, if there are a relatively low number of sixes being hit at St Lawrence Stadium in Canterbury, one might surmise that there are plenty of singles, twos and boundaries being hit and this does seem to be the case. We find that Canterbury shows up ranked 3rd in terms of balls/4, which could either be a reflection of the type of batsmen that Kent tend to select in their T20 team or the fact that there are long boundaries that it is hard to clear. It also appears ranked 1st in terms of the fewer dot balls at just 33.2%.

The next question is whether facing spin or non-spin bowling tends to make a difference at certain grounds. The table below shows a selection of seven grounds and the balls/6 against spin and non-spin bowling:

Being one of the biggest six-hitting grounds, it is no surprise to see that the M Chinnaswarmy Stadium is a ground where it is relatively easy to hit sixes off almost any type of bowling, particularly when you have the likes of Chris Gayle and AB de Villiers playing there regularly. Similarly Warner Park is a ground where sixes can be easily hit off all types of bowling.

The Dr Y.S. Rajasekhara Reddy ACAVDCA Cricket Stadium, a venue used by a number of IPL franchises over the past five years, is a ground where we can see a large discrepancy between the ease of six hitting off spin bowling compared to non-spin bowling. The Maharashtra Cricket Association Stadium in Pune is another ground where we see this discrepancy between spin and non-spin bowling.

The Brisbane Cricket Ground interestingly appears to be easier to hit big shots off the quicker bowlers - a trend that we also see at the Sinhalese Sports Club Ground in Colombo, which may reflect the fact that Sri Lanka are generally known for producing quality spin bowlers as opposed to quicker bowlers.

Finally, we see a ground like Grace Road, where it seems that it can be tricky to hit boundaries off any type of bowling with figures upward of 30.0 balls per six against both spin and non-spin bowling.

Who are the best T20 Bowlers in World Cricket?

Recently, I looked at how we could rank the best batsmen in T20 cricket, comparing two existing rating systems and developing my own system to compare and contrast with the ICC ratings and the Cricket Ratings systems (here). Obviously though, there are two crucial elements to cricket though - batting and bowling.

In this article, I want to take a look at the best bowlers in T20 cricket. As before, let us first look at the ICC, who have developed their own rating system to grade bowlers. Their current rankings are shown below:

As with the batting rankings, this only takes international matches into account and arguably this top 10 might cause more arguments than the batsmen. One of the hardest aspects of grading bowlers in T20 is deciding on the balance between the weights given to economy and wicket-taking. As the ICC mention about their rankings, "while taking wickets is still important in T20 cricket, the T20 rankings give more credit to a bowler for economy. A bowler that takes 0-15 off 4 overs gets more credit than one who takes 2-35.

This weighting toward economy is perfectly valid, but if you are only taking the pure figures, one would therefore expect bowlers that bowl during the powerplay at the start and during the death overs to be overly penalised by this system. Indeed, we see that the majority of the bowlers in the top 10 of the ICC rankings are spinners that will bowl most, if not all, of their overs during the middle overs.

So, let us move on and take a look at the Cricket Ratings system. As described previously, this takes domestic competitions into account and adjusts for the quality of competition in each different event to allow performance to be compared across different countries. It breaks down performance into two aspects, adjusted average and adjusted economy. The rankings are shown below:

Now, we see that there are only two bowlers that appear in the Cricket Ratings top 10 that also appeared in the ICC top 10 - Sunil Narine and Ravichandran Ashwin. Here, we see an increase in fast bowlers appearing toward the top of the list with Lasith Malinga, Mustafizur Rahman and Bhuvneshwar Kumar in particular rating highly. There are also a couple of unexpected names in this list. Benny Howell, Kevin O'Brien and Rikki Clarke are names that have performed well in the T20 Blast in recent seasons that might have gone somewhat under the radar.

It is interesting here that there appears to be equal weighting given to wicket-taking and economy, which is in contrast to the ICC system.

Now, my system is based on my T20 database that contains almost every T20 match played around the globe in the past 5 years for which there is ball-by-ball data. As with both the ICC and Cricket Ratings systems, my system takes two aspects into consideration - wicket-taking and economy.

The economy aspect is based around the number of runs that a bowler concedes compared with the expected runs that an average bowler would concede on that delivery. For example, off the opening ball of the match, the average T20 bowler would be expected to concede 0.78 runs, while if he were bowling the final ball of the final over, he would be expected to concede 1.72 runs. By comparing what a bowler should concede with what he actually concedes, we can compare economy rates regardless of when during the innings the bowler was actually bowling.

In addition, I adjust the expected runs based on the ground at which the match is being played and for the batsman that is facing the ball. For example, if you were bowling the first ball of the 19th over to Khaya Zondo at Queen's Park Oval, you would expect to concede fewer runs than bowling the same ball to Chris Gayle at the M Chinnaswarmy Stadium.

For the wicket-taking aspect, I again compare the actual wickets taken to the expected wickets, taking into account once again, the batsman and the stage of the innings. For example, you would expect to take 0.024 wickets with the first ball of the first over of the innings, but bowling the last ball of the innings, you would expect to take 0.134 wickets. Similarly, you would expect that to vary depending on whether you were bowling to Virat Kohli or to a #11.

As with the ICC system, I have decided to weight my ratings slightly more toward economy, rather than wicket-taking, simply due to the fact that the confidence interval for economy is narrower than wickets due to the relatively low frequency of wickets compared to runs. The top 10 from my ratings are shown below:

Based on my ratings, Australia's Mitchell Starc is the best T20 bowler in world cricket. Interestingly, Starc did not show up in either of the previous top 10 rankings, but I imagine that plenty of people would not be surprised to see him near the top of the list. He performs well in economy and he is one of the most dangerous wicket-taking bowlers in T20 cricket.

Another Australian, although less-heralded, Jason Behrendorff, appears second in my ratings, as he did in the Cricket Rating system. It would be interesting to see Behrendorff tested outside of the Big Bash and, at only 26-years old, there is still time for a franchise in another competition to take a chance on what would likely be a relatively cheap player.

There is likely to be little surprise to see Sunil Narine rounding out the top 3. His economy rating is simply outstanding and his low wicket-taking performance is likely partially affected by batsmen simply looking to play out his overs without taking any risks.

The trio of Benny Howell, Azeem Rafiq and Liam Dawson represent the T20 Blast strongly here and Dawson's performances have been rewarded recently with a T20 England debut. Benny Howell is a name that has appeared on both my system and the Cricket Ratings system and he is one that England should maybe keep an eye on. Strong performances in both the economy and wicket-taking aspects suggest that he is a very talented T20 bowler.

As with the batsmen ratings, let us take a quick look at the bottom 10 in the ratings. These are shown below:

This does not make excellent reading for India with both Ishant Sharma and Umesh Yadav, both of whom have played internationally for India and who are both regulars in the IPL, showing up in the bottom 10. Indeed, Ishant Sharma scores very poorly on both aspects, suggesting that he is expensive and rarely takes wickets.

Regarding a couple of the other names that showed up highly in the ICC ratings, we find Samuel Badree at #13 in my rankings, Jasprit Bumrah way down at #123, Imran Tahir at #59 and Kyle Abbott actually shows up a long way down at #232, well below the average T20 bowler.

Comparing a couple of names from the Cricket Ratings system, Lasith Malinga is at #23, Kevin O'Brien is at #51, Bhuvneshwar Kumar is at #57 and Ravi Ashwin is at #24. Generally, there appears to be slightly more consistency between my ratings and the Cricket Ratings rankings than with the ICC rankings.