Saturday, 15 March 2014

Is it game over if you lose more than 2 wickets in the powerplay?

I recently observed this conversation on twitter:

It immediately made me wonder if Aakash was correct. Do you lose if you are more than 2 wickets in the power play of a T20 International.

I decided to find out. I felt that it was probably best to only look at situations where a team had batted first, as there is not any external scoreboard pressure (or lack thereof) interfering with the batsmen's mind sets.

I looked at every match where there was a result inside 20 overs (I ignored matches that had ended in a super-over or bowl-off) and looked at how many wickets down the team were after 6 overs. I didn't count "retired hurt" as a wicket, despite there being a change of batsmen and the batting team losing momentum similar to when a wicket falls.

Once I did that I came up with some quite interesting numbers.

Wickets DownWinsLosesWinning %

It's fairly clear here that losing wickets early hurts the probability of winning. This is not really a surprise, often teams bat their best batsmen at the top, and the subsequent batsmen have to take fewer risks if there are not many wickets left above them. However while there are a lot of incidents of teams losing 1 or 2 wickets, our sample size is quite small for the other number of wickets. I've graphed it, adding in a 95% confidence interval. This indicates what range we can expect the actual winning probability to lie in per wicket loss: The shorter the line, the more reliable the data.

We can clearly see the trend here. But we also notice the huge gap between being 2 down and being 3 down. There does seem to be a difference between losing 2 wickets or losing more than 2 wickets.

Accordingly I broke it down into 3 groups. Less than 2 wickets, 2 wickets or More than 2 wickets. Here's how that looks:

Roughly teams win two thirds of the matches where they lose less than 2 wickets, half of the matches where they lose two wickets and about a quarter of the matches where they lose more than 2 wickets.

I also broke it down further by team, and this holds true for almost every team. The only team that has won more than half of their matches when batting first and losing more than 2 wickets in the power play is Ireland. (Interestingly Ireland has the 4th best winning record of any team batting first, and then they are not far behind Pakistan, Sri Lanka and South Africa).

Sri Lanka win just under 80% of t20's when they lose 2 or less wickets in the power play, but 20% when they lose 2 or more wickets. England win just over 60% if they keep their wickets in hand, but only 20% when they lose 3 or more in the power play.

With the World T20 getting underway, how the teams approach the first 6 overs could be a fascinating thing to keep an eye on.

Sunday, 9 March 2014

Who are the most reliable 6 hitters

I noticed that the ICC have set up a new game, where you need to pick a player who is going to hit a 6.

This is an interesting option, as there are not many stats out there for how reliable batsmen are at hitting 6's. We know how many 6's a player has hit, but how regularly they hit them is another issue. For example, Aaron Finch has hit 21 sixes in the 9 matches he has played in the last 2 years. However those 21 sixes came in just 4 innings. In the other 5 matches he didn't hit any. Once he gets going he really starts to pepper the boundary. In comparison, Ziaur Rahman from Bangladesh has hit 10 sixes in the 11 matches he's played in that time. However he's hit those 10 sixes in 6 matches, meaning there are only 5 that he hasn't hit a six in. In other words Finch has hit more sixes per match, but Rahman is significantly more reliable.

As the ICC game is about either hitting a six or not, the most important stat is their reliability, not their sixes per match.

To help out anyone who is playing that game, I've compiled a list of the 6 hitting reliability of players who had played 5 or more matches in the last 2 years. I've listed everyone who has hit a 6 in 40% or more of the matches.

If you want to join my league - here's the link.

PlayerMatchesSixesInnings with a 6P(hits a 6)
SE Marsh (Aus)55480%
DR Smith (WI)916666.7%
Yuvraj Singh (India)1121763.6%
MDKJ Perera (SL)1114763.6%
AM Rahane (India)54360%
SR Watson (Aus)1430857.1%
RR Patel (Kenya)1417857.1%
MJ Guptill (NZ)1415857.1%
HD Rutherford (NZ)79457.1%
MN Waller (Zim)75457.1%
Gulbadin Naib (Afg)1112654.5%
Ziaur Rahman (Ban)1110654.5%
MEK Hussey (Aus)118654.5%
Mushfiqur Rahim (Ban)1311753.8%
BB McCullum (NZ)1726952.9%
KA Pollard (WI)1725952.9%
DJ Bravo (WI)19181052.6%
MN Samuels (WI)1626850%
MJ Lumb (Eng)1213650%
R Gunasekera (Can)84450%
JL Ontong (SA)66350%
MW Machan (Scot)64350%
CH Gayle (WI)1526746.7%
DA Warner (Aus)1522746.7%
LMP Simmons (WI)1112545.5%
MR Swart (Neth)1112545.5%
DA Miller (SA)117545.5%
AD Hales (Eng)2018945%
AJ Finch (Aus)921444.4%
Ahmed Shehzad (Pak)1614743.8%
Mahmudullah (Ban)1412642.9%
Mohammad Shahzad (Afg)1411642.9%
Asghar Stanikzai (Afg)75342.9%
LJ Wright (Eng)1920842.1%
DT Johnston (Ire)129541.7%
Shakib Al Hasan (Ban)127541.7%
Mohammad Hafeez (Pak)25201040%
GJ Bailey (Aus)2016840%
F du Plessis (SA)1511640%
RS Bopara (Eng)1010440%
NJ O'Brien (Ire)52240%

Thursday, 6 March 2014

Who should win the NZ cricket awards

I was asked by Tony Veitch to put together some stats for the different awards on offer for the New Zealand Cricket Awards tonight.

I could have just brought up a list of averages, but that's really not the CricketGeek style, so I decided to delve into things a little more closely.

One of the difficult things in cricket statistics is to compare bowling success with batting success. For example, which is better taking 5/84 or scoring 172? We need a device to compare the two disciplines.

I decided to compare each player's year with the historical averages for their position. For example, for batting I compared the batting average with year end batting averages throughout history. I had a cut off of 10 innings, as making a cut off much higher than that excludes too many players, as most teams play less than 10 tests per year. I then compared a player's average to the historical average of averages, and the standard deviation of averages to generate a z-score. (For more on Z-scores, see This NFL blog post)

I used batting average and bowling average for test cricket, as really what we care about is scoring runs and taking wickets. I wasn't totally happy with the results, as there was no advantage for the players who had maintained a high standard over a number of games, rather than just one. (James Neesham, for example, averaged 171 this season, but only over one match).  I first filtered out anyone who hadn't either batted in 10 matches or who had bowled less than 100 overs. Then I multiplied the z-score by the square root of the number of innings that they had applied their skill in, in order to get a fairer list. It only caused a couple of positional changes, but the new lists looked more appropriate.

Here's the test lists.

Player - SkillAverageRanking
LRPL Taylor - batting81.6012.3
BB McCullum - batting52.735.0
TG Southee - bowling20.073.8
TA Boult - bowling22.363.6
KS Williamson - batting47.213.4
BJ Watling - batting42.272.0
N Wagner - bowling30.421.1
CJ Anderson - bowling30.541.0
CJ Anderson - batting32.70-0.3
TA Boult - batting32.25-0.4

I would give the award to Ross Taylor. He scored 816 runs at an average of 81.60. He past 50 in half of his innings. McCullum, Southee, Boult and Williamson all had great years, but Taylor's average really makes his numbers stand out.

Next I looked at the ODI lists.

Here I decided to use the batting and bowling index developed by S Rajesh from Cricinfo (and me separately). Again I compared the players index to the historical data.

Here's the list:

Player - SkillIndexRanking
CJ Anderson - batting 84.4816.1
LRPL Taylor - batting 43.776.9
MJ Guptill - batting 44.226.4
KS Williamson - batting 39.044.7
MJ McClenaghan - bowling 23.871.1
NL McCullum - batting 26.230.9
JDS Neesham - bowling 23.690.8
CJ Anderson - bowling 24.850.7
KD Mills - bowling 25.970.7
L Ronchi - batting 22.93-0.1

Again a batsman takes the title. This, however was not particularly surprising. Anderson was immense with the bat, and generally the games were played on high-scoring pitches, which don't really flatter bowling statistics.

For the T20 award I used batting index, but my own metric for bowling. In a previous post I showed how each wicket worked out to roughly 5 runs in a t20. Accordingly we can take 5 runs off a bowler's total for every wicket they have taken. They then get a modified run rate. I used this to compare the NZ players' years to the historical data. This is a little less relevant, as there is not a lot of historical data (about 1/10 the quantity of test and ODI information) and also New Zealand only played 6 matches, so the sample size is very small.

Here is the list:

Player - SkillIndex/Modified run rateRanking
L Ronchi - batting221.1114.7
BB McCullum - batting101.084.1
AF Milne - bowling2.752.9
AP Devcich - batting73.341.7
C Munro - batting60.041.5
JDS Neesham - bowling5.000.5
JD Ryder - batting44.020.0
NL McCullum - batting42.25-0.1
NL McCullum - bowling5.64-0.3
HD Rutherford - batting40.02-0.3

Luke Ronchi is a bit of a surprise here, but I remember looking up his stats and being surprised as to how effective he has been in t20s recently. During the course of the year he averaged 133 at a strike rate of 166. Those are quite ridiculous numbers.

The last major prize left is the Sir Richard Hadlee Medal, for the best overall. For me that goes to Brendon McCullum. He managed to attract the attention of the whole nation with his 300, and he also captained the side particularly well across all the formats. There would be a fair argument for Taylor and Anderson, but for me, McCullum needs to be acknowledged some how, and that award seems appropriate.

Who would you give the overall award to?

Sunday, 9 February 2014

Mini-session Analysis 1st Test NZvInd, Eden Park 2013/14

Here is the final mini-session analysis for the first test between New Zealand and India at Eden Park, Auckland

A mini-session is (normally) half a session, either between the start of the session and the drinks break or the drinks break and the end of the session. Occasionally a long session will have 3 mini-sessions where it will be broken up with 2 drinks breaks.

Monday, 13 January 2014

The McCullum Conundrum

There seems to be a regular pattern to cricket related chat in New Zealand. It often refers to Brendon McCullum.

A number of people want him dropped. They mention that he has only scored 1 hundred in the past 3 years. That he averages in the low 30's in test cricket and that he's too arrogant.

A roughly equal number of people want him to remain in the team, talking about how "form is temporary, class is permanent" and mentioning his successive fifties against England, his double century in India and his good captaincy.

While engaging in one of these fruitless (but fun) discussions, I noticed something surprising in his numbers. His away form has been horrible recently, but his home form has been remarkably good.

Here's what I mean. He has played exactly the same number of tests both home and away. In 41 tests at home McCullum has scored 2531 runs at a respectable average of 41.49. However, in 41 tests away he has scored 2153 runs at a much more pedestrian 29.90.

That's a fairly significant difference, but it wasn't always like this. Up until the end of 2010 his numbers were fairly similar. It's only since then that things have changed.

Home Away
Year Matches Runs Average Matches Runs Average
2004-2010 29 1669 37.93 27 1563 36.34
2011-2013 12 862 50.71 14 590 20.34

His career progress in terms of home and away is visually quite interesting.  Here are a couple of graphs to show the difference:

If we break it down by year we also see something interesting:

We can see that there has been a big difference in average per year at home and away from 2011 onwards.

This isn't due to the teams that he has played either.  In that time McCullum has played Zimbabwe, England, South Africa and West Indies both home and away, and has played against Pakistan at home, and Sri Lanka, Australia and Bangladesh away. There is no particular difference in the strength of his opponents at home and away in that time.

I also looked at a weighted 7 match average. This is where I look at an average for 2 games either side of a match, and 3 games either side of a match and average those two averages. It's a reasonable guide to the form that a batsman is in, and it normally covers 12-14 innings, so it is a reasonably reliable guide.  Again I broke it up into home and away.

We can see here that his career was tracking along similar lines for a long time, but then they started to split. The latest weighted average is at just over 57 for home matches, and just under 16 for away matches.

If McCullum was able to replicate his home form away from home, there would be no more questions asked about his place in the side.

I wondered if part of the difference was due to where he batted. He has moved around the order a bit recently, and wondered if he was more likely to play in positions that didn't suit him away than at home.

However I quickly checked his averages at different positions options, and found that there was a significant difference in his average at home and away for every group of batting positions. As an opener he averages 51.80 at home and 32.13 away. Batting from 3-5 he averages 37.40 at home and 30.57 away. In the lower order (6-9) he averages 41.26 at home and 27.72 away.

The question then remains as to what he can do about it. How does he become good at batting in conditions he can't practice in?

One option is to play a season of first class cricket as an overseas player.  He has played 3 matches in his career as an overseas pro, in 2007, and was quite successful, but picked up an injury doing so.

The difficulty is finding somewhere that takes overseas players, and plays their first class cricket when NZ is not playing, where the standard is high enough to be useful. Martin Guptill played second division cricket in England, and dominated, but that did not translate to his test game improving significantly.

Given that he's captain, it makes it more difficult for him to be able to find time to play in England also. But that does bring another question. Is the difference due to captaincy? Are the difficulties of captaining a side on tour affecting his game?

This is a difficult question to answer initially, as he wasn't captain at the start of his career when he was going well, but he also wasn't captain all the time when his away form started to dive. I first of all looked at the over all numbers, which are quite different. When he was not captain, he averaged 39.11 at home and 32.55 away. When he is captain his numbers are truly astounding. His home average is 59.85, while his away average is 15.18. However, if we just look at the matches where he was not captain since 2011, his averages are 44.30 at home and 23.50 away. It seems as if the impact of captaincy is exaggerated due to the fact that he is captain at a time where he is already struggling away from home. Also his away matches as captain have been against South Africa, England and Bangladesh. McCullum has struggled against left-arm spin at times, so Bangladesh, with 3 left arm spinners, is hardly the best tour option for him personally. (Just under half of his dismissals as a captain have been to left arm spin bowlers) Also England and South Africa are two of the more difficult places to tour.

If Brendon McCullum can find his form away from home, he is capable of becoming one of New Zealand's greatest ever captains. He has at his disposal probably the most solid batting line up New Zealand has ever produced combined with what is turning into one of the most lethal opening bowling attacks ever.  It may be worth playing a tour without him, in order to get him to play some serious first class cricket overseas and allow him to improve his away game.

Wednesday, 8 January 2014

Mini-session Analysis, 2nd test, Pakistan vs Sri Lanka, Dubai 2013/14

Here is the final mini-session analysis for the second test between Pakistan and Sri Lanka at Dubai International Cricket Stadium, Dubai, United Arab Emirates

A mini-session is (normally) half a session, either between the start of the session and the drinks break or the drinks break and the end of the session. Occasionally a long session will have 3 mini-sessions where it will be broken up with 2 drinks breaks.

Monday, 6 January 2014

Don't steal Corey Anderson's moment

Michael Jeh is a fantastic writer. When I see that he has written an article, I often read it.

Recently he wrote an article about the New Zealand - West Indies match in Queenstown, in it he suggested that Corey Anderson's innings was more a case of bad bowling than good batting. He even suggested that the game contained "possibly an unofficial record for the most full tosses bowled (including junior cricket!)"

I hadn't had a chance to see the innings before reading the article, and so I naturally assumed that Anderson's innings had involved him hitting a number of full tosses for 6. I was quite surprised, therefore, when I watched the highlights that I didn't notice a single full toss before he got to 100.

A few days later, I had a chance to sit down and watch the game closely, and actually see if Jeh's criticism was valid. Not every ball is shown on the highlights, so I wanted to be careful to not judge his article based on the work of the Sky editors.

After watching it I noticed a few things.

There are a few reasons I can think of why a bowler will deliver a full toss. Here is a list of some possible reasons:

  • Perhaps they decided that a particular batsman has trouble with full tosses. 
  • Perhaps they want to bowl a slower ball, and they know that slower balls are much more effective if the batsmen are attacking them. As a result a wide half-volley or a full toss often pick up wickets. 
  • A good tactic for spin bowler who sees a batsman charging down the wicket is to throw in a flat full toss. The batsman often ends up just hitting the ball straight back to the bowler.
  • Perhaps they just executed a yorker/full ball badly.
  • A bowler who has been hit a number of times sometimes just wants to get through their over, and doesn't focus as much on where the ball lands.
Some of these are a result of poor skills, but some of them are actually caused in reaction to the batsmen. It is important that we identify which is which before we criticize too harshly.

Jesse Ryder has a history of getting out to full tosses. It is not a good idea to bowl one every ball to him, but it is a valid tactic to occasionally bowl one to him, especially outside off stump, where he has a tendency to mistime them and hit them at catching height to cover. This is a risky tactic, and not one you would try every delivery, but it is a valid option occasionally.

Corey Anderson, however, doesn't have a reputation as a bowler who is likely to get out to a full toss. He is possible only behind James Franklin and Colin de Grandhomme in his ruthlessness at dealing with full tosses.

The first 5 full tosses were all bowled to Ryder. The first one was in the 9th over. There was not a single full toss in the first 53 balls. The first one was mistimed for a single. In the next over the West Indians bowled another. It was also mistimed for a single. A couple of overs later Ryder received two in a row. He hit the first for 4, but failed to score off the next one.

About 3 overs later Ryder received another full toss, and again managed only a single off it.

The first full toss that Anderson received was the ball immediately following him bringing up his hundred. There was not a single full toss in the 36 balls that he took to get to 101.

Ryder got 2 more full tosses. The first he managed to score 2 off and the second one dismissed him. Ryder scored 9 runs and was dismissed off the 7 full tosses that he faced. Off the other 44 balls he scored 95 runs. Overall Ryder scored at a higher rate off the balls that bounced than off the ones that didn't.

Anderson received 4 full tosses. He hit two of them for 6 and two of them for 2.

Overall the "unofficial record for the most full tosses bowled" is apparently 11. Only 2 of those 11 were in the first half of the innings. They were a result of good batting, putting the bowlers under pressure and getting them to go searching. 

They also didn't actually contribute that significantly to the overall score. Anderson and Ryder scored 25 runs off the 11 full tosses. This equates to 2.27 per ball. Off their other 87 balls they scored 210 runs, or 2.41 per ball.

The bowling performance by the West Indies may not have been the best ever, but the real story was the extraordinary batting. To focus on the bowlers bowing too many full tosses is to steal the glory that Corey Anderson and Jesse Ryder richly deserve. It is a disappointing angle for such a high quality writer to take, and makes me wonder if it would have been taken if it had been Warner or Dilshan scoring the runs.