Saturday 21 January 2012

Running between wickets

I have long been a believer that running between wickets is the most important thing in a limited overs cricket match.

More important than catching. More important than bowling, more important than hitting boundaries.

It is difficult to quantify this, but the focus on it has meant that when I look at statistics I look at them slightly differently sometimes.

One of the main things that I look at when assessing a batsman is their activity rate. This is the number of run runs (ie runs that did not come from boundaries) divided by the deliveries that they did not hit to the fence.

This statistic is a good guide to the form of a batsman. The ability to accumulate runs without really taking risks is really the sign of a classy batsman.

But I wanted to look at who was the best between wickets. This has in part been due to watching Azhar Mahmood playing for the Auckland Aces. He has batted very well, but his running between wickets (and fielding) has been awful. He scored a brilliant hundred against Canterbury, but there were about 6 or 7 times that he wasn't quick enough to get back for a second, and so the team missed out on a number of runs. (He also let in about 6 or 7 runs in the field).

He has been out run-out in about half of his innings in the HRV Cup, which got me thinking about accounting for a batsman's propensity to get run out when I'm assessing how good they are between wickets. It would be better again to check how often there is a run out when they are batting, because a bad call that runs your partner out is as bad as a bad call that runs yourself out. However that data is very hard to get, and so I have to use what I do have.

So I decided to work out a metric for running between wickets. I started with looking at who scored the most run runs per dismissal in ODI's (ignoring how they were out). A couple of the results were quite surprising:

(all figures here are over the last 5 years, in ODI cricket. I have only looked at batsmen who have faced more than 1000 deliveries in that time period)

Run runs per dismissal
NameMatchesRunsRun runsRun runs per dismissal
RN ten Doeschate (Neth) 27128275041.67
S Chanderpaul (WI) 632549165338.44
IJL Trott (Eng) 401798122234.91
TLW Cooper (Neth) 1886454033.75
MS Dhoni (Asia/India) 1374762289033.22
HM Amla (SA) 542705155532.40
JH Kallis (SA) 773154190431.21
MJ Clarke (Aus) 1123871253730.20
MEK Hussey (Aus) 1153408227429.92
JP Duminy (SA) 772273155729.38

I had to look up Tom Cooper, as I really knew very little about him. It will be interesting to see if he can keep that up as the Netherlands start to play more matches. From a quick look there are a lot of players that learned to play their cricket in the Southern Hemisphere, Dhoni and Chanderpaul being the exceptions.

Next I looked at the activity rates:

Activity Rates
NameRun runsNon-boundary ballsActivity rate
Shahid Afridi (Pak) 85712540.683
JP Duminy (SA) 155725200.618
MEK Hussey (Aus) 227436990.615
DL Vettori (NZ) 5699280.613
JR Hopes (Aus) 73011940.611
HM Amla (SA) 155526650.583
DJ Hussey (Aus) 5629670.581
MV Boucher (Afr/SA) 63611060.575
IJL Trott (Eng) 122221470.569
AB de Villiers (Afr/SA) 215638020.567

The name at the top of the list here was a big surprise. When you think of Shahid Afridi, running between wickets is certainly not one of the things that you think of. However he has scored extremely quickly from run runs. Looking at different time periods and in different forms of the game another couple of names that occur regularly are Prior, Pollard and Dilshan. All three are thought of more as big hitters, but are also very effective at turning over the scoreboard. (Although this may be partially due to opposition captains being happy to conceded a single to them.)

The next thing that I looked at was the runs per run out. There were a number of batsmen who were not run out once, so I couldn't provide an average. In the interest of comparison, I have counted them as having been run out once for the sake of the average.

Run runs per run out.
NameMatchesRun runsRun outsRun runs per ro
AB de Villiers (Afr/SA) 97215602156
Yuvraj Singh (Asia/India) 115168901689
WU Tharanga (Asia/SL) 95147701477
Mohammad Yousuf (Asia/Pak) 60133201332
IJL Trott (Eng) 40122211222
EJG Morgan (Eng/Ire) 70114201142
V Sehwag (Asia/India) 77106701067
E Chigumbura (Afr/Zim) 909280928
S Chanderpaul (WI) 6316532826.5
AD Mathews (SL) 627991799

There are a number of batsmen who have been very careful between the wickets. However I think it is probably stretching the truth a little to suggest that Yuvraj Singh is one of the best between the wickets in the world.

So I combined these results using a variety of formulations until I found one that seemed to give the right results. (I can provide it if anyone is interested, it basically multiplies the activity rate by a modified run runs per dismissals, giving a higher weight to run out dismissals). Here were the best runners using my scoring system:

NameMatchesAverageActivity rateScore
IJL Trott (Eng) 4051.370.56922.82
AB de Villiers (Afr/SA) 9752.460.56722.79
RN ten Doeschate (Neth) 2771.220.54222.41
Mohammad Yousuf (Asia/Pak) 6043.10.53720.94
S Chanderpaul (WI) 6359.270.49219.64
JP Duminy (SA) 7742.880.61817.94
A Symonds (Aus) 4144.30.54917.93
MEK Hussey (Aus) 11544.840.61517.46
MS Dhoni (Asia/India) 13754.730.56216.72
Misbah-ul-Haq (Pak) 7744.260.52916.28

So the top of the pile is Trott, de Villiers and ten Doeschate.

An extra statistic that I found interesting was who relied on run runs the most and the least. This is a list of the run run percentages for batsman, ie for every 100 runs they score, how many are not in boundaries. The two surprises there are how high Kandamby is (I've only seen him bat twice, and both times he seemed only interested in finding the rope) and how low Herschelle Gibbs is. Gibbs is so quick between the wickets it is unusual that he scores such a low percentage by running.

NameMatchesRunsRun runsRunning %
GD Elliott (NZ) 3771650069.8
JP Duminy (SA) 772273155768.5
P Utseya (Zim) 8284957968.2
IJL Trott (Eng) 401798122268.0
AR White (Ire) 5165544567.9
SHT Kandamby (SL) 3484757167.4
MEK Hussey (Aus) 1153408227466.7
HH Gibbs (SA) 54175272841.6
JD Ryder (NZ) 37109445241.3
AC Gilchrist (Aus) 37112745740.6
PR Stirling (Ire) 34134954140.1
ST Jayasuriya (Asia/SL) 70198875638.0
CH Gayle (WI) 73245489836.6
V Sehwag (Asia/India) 773269106732.6


  1. I think this stat would be a lot more skewed toward middle order batsmen as the openers always bat during the powerplay and it's harder to pinch singles and twos with more men in the ring. Another thing to be looked at is the number of deliveries off which batsmen have run runs, as just taking the total number of runs doesn't differentiate between ones, twos and threes.

  2. Those are all good points. It is difficult to find the stats for types of runs in a form that is easy to use.

    I also hadn't thought of the bias towards middle order players. I wonder if the new powerplay rules will correct that?

  3. Mykuhl, made a similar analysis a few months ago, but wasn't happy enough with results to post anything. You can see the tables here. The main aim was to determine runs +/- to an average given running.

    The problem is boundaries have a strong positive correlation with runs per ball, and a strong negative correlation with percentage of runs run. So it is very hard to distinguish between both a strong runner (lots of run runs) and an attacking batsman (high strikerate) and a strong runner (high percentage of run runs) and a defensive batsman (nurdling).

    Would be interested in any ideas you have for resolving that. I suspect it would work better at ball by ball ratios (1s over .s, 2s over 1s, 3s over 2s). I don't have ball by ball data though.

  4. When I looked at my rbw score and the batsman's average they correlated quite well. However when I looked at strike rate and rbw score there was close to 0 correlation (r^2 of 0.07).

    I've found that activity rate is better than run runs per ball, as it removes a bias against players that hit boundaries.

    Interestingly the stat seems the most meaningful in test matches, where you would think it is the least important. Martin Guptill seems to score much better in tests when he gets easy runs (ie not against Australia, who cut them off). Usman Khawaja struggles to hit singles, and as a result is constantly under pressure when he is batting (and hence is no longer a current test batsman).

    I had a go at what you are trying to do in one of my early posts,

  5. Sorry, I didn't make that clear. I found a strong correlation in tests between what you call activity rate (run runs per non-boundary ball), and boundaries hit per ball. Which confounds thinking about the quality of run runs. That is, attacking players generally score more singles too, because they are going after more balls generally. The table I linked to corrects for boundaries to try and adjust for "attackingness", but that ended up with a strong negative correlation instead (over-correction basically). I'll play with that though, because it ought to be possible to correct properly.