Tuesday 25 January 2011

A different kind of hundred partnership

In a test match if a pair of batsmen put on a hundred partnership, it generally increases the teams score by that number of runs. This seems fairly obvious. The exception comes when a team is trying to put together some quick runs before declaring. In this case the team are often going to declare at a certain score regardless of how many wickets have fallen. In this case if a team is 220/3 or 220/7 they will still get 220. A 2nd wicket partnership of 100 if the team declares at 220 has not necessarily added 100 because the team may have got to that total even if that partnership had been 0. Likewise if a team is chasing a winning total the same thing applies.

A similar situation happens in one day cricket. A team has two resources that they have to convert into runs, wickets and balls. If a team that has 1 over left and 7 wickets in hand, the wickets are completely irrelevant. But at the start of an innings the wickets are very important. A slow 100 run partnership at the end of a match can actually cost a team runs that they should have got. There are famous stories of this happening, including one where Ian Botham deliberately ran out Geoffrey Boycott because he felt Boycott was scoring too slowly.

Fortunately Frank Duckworth and Tony Lewis developed a system of describing the number of resources left to a team at various combinations of overs and wickets. While their system was designed solely to help find fair targets/decisions in rain affected matches, it is a useful tool for analysing progress throughout games.

I used their basic version to look at all the partnerships over 70 in ODI's this year. The key things I was looking at were the predicted score then they came to the wicket, and then the predicted score when their partnership ended. I already used this technique once in an earlier post Duckworth Lewis as a prediction tool

The problem came with opening partnerships, as there was no data to use for the inital prediction. As the Duckworth-Lewis system was based on the idea of 250 being the expected score for a team, I used this as my starting point.

Using this system I generated this list of the top partnerships:

Batsmen NamesScoreStartEndDL Difference
MJ Guptill, JD Ryder84-84/1, 10.0 Overs282
SM Davies, AJ Strauss90-90/1, 12.1 Overs245
BJ Haddin, SR Watson110-110/1, 19.4 Overs147
HM Amla, MN van Wyk9716/1, 2.3 Overs113/2, 22.2 Overs133
SE Marsh, CL White10033/4, 12.3 Overs133/5, 32.6 Overs128


Haddin and Watson scored 26 more than Ryder and Guptill, but because they took 9.4 overs more to do it, their partnership is not rated as highly. The biggest problem with this way of measuring partnerships is that it penalises a pair if someone else did particularly well. For example, when Haddin got out, Michael Clarke came in an added 103 with Watson, in 19.1 overs at 5.37 which is not a bad effort in the middle overs. However the DL predicted score when Clarke arrived was 397, and it reduced to 322 by the time he departed, giving them a score of -75. However the team were chasing 295, so effectively they are being penalised for the good start, and for them pacing the innings appropriately.

I'll do an update of this table at the end of the group stage of the World Cup and it will be interesting to see who comes out on top.

No comments:

Post a Comment