Thursday 13 January 2011

Duckworth Lewis as a prediction tool

A few years ago I signed up with bet365.com to see how good my cricket prediction skills were. I'm not really a gambler, so I quite liked the fact that I could make 20c bets, where the aim was really to see how good my skills are.

I discovered that I'm quite good at betting on cricket, but terrible at betting on American sports, Rugby League and Football. Overall the different sports have canceled each other out, I put in $60, I've taken out $75, and I have about $7 in there at the moment. (As I said I'm not a big gambler). My highlight was probably my very first bet, Bangladesh to beat India at $8.00 at the world cup.

One thing that I found to be quite useful was a score predictor that I made using the Duckworth-Lewis tables. While they are not fool-proof, they are certainly better than any method that I've found. They just need to be applied using common sense. If New Zealand are playing bump the score up a little (given that New Zealand bowlers are better at batting than New Zealand batsmen), if a minnow is playing drop the score down a little (non-test teams tend to suffer terrible collapses - tail often starts at 5 or 6)

The tables are also not much use for the first 5-10 overs, but the scores normally start to be realistic about over 15.

For example, in the South Africa - India game last night, after 20 overs the DL predictor had South africa getting 273 (they eventally got to 289) while after 20 overs it had India getting 161 (they got 154). These are fairly good predictions after 40% of the innings.

Another interesting game is the first match between Bangladesh and Zimbabwe from December 1. After 20 overs Zimbabwe were 84/3, with a DL score of 219 (they eventually got 209). Bangladesh were cruising at 76/1 (DL score 270). However a Ray Price wicket followed closely by a run out left Bangladesh in trouble after 22 overs at 83/3 (DL score of 205) and this turned out to be the turning point, as they eventually were bowled out in the final over for 200, falling 9 runs short.

It is also a useful tool to quantify contributions of partnerships. Just after de Villiers and Duminy came together it was 82/3 off 14 (DL score of 247). Just after de Villiers departed it was 215/4 off 36 (DL score of 335). The difference that their partnership made was an increase of 88 to the DL score, which is really quite impressive.

> As a side note their partnership of 131 contained only 36 runs in boundaries, and yet it came at a run a ball. More evidence of the importance of running between wickets (and fielding)

3 comments:

  1. D/L is a confusing mechanism for many .. and I freely admit to being a techno numpty.

    What is the easiest D/L calculator/prediction tool, for idiots.

    Thanks

    ReplyDelete
    Replies
    1. That's a good question. I don't think there is a calculator out there. I'll see if I can write one for this site.

      Delete
    2. There is now a link for a predictor on the site. 2nd down on the list of links on the right of the page.

      Delete