Duckworth-Lewis vs Best-Scoring overs method – who ya got?

In an earlier post I promised to monitor domestic and international one-day matches this summer, and hypothetically compare the Duckworth Lewis (DL) method to the old best-scoring overs (BSO) method. From the 20th over onwards in the second innings, each method will predict a winner by providing a target score, as if the match was suddenly rained out. So there will be a prediction after over 20, 21, 22… all the way up to 49; 30 predictions in total if the match goes that distance (no points awarded for predicting the winner after over 50).

[Update as of 3/11/2013 – I’ve made the executive decision to add the highest-scoring consecutive overs method to this.]

In the first match Tasmania spoiled the experiment by not lasting until over 40 – but all methods correctly chose the winner correctly 15 times, until the match ended at the end of over 35.

01tas

You’ll notice that the BSO line is pretty smooth, but the DL is jagged – that’s because the target changes markedly each time the chasing team loses a wicket. (In this match, Tasmania were kind enough to demonstrate this frequently.)

In the second match, Ed Cowan of Tasmania upset BSO’s applecart by belting 14 in a 20-run over off Fawad Ahmed – this concentration of runs pushed up all the BSO targets and predicted a Tasmanian win throughout. Whoops. As you can see, the target red line exceeds Victoria’s score in purple, which is in turn above the green DL target. BSEQ, meanwhile, predicted a Tasmanian win until the 36th over, when it came back from a kitchen with a drink and pretended to have been on Victoria all along.

02vic

DL scores 25 for this effort, BSEQ 8 and BSO 0.

The third match was a repeat of the second, with Victoria’s purple line cutting a perfect path to victory over South Australia, above the DL target and below the inaccurate prediction of poor old BSO – BSEQ’s change of heart came in over 34; the fourth pretty much followed suit as Victoria eased past NSW; BSEQ was on the winner all the way, had a panic attack and switched sides as the crunch overs approached, then jumped back on the bandwagon just before the end, saying “See? I was right all along.”

03vic

Another fail for BSO, another late change of mind for BSEQ.

04vic

BSEQ doesn’t know what to think, BSO isn’t even watching the game.

It got a little interesting in the fifth match, as Tasmania just got there in the last over over SA, with 2 wickets in hand. BSEQ was wrong all the way through – but in the closing overs, only by a run or two.

05tas

Only DL knew what was going on – and just barely.

In match six, Queensland stumbled along to 202, but WA couldn’t get anywhere close. All three methods predicted this, through the faint sound of snoring.

06wa

SA played Tasmania in match seven, and so many cool things happened that it needs a dot point list:

  • Tasmania won by one wicket on the second last ball of the match
  • everyone remembered why one-day cricket is fun to watch
  • SA vs Tas: 2 matches in 3 days, Tas wins both by a combined 3 balls and 2 wickets; SA bulk-orders Ben Hilfenhaus voodoo dolls
  • only DL made any correct predictions (you can see Tasmania’s line snaking along underneath them all), but it switched to the wrong prediction after over 40 and stayed there right to the end – if a pitch sprinkler malfunction had occurred after over 49, it’s SA’s match

07tas

In match eight, Queensland put up a big pile of runs and Victoria couldn’t chase them. But (as I’m sure Cameron White will be relieved to hear) DL got on board Queensland just in time to post another perfect slate of predictions.

08qldAnother close match followed, with NSW’s Sean Abbott having a better day than anyone else on his team or WA’s – and better than most of the win predictions too. Frankly, it was all Abbott’s fault – making a late-innings rapid-fire and unbroken partnership of 85 with Steve Smith derailed DL’s early confidence. BSEQ and BSO once again well off the pace.

09nsw

Next, Tasmania pulled yet another Houdini act, this time against Queensland. After lagging behind all predictions, they found themselves with 2 wickets in hand, 4 overs left, and 30 runs needed – and it was no problem.

10tas

Here’s the scoreboard at the halfway point – not great reading for anyone but DL fans:

scoreboard

On to match eleven. And I know what you’re thinking. How long can the Statsman continue to make pithy remarks about every match and chart? The answer is: only this long. Here are the remainder of the matches, without editorial content unless absolutely necessary.

WA vs Victoria, and cricinfo has all the data, but for some reason won’t publish the over by over table. 😥 <– sad Statsman.

Match 12, NSW vs Queensland:

12qld

Dear cricinfo, why are there typos in your over table for this? Aren’t the figures generated from the (presumably correct) ball-by-ball data? Only 21, not 26 in Qld’s 7th over and 3, not 7, in NSW’s 40th. And this isn’t the first time. Hugs, Statsman.

Match 13, WA vs SA:

13sa

Match 14, Qld vs SA:

14qld

All methods predicted the bleeding obvious.

Match 15: Tas vs WA:

15tas

DL fail! As Tasmania dropped 7 wickets for 53 runs. So maybe it was Tasmania fail.

Match 16, NSW vs Vic:

16nswIn an ironic twist, matches 17 and 18 were interrupted by rain and therefore decided by… Duckworth-Lewis. Let’s say it was accurate.

By this point, the competition had reached the finals stage. (This should be obvious, because every team had played every other team once. And then played one further game. Hmmm.)

In the eliminator Victoria made over 320 and for the second time in a week, NSW found that task pretty easy:

19nsw

In the final, NSW got a dose of their own medicine from Qld:

20qld

Queensland celebrated a great tournament victory, but I doubt they were any happier than I am to stop making these graphs. Final scorecard:

score

Duckworth-Lewis – twice as accurate as either of the other methods – or, to look at it from the other angle, BSEQ would produce four times as many wrong results. But that’s not the end of the story – stand by for a further post in November.

Advertisements

Click here to like, or tell me whether this is genius or gibberish

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s