DLS System is not that complicated: Dr. Steve Stern

ICC adopted the DLS since the World Cup, it was implemented in a few games, and I wanted to ask what is different from the erstwhile DL system?

The S actually just stands for my surname, Stern. I have been working with Frank Duckworth and Tony Lewis for a number of years about how the method was coping with modern scoring trends. The new piece that I put in is designed to cope with the modern scoring trend of high scoring games and was particularly true in the world cup as you saw there were a few matches over 400 runs. The old version, the original DL just didn’t quite cope correctly with them because it was built on data that was only for matches that were 300 or below.

Please explain in simple terms how the system works. I don’t want to go in to Technical details, I would like for you to explain it to someone who asks “How does the DLS work”?

The simplest explanation I have is based on the scoring data that we have for the last five or six years, what we do is we figure out what proportion of a team’s score they accumulate in any particular period of the game.  For Example, “What proportion of team’s score do they get in the 35th over if they have seven wickets in hand?” The Duckworth-Lewis and the Duckworth-Lewis-Stern say well that is 3.2%. That way if that over is missing because of rain, we tabulate all those percentages for all the missing overs and determine that this team missed out on 33% of their scoring opportunity and that way we can rescale what they actually scored to what they probably would have scored, had they actually been able to play the overs that were lost.

The idea is the same for both Duckworth-Lewis and Duckworth-Lewis-Stern, but what Duckworth-Lewis-Stern does is it uses the modern scoring data to basically look for the really high scoring matches were those percentages have to be modified from what they would have been if the old version of Duckworth-Lewis were in place. I find this interesting because then the world cup game in Canberra between WI and ZIM, the one where Chris Gale scored his 200 ended up being a very high-scoring match. The West Indies scored 370 odd runs (372/2). In that case, there were only a few overs lost for Zimbabwe so it really didn’t make a huge amount of difference; but if it had been a big number of overs lost then the Duckworth-Lewis would have given a rather different picture of the match than Duckworth-Lewis-Stern. Duckworth-Lewis-Stern would have been a more fair adjustment to the target because it’s based on information that is now available that Frank and Tony just didn’t have when they were first developing their method. There just are a lot more high-scoring matches these days. In addition there is the 20-20 matches where scoring happens at a higher rate.

You mentioned you use past data over a period of five to six years, you very well know that all overs are not the same and all wickets or not the same. For Instance, getting Chris Gayle out is not equal to getting Trent Boult out. The same way a rain interruption in the 20th over is not the same as a rain interruption in the 37th over, when teams are really pressing to score more, does the system create an adjustment based on the batsmen and bowlers or it just works on historical team scoring records?

It certainly creates adjustment based on the situation in the game. Therefore, if there are seven wickets down it assumes that the batsman is probably going to be a lower order batsman. It doesn’t make a specific adjustment for individual players. That would be a bit too difficult to implement directly. There isn’t enough information to make that possible and it will get very complicated.

How difficult is it to read the DLS chart? I’m asking this since we have seen a fair share of goof ups by players and coaches.

It’s just a big table. I don’t find it very difficult at all.  To be fair, the mistakes that were made were minor ones and it’s just that they turned out to be very catastrophic for the team. For example, people very commonly make the mistake, even though it says very specifically on the chart, that the scores on the chart are the targets to win when in fact they are the targets to tie.

Then the other thing that does happen every once in a while is there are two versions of the charts, one is just over by over, and that is the one that most people look. If rain arrives in the middle of an over there are adjustments that actually go ball by ball. Obviously, that won’t fit on a single page.  So there are several pages, and people read the table and they read it wrong. There is a column for each number of possible wickets available, and they read the wrong column because I think what happens is they don’t start counting from zero. The first one is zero wickets down but they think that one wicket down and so they count across to the wrong column, but really the table is pretty simple.

The official scorers have a computer program which I wrote. They don’t use charts at all. They have a little Java applet and they enter all the information that is relevant to the match at the time and then it pops up the target score for them. I am not sure why the teams don’t use that. The application I created also creates the tables and they get printed out. I am not sure why the coaches don’t just use the program. It would be much easier than the charts.

Are you the custodian of the system now?

I have been for about more than 18 months now, yes. Frank and Tony were both in their 70s and decided to retire at the end of 2013. The ICC then was looking for someone. Frank and Tony suggested me; I had a quick interview with some of the people with the ICC. They knew me since I worked with Tony and Frank. We have a three-year contract, which then gets renewed. Presumably, if they are happy with what I have done, they will renew that for another three years and we will go from there.

How does the system work for T20 games? There is always criticism when the target gets reduced in the T20 game. How does DLS account for that?

The statistical data makes it very clear that T20 matches on average looks exactly like the end of a 50 over match. So the suggestion that they are a different game is not really true from a purely statistical perspective.  Of course, there are differences in the rules and the actual play itself.  So the main issue with the DL method was that it didn’t quite get things right for high scoring situations and typically T20 games are high scoring.  The DLS starts to make noticeable changes in a one-day match if the scores get above 330.

In T20 matches, that corresponds to about a score of 170-175.  If the scores are lower than that then DL does a fine job for either format of the game.  It is just that 175 in T20 is very common compared to 330 in an ODI although it’s becoming more common these days. Anyway to give you an example, The T20 world cup 2010 between WI and England got interrupted by rain and England had scored 190 in 20 overs. That was a very high score and the DL method set a target of 60 in 6 overs, The DLS would have changed the situation to 66 of 6 overs.

At the end WI won the game with a ball to spare and if DLS was implemented, they would have had to hit a six off the last ball to win the game. While six runs might not look huge, it makes a significant difference in a shortened game. An extra run an over will make a significant difference to the result.  In the context of the game, WI won with a ball to spare. The other thing about that target that people don’t remember is that England bowled 5 or 6 wides. If you take those 5 runs away plus the extra balls they delivered, 66 would have been a much harder target for the West Indies to get. So England did themselves no favors in how they bowled.

Two matches from the past before the Duckworth-Lewis was even there. Both the matches from the 1992 World Cup. The first one was between India and Australia. Australia scored 237 runs in 50 overs. India had to chase 236 in 47 overs, Rain interrupted play with 16.2 overs being bowled and India on 45/1. What would have been total for India if Duckworth-Lewis System was in place? Needless to say India lost that Game by 1 run.  The other Game is the famous England-South Africa Semi-Final in the same tournament, SA walked off the ground needing 22 of 13 balls, and when they came back still needed 22 of 1 ball. What would have been total for South Africa if Duckworth-Lewis System was in place?

For the India – Australia game, The DLS target for that scenario would have been 229 in 47 overs. So, India would have needed a further 184 runs in 30.4 overs with 9 wickets in hand. With D/L India would have won with four balls to spare, Since India reached 231/7 in 46.2 overs.

Everybody asks about that semi-final game. In fact, some people think it was Duckworth-Lewis that made that problem and I would like to clarify it was not DL.  So to answer that scenario, SA would have had to hit a boundary of that last ball to win that game.  They would have required 4 runs of 1 ball. It would have been hard but at least it wouldn’t have been impossible.

On that note, I thank you for your time Dr. Stern on Behalf of Cricket Web.

I’m glad to have chatted with you and hope this helps people understand the DLS system a little better.