WASP takes playing conditions into consideration unlike D/L method, say creators
WASP has been used in the ODI series between New Zealand and India. Photo Courtesy: Paul McClean’s Youtube channel
The Winnings and Score Predictor (WASP) has become the talk of the town with cricket fans as it flashes across the scorebar during the ODIs between New Zealand and India. Nishad Pai Vaidya caught up with Dr Seamus Hogan and Scott Brooker, the creators of WASP.
The Winning and Score Predictor (WASP) has become an interesting topic of discussion during the ongoing One-Day International (ODI) series between India and New Zealand. To understand a lot more, CricketCountry spoke to Dr Seamus Hogan and Scott Brooker, who created WASP and asked them simple questions a cricket fan may have about WASP.
CricketCountry (CC): How did you get the idea to develop the WASP for cricket?
Seamus Hogan (SH): When watching baseball I had thought about how dynamic programming would be the ideal way to analyse that sport. Many years later, back in New Zealand, an opportunity came to use it for cricket. Using dynamic programming, we could analyse sport and determine what could happen in a particular game. And, my PhD student, Scott, joined me later and we created WASP. It was first used in the HRV Cup in 2012.
Scott Brooker (SB): Yes, it came because of a liking for cricket and Seamus has spoken to Rugby teams as well about this, though nothing has come as yet. It evolved slowly. We were initially looking to see how much risk was justified by batters in different situations by asking questions like: a) How high does the probability of being run-out have to be before it is worthwhile attempting a risky single in different situations? b) If the “death” starts at 40 overs when a team has lost four wickets, how soon would it start if they had lost only two?
SH: To do this, we needed to model the expected number of additional runs a team would score in the first innings based on the number of balls and wickets remaining. And in the second innings, we needed to model the probability of winning based on balls and wickets remaining and runs required. Later on, we realised that these models gave us a way of tracking how well a team was doing throughout the course of a match, which is what WASP is.
CC: What has been the response like for WASP? Seamus: When the WASP was introduced in 2012, my blog post explaining it was read by about 300 people. Later, during the England series, it attracted about 1,000 hits per game. But, in 24 hours after the first ODI against India, we had 15,000 people visiting the blog. There has been a growing interest in WASP since the New Zealand-India series has started. This is the first time it has been exposed to a huge Indian audience.
CC: In simple words, how would you explain WASP to a layman?
SH: In the first innings, the Score Predictor is what the average batting team would score on average against the average bowling team given in the batting conditions in that match, given how much the batting team has scored to that point and the number of wickets and balls remaining. This is based on how other teams have fared in the same situation. When we say the “average” batting or bowling team, we mean the average top-eight international sides based on results in ODIs (or T20 games) played between them since 2008.
SB: In the second innings, the Win Predictor is the probability that the average batting team would beat the average bowling team given the batting conditions, given the number of additional runs required, and the number of balls and wickets remaining. So if WASP is, say, 20 per cent, it is saying that if an average batting team were playing an average bowling team in this situation in these conditions repeatedly, they would be expected to win only one game out of every five.
CC: You use a bit of historical data and bring it into equation. Why was 2008 used as the starting date for the data? Lots of things have changed since then such as the teams itself for example, Sachin Tendulkar isn’t playing for India now and nor is Daniel Vettori a regular in the New Zealand ODI side.
SB: One thing about WASP is that it takes into account the average batting team taking on the average bowling team. It doesn’t take into account the players out there or their quality. For example, when MS Dhoni and Virat Kohli were batting in the first ODI, WASP showed that India’s chances of winning were nine per cent. Now, we know that they are fantastic players, especially while chasing, but WASP doesn’t factor that. Both India and New Zealand may have changed their players since 2008, for example, Daniel Vettori isn’t currently in the side. But, WASP takes into consideration that there would be an able replacement, although no two players can be the same. We thought 2008 would be the right mark. Taking only the previous game into account wouldn’t give a proper reading, nor would it be right to take the 43 odd years of one-day cricket into the equation.
SH: Also, the rules keep changing. The powerplay regulations have evolved, there are two new balls and a general increase in the scoring rate. That is why we have the experts setting the par score at the start of the game. So far, in my view, the broadcasters have set a low score and have been a bit conservative. But, the experts factor in various things to give that count.
CC: The Slog overs form a crucial part in an ODI as the score would just swell up with a few big hits. Does WASP take that into account?
SB: WASP does take into account the fact that batsmen will take risks during the slog overs. If you are 200 for one in 35 overs, your slog overs may start earlier when compared to a team that loses four heading into the 40th over.
CC: What is the main difference between the Duckworth/Lewis method and WASP?
SB: WASP is very different from the D/L method. The main reason for this is that the D/L uses a “resources-lost” criterion while WASP uses a “probability maintenance” criterion. Essentially, the target should be set so that the probability of winning is the same after a reduction in overs as it was at the time of the interruption. In addition, the concept of a par score can add accuracy and take into account different conditions, though the method of selecting a par score if WASP were to be used for setting targets would require some further analysis and discussion.
SH: We would love if the WASP can replace the D/L method as it does take into account various factors and gives you a more accurate picture. The D/L method has certain problems.
We did not give it the name Winning and Score Predictor, we merely made the model. The economist in me would call it the Winning and Score Projection, as that is what it does
CC: You have also left room for expert adjustment to WASP before the start of the game. What is it about and why has it been put in place?
SH: An advantage of WASP is that it can be adjusted so that it is appropriate for the conditions on the day and factor that into give a projected score or probability of winning. But, it is the broadcaster who has to set that. Using dynamic programming, we can take such things into consideration. So, on large grounds, you may get a 250 when compared to a smaller venue with a good batting surface, where you may get 300. Ultimately, it is up to the armchair critic to decide whether they agree with the par score set by the broadcaster. At Seddon Park, you may get a certain score in a set of conditions, but you may not get the same on another ground.
SB: The experts will determine the adjustment factor before a game. We have left that to their opinion. If we do not, then regardless of a flat pitch or a bowler’s paradise, WASP will predict 250 or 260 at the start of the innings, but if you bring the expert opinion into the picture, it might start at 220 or 290, depending on the conditions, and then adjust depending on what happens in the game.
CC: Let me give you two match situations. New Zealand is facing Australia and are 150 for two in 28 overs. In another game, they are playing Australia and are in the same situation. Would the WASP readings of the eventual score be the same?
SB: The reading would exactly be the same. WASP takes into account what an average batting team would score against an average batting team. It may give a figure like 280 for the situation you mentioned. But as we know, Australia’s bowling attack is strong and New Zealand might score about 270. Whereas, against Canada, they may push to 300 or even 310.
CC: WASP isn’t exactly a predictor is it?
SH: We did not give it the name Winning and Score Predictor, we merely made the model. The name was given by someone else. The economist in me would call it the Winning and Score Projection, as that is what it does. What it intends to do is to give you a projection of what is the likely outcome assuming average quality teams, not a prediction taking all information into account. But you can look at the WASP and adjust the number up or down based on how the remaining batsmen and bowlers may be better or worse than average.
SB: Also, you look at a game of football and you can see who is winning. If Brazil and New Zealand are playing, and if the latter leads 2-1, you may say that Brazil may hit back, but until then, it is New Zealand who are leading. But, in cricket, only one team at a time is scoring at a time and WASP tells you who is leading the game currently. And, if it did replace the D/L method, we would keep 50 per cent as the mark. The team whose WASP index would be more than 50 when the game is interrupted would win.
CC: What do you see for WASP in the near future? Are you trying for something similar in other sports such as Rugby?
SH: We hope that it does become a part of Indian Premier League (IPL) and Indian domestic cricket and is used there. Also, we have an App in mind which will help a user determine what score a team can reach etc. They can change the par score as the experts do of course. So, we do look forward to such things in the future.
SB: General dynamic programming can be used in all sports. It mainly has been used in finance. When it comes to a sport like cricket, it can give you a figure as there are many numbers involved. But, for a sport like rugby, it will help you with your strategies — whether or not you should take the ball in the opposition’s half etc. We would love to do something similar in other sports, if those sports were interested. Seamus has had some discussions with a rugby team, but that hasn’t led anywhere yet. SH: It has been used in America quite often. Gridiron football team New England Patriots’ coach Bill Belichick has used it. Economics has been in sport for quite some time. We see a continuation of the very interesting discussions about WASP that are occurring on the web and social media at the moment. We hope to see some very good games in this current series where WASP can give the viewers a statistical perspective on what is happening in the match. And we will tweak the code a little to make WASP easier to use in rain-reduced matches.
(Nishad Pai Vaidya is a Correspondent with CricketCountry and anchor for the site’s YouTube Channel. His Twitter handle is @nishad_44)