Fielding Independent Wicket Share (FIWS) : A novel way to measure bowler performances

- Omkar Bhatavdekar and Apoorva Bhide

Introduction

1.2 Malinga to Warner, OUT

The famous Malinga yorker makes an early entry, Warner's feet go nowhere, his bails fly a distance after the ball crashes into the stumps, Malinga provides a big early breakthrough, as he does so often, is there anyone with a better yorker in world cricket today?

In his first match of IPL 2011, a then-27-year-old Lasith Malinga strode out to bowl for the Mumbai Indians in Delhi. His second ball that evening to a young David Warner was a delivery that’s always represented Malinga in cricket consciousness: a full, swinging yorker at 143 clicks that threatened the base of the stumps. Warner had no answers.

As it turned out, neither did young Unmukt Chand who walked in to replace him. Chand was then eighteen, playing his first IPL game in a packed Kotla. The first ball he faced was a low full toss that he kept out. His second was an off cutter on length which he tried to hoick over midwicket, and lost his stumps in the process. No one in the lineup fared much better. Delhi ended up folding for 95. Malinga took 5-13.

If you’ve followed Malinga - or cricket - at all in the last ten years, this is a familiar story for you.

Another day, another left-hander. On a turning track at the Wankhede in 2019, when Malinga took the ball for the 18th over, RCB were 144-2, with two set batsmen at the crease in Ali and de Villiers, looking set for a big score. Ali was batting on 50 (31), striking at 161.

Malinga’s first ball to him was - as it so often is at the death - an off-cutter on a length. Ali had sent about the same ball from Behrendorff sailing over cow corner exactly seven balls ago. This time he didn’t connect. He chipped it straight to midwicket. A couple of balls later, another off cutter on a length held up in the pitch. Stoinis, his chewing gum not even wet, held out a clueless bat, spooning an easy catch to short cover. Two more off cutters produced two more wickets in the 20th over, and Malinga finished with 4-31.

If you’ve followed Malinga’s bowling over time, this is also a familiar story for you.

Looking at the statistics for the two seasons - Malinga had 28 wickets in 16 games at a scarcely-believable 5.95 in 2011, but took 16 wickets in 12 games at 9.76 in 2019 - tells you a story.

But it’s not the whole story.

On an intuitive level, the Malinga bowling in 2011 and the Malinga bowling in 2019 are not the same bowler. The clicks have dropped. Playing for a decade around the world is hardly ever kind to a fast bowler, let alone one with an action as taxing as Malinga’s is. The knees are barely there.

The back has borne a lot.

But then, 2019 Malinga still has those 16 wickets in 12 games. The Malinga of 2019 is, you can argue, a more experienced, canny bowler. On a pitch like that one in Wankhede, to someone running after glory in the last ten balls, his bowling is like quicksand. He still gets those wickets, though he might not get them the same way.

This led us down a path of trying to find the dimension that captures this change. Meet Fielding Independent Wicket Share, or FIWS. It’s a concept heavily inspired from what are called DIPS in baseball.

We’ll come back to Malinga in a bit.

Definition

A formal definition of FIWS would be something like this. A bowler’s FIWS is the proportion of the wickets they take that are aggressive in nature (i.e. a bowled, LBW, or caught behind). In essence, this is an attempt to separate out the wickets that are taken due to the bowler attacking the batsman, versus the batsman making a mistake while taking a risk and getting out.

Now of course, those two things aren’t mutually exclusive. So think about it this way. A batsman’s role in cricket is twofold: he has to defend his wickets so that he can stay at the crease, and - looking at you, Pujara - he has to score runs. An aggressive wicket, as we’re classifying it, is a bowler failing a batsman at the first objective. A non-aggressive wicket is a batsman losing his wicket trying to go for the second.

We believe that FIWS illuminates a particular dimension of a bowler’s skill: being a low-FIWS bowler is a different skill from being a high-FIWS one (Bear in mind: this doesn’t tell you anything about how effective the bowler is at this skill). This isn’t designed to be a world-changing metric that will tell you everything you need to know. We just think this is a metric that can better help us understand bowlers’ performances. And it can be derived using the most preliminary information widely available (mode of dismissal)!

Here’s a list of bowlers with the lowest and highest FIWS numbers in the league in the last five seasons till 2020:

Bowlers with lowest FIWS in the last 5 years in the IPL

Bowlers with highest FIWS in the last 5 years in the IPL

Establishing FIWS as a metric

In this section, we’d like to establish FIWS and try to explain and ground in data the intuition behind it. This section is a bit more technical and can be skipped by those who were too cool to pay attention to statistics in school. (Kidding. You can skip this section without missing a lot, but you should be able to get the points made here even if you gloss over the statistics.)

First, some helpful cutoffs.

The league average for FIWS for bowlers with at least 10 wickets (this cut off was set so that bowlers who roll their arm over for a game and take 2 wickets do not skew this data) is 43. Based on Average ± SD, here’s what we think are reasonable cutoffs to distinguish between bowlers.

Next, we’d like to visualise the intuition behind the seemingly arbitrary segregation of wickets, and show you exactly what that segregation captures. The definition of FIWS rests on the idea that these are two “types” of wickets, so let’s see if that actually plays out.

Here’s the % of balls that produce a wicket in each over of the two innings in IPL history:

This arguably typifies the way T20 is currently played, Narine-sized floater experiments notwithstanding: wickets are more likely to fall in the powerplay due to either bowlers’ skill or due to some risk being taken. They fall much less often in the middle overs, until they become pretty likely again at the death, purely due to more risk being taken in that phase.

Now here are those same graphs, split by Fielder-Independent (FI) and Fielder-Dependent (FD) wickets:

As we’re defining it, FI wickets are aggressive wickets while FD wickets are usually due to a batsman making a mistake while taking a risk. The graph shows that playing out to a big extent: while FI wickets are also more likely at the death, a large part of the increased likelihood of a wicket falling at the death is due to an increase in FD wickets, because they’re driven by batsmen taking more risk.

That seems okay for a breezy visualisation, but we can actually test this out in a particular scenario. In the second innings, the required run rate (RRR) provides a proxy for how much aggression is required to be shown by the batsman at any point. If we look at the probability of getting each wicket at different values of required rates (excluding values above 36, because the match is then lost), here’s what we see:

After an RRR of about 9, when the pressure really starts to build, FD wickets increase a lot faster than FI wickets - in fact, while FI wickets increase slightly with required rate, they’re roughly as likely at an RRR of 18 as they are at 36. FD wickets are a lot more dependent on the amount of risk a batsman takes. To us, this shows that the intuition behind the definition probably holds - that these are two different types of wickets.

For the more mathematically-minded, FD wickets (mean = 0.0515) were generally more likely to fall than FI wickets (mean = 0.0257) at different RRR values, and this effect was significant

[t(56.5) = -4.9173; p <<0.001] and showed a large-sized effect (r =0.547). (Quade’s ANCOVA test was also performed but yielded similar results)

But we don’t want to stop here! We want to establish that these are two different types of skills - that being a high-FIWS bowler and being a low-FIWS bowler are two different things. This should mean that a bowler’s FIWS should be roughly similar in different seasons. So let’s see if that plays out.

Let’s look at performance in consecutive years. Filtering for all bowler-season combinations where a bowler took at least 7 wickets in two consecutive seasons yields 260 cases. Now, we’re going to plot the performance of a bowler in one season against his performance in the next. So, X-axis is a bowler’s FIWS in one season, and Y is the same bowler’s FIWS in the next season:

A bowler’s FIWS in one season and his FIWS in the next season is significantly correlated, rs = 0.2688, p < 0.001 [t(258) = 4.483].

This is a small-to-medium effect size, which tells us that bowlers’ performances in this regard are somewhat correlated. FIWS looks to be capturing something that is characteristic of a bowler’s performance in a season, or something that doesn’t change a lot between consecutive seasons. So FI and FD are two different types of wickets, and bowlers take a roughly characteristic, correlated share of them in different seasons.

Now, let’s look at what that something can tell us.

Applications

1. How to use your resources efficiently?

When they returned to the IPL in 2018, the Chennai Super Kings got two promising young bowlers from the Rising Pune Supergiant: Deepak Chahar and Shardul Thakur. Just 25 and 26 years old, they were technically both young, uncapped Indian fast bowlers. But they were, and have been, two very different kinds of fast bowlers, even to a non-expert viewer - Chahar relies much more exclusively on swing to attack the batsman and get his wickets than Thakur does. Their FIWS bears this out: Chahar’s career FIB is 42.2%, while Thakur’s is 17.8%.

As things go, this has also been reflected in the way a shrewd captain like Dhoni has used them. Chahar bowled only seven overs for RPS in 2017, but they were all in the Powerplay. He has since bowled more than 70% of his total overs in that phase in each season till 2020. Thakur has bowled 38%, 34%, 53% and 21% of his overs in the Powerplay in this period.

Looking at their wickets by phase lets us see this even better. Chahar’s FIWS in his powerplay overs is 42%, which shifts slightly to 45% when he bowls at the death. Thakur’s FIWS in his powerplay overs is 36%. It becomes 14% when he bowls at the death.

They may both be 28-and 29-year old Indian right arm fast bowlers, and they’ve both gone on to play for India since. But they’re not the same bowler. And using them according to their skillsets has allowed CSK to get a lot out of them. Their FIWS can help us take a peek at those skillsets without needing a lot of data.

2. Developing a new weapon

Bhuvneshwar Kumar arrived on the international scene when he turned up at the Chinnaswamy one day and destroyed Pakistan’s top order for fun. Since then, his calling card has been his ability to generate prodigious swing, which has made him a phenomenon both in the IPL and internationally.

Bhuvi’s powers were arguably already at their peak in 2016, when he got the purple cap and played a key role in SRH’s title win, with 23 wickets in 17 matches. In early 2017, though, he stepped things up further when he developed a knuckleball, unleashing it in the 2017 IPL season.

Without getting into too much of the physics of the knuckleball, a knuckleball is delivered such that the bowler barely has any control over its trajectory. It’s a different beast from the regular slower ball. It doesn’t just beat the batsman for pace. It deceives him with dip, bounce and trajectory. Andrew Tye was one of the first adopters to really successfully use one, and it’s now a variation possessed by a lot of bowlers. A knuckleball makes you deadly when the batsman is trying to attack: Tye has the lowest FIWS in the league.

Bhuvi won another purple cap in 2017, taking 26 wickets in 14 games. He was arguably a good aggressive bowler, but the knuckleball allowed him to solidify another skill: he was now much tougher to hit. In 2016, he had taken 57% of his wickets at the death. In 2017, he took 69% of his wickets at the death.

And his FIWS can help us see this change: between 2016 and 2017, his FIWS went from 50% to 36%.

3. Action Changes: What’s the deal with Sunil? (yeah.)

11.4 Narine to Rohit Sharma, OUT

Opened him up! That was a beauty from Narine. Pitches middle, hits off, and Rohit clearly didn't pick it. Was looking to go on the back foot and work it through the leg side, initially, and then tried to adjust and play with a straight bat, but it slipped quickly past his bat-face and clipped his off bail

Sunil Narine has been one of IPL’s bonafide match-winners. An IPL career economy of 6.78 (before the start of the 2021 season) with a career SR of 21.94 is an outstanding achievement in itself.

Narine’s faced his ups and downs. His action was first reported in the 2014 Champions League T20, where he was actually banned from bowling for KKR in the final. His career has since been dogged by issues with that bowling action. He was reported again and given a warning for his action in IPL 2015, and he had to miss the 2015 World Cup. The ICC cleared his action just before the IPL after a major action remodeling in 2016 (“In the end, it all worked out.”, as the article confidently declares), but he has since been reported again (once each in the 2018 PSL and the 2020 IPL).

Narine had taken 74 wickets in 54 innings at an economy of 6.10 at an SR of 17.4 before this action remodel. He has taken 52 wickets in 61 innings at an economy of 7.69 and an SR of 27.0 since.

Here’s his FIWS over time:

Something has clearly changed, and it’s not just how many wickets he takes. Bowling with his new action, Narine may still have guile, but he isn’t able to take wickets the same way anymore. He’s not the Narine of 2015, and expecting him to bowl like one may be setting too many people up for disappointment.

---------------

19.2 Malinga to Compton, OUT

Timber. You don't try such cuteness against Malinga, but this seems to be Compton's favourite mode of operation. He shuffles across the line to try to scoop, Malinga is full and straight. Not quite a yorker, but too full for that shot. Middle stump knocked back

In the semi-final of the Champion’s League T20 2011, Malinga walked up to bowl the last over against Somerset with the job nearly done. The two set batsmen, Kieswetter and Buttler, had both been dismissed in the 19th over, and Somerset still needed 15 to win. Malinga proceeded to get two batsmen bowled with full and straight balls, and kill the chase entirely.

Malinga has had a long and successful career, bowling primarily as a strike bowler. His career FIWS is 55%. But that FIWS has had a long journey as well. Here’s his FIWS over the course of his career:

Malinga started his career in 2009 with an FIWS of 66.67%. He ended in 2019 with an FIWS of 37.5%. The Malinga of 2019 was a different bowler from the Malinga of 2011. In fact, we believe Malinga was able to have a long and successful career in part because when his primary skillset started waning, he was able to develop another one, and be nearly just as successful with it.

So, when he came on to bowl the 20th over in the CLT20 semi-final in 2011 against the likes of Nick “cuteness” Compton, he trusted his yorker and destroyed some furniture.

And when Malinga loaded up to deliver what turned out to be the final ball of his IPL career, with the IPL 2019 trophy on the line, having been smashed all around the park in his first three overs, with Thakur waiting for a hoick, he trusted... his off-cutter.

19.6 Malinga to Thakur, OUT

Mumbai Indians win IPL 2019! Slower ball on middle, and the leg-side swipe has been missed by the batsman. Malinga appeals, and umpire Menon raises his finger. Lasith Malinga has redeemed himself after that 20-run over.

--------------

We hope that this metric gives you a small new dimension to look at the game. While we’ve written about a few illustrative scenarios where we think FIWS illuminates a key difference, there are several areas we’d like to explore with this metric. Here’s a few questions we think we can explore further (any suggestions and analyses are welcome!):

• How does FIWS vary with venue? We think it’s likely to vary with specific factors relating to the pitch, but data that granular isn’t easily available.

• On a related note, which skills are better transferable? For example, do high-FIWS bowlers from the BBL do well in the IPL, too?

• How can FIWS inform team combinations? Does a particular combination of high- and low-FIWS bowlers work best?

• Do batsmen have a characteristic FIWS too? Are there some batsmen who are a lot less likely to get out caught or stumped, for example? Are specific batsmen weaker or stronger against low- or high-FIWS bowlers?