CricAlgorithmics gets into Predictive Mode for the Pakistan vs Sri Lanka limited over series
Predicting the score of a team and the outcome of a match has remained one of the biggest challenges in cricket and also one of the fascinations of the game. Such is the complexity of the sport and its technicalities that it is impossible for any stats system to achieve 100% accuracy. But a good stats system should be able to predict team totals with as less error margin as possible and the match result with the maximum possible success rate. The challenge becomes more daunting when it comes to individual batsmen and bowlers.
TVConal’s CricAlgorithmics is an analytics engine in cricket equipped with a wealth of digital data and AI models to discover deep insights about the game and generate predictive and inferential data. It delivers unique AI-powered insights to engage fans, help teams analyze performances, scout oppositions or recruit new players, and to support commentators or editors in live coverage of the match by generating ideas and unique talking points. CricAlgorithmics uses a completely different approach for match analysis and player comparison based on the individualized, situation-based prognosis. Using it one can not only predict the probability of victory for a team in a match but fairly accurate predictions can be made about several nested layers of parameters.
Predictive models are developed to assess how well a player (batsman or bowler) is likely to perform in a given situation with respect to an opponent team, an opponent player, in a specific partnership, and at a particular ground and time. We look at some examples where CricAlgorithmics’ predictive analysis gave a fairly accurate measure of team scores and player performance from the recently concluded limited-over series between Pakistan and Sri Lanka. We will also provide some interesting and unique observations.
PAKISTAN WERE FAVOURITE TO WIN BEFORE EVERY MATCH Pakistan were the higher ranked team (6 th vs 8 th ) during the ODI series. Not surprisingly, they had a higher probability of victory at the start of every match. They had a win percentage of 77% at the start of the second ODI in Karachi and 65% at the beginning of the third ODI. They went on to win both the matches comfortably, by 67 runs and five wickets. Interestingly, Pakistan were also favourites to sweep the T20I series – they were the number one ranked side in the world in the format, and by some distance. Sri Lanka were ranked at Number 7. CricAlgorithmics gave them a win probability of 63% for the series opener. The win probability rose to 71% for the second T20I despite of the loss in the first match. And despite two losses it further went up to 83% at the start of the final T20I. This basically suggested that the losses were treated as aberrations – they were completely against the expected result. Pakistan were the stronger team based on the chosen XI and past performances. Thus, Sri Lanka’s sweep was a massive upset.
HOW THE STORY UNFOLDED IN THE SECOND ODI
– CricAlgorithmics predicted (at the start of the second ODI) that Pakistan would
score 293. They made 305.
– The Expected Score for Fakhar Zaman was a run a ball 61. He ended with 54 off
65 deliveries. Similarly, the Expected Score for Haris Sohail was 45 off 46
deliveries. He went on to score 40 off 48.
– Babar Azam was expected to score 50-plus. He registered a century.
– After Pakistan posted 305 for 7, their Win Percentage shot up to 90% at the half
way stage (from 77% at the start of the match).
– Azam scored his 11 th ton in this match. He reached his fifty in 55 deliveries.
According to CricAlgorithmics, at that stage, Pakistan at 155 for 2 after 31 overs,
had a win probability of as high as 84%. Amazingly, at this point, the stats system
predicted that Babar would score 116 off 114 deliveries. He ended with 115 off
– Sri Lanka lost early wickets and were reeling at 28 for 5 after 10.1 overs. Their
Win Percentage was as low as 7% at this stage.
– Such was Pakistan’s dominance throughout the match that their Probability of a
Win never went below 65%.
– Usman Shinwari was the highest impact player for Pakistan in the match. His
fifer in the match included 4 top-middle order wickets. After reducing Sri Lanka to
28 for 5 in the 11 th over with a win probability of as low as 7% with an initial spell
of 3-16 in 5 overs, he returned to break a potentially threatening 177 run stand for
the sixth wicket between Shehan Jayasuriya and Dasun Shanaka ending all hopes
of a Sri Lankan victory.
Shinwari has had a great start to his ODI career having picked 34 wickets in just 17
matches at an average of 18.61 and strike rate of 22.5 including 3 four-wicket
hauls and 2 fifers.
PREDICTIVE PLOT OVER THE COURSE OF A MATCH
CricAlgorithmics’ Predictive Plot plots the Probability of Victory for both the
teams during the course of the entire match – in this case the third ODI at Karachi.
Pakistan were favourites to win based on the starting XI of both the teams with a
starting Win Probability of 65%. They continued to hold the edge (win probability
above 60%) even as Gunathilaka and Thirimanne resurrected the innings after the
early loss of Avishka Fernando.
But as the partnership gained momentum, Pakistan’s chances started reducing and
went constantly below 60% after ball 90 (15 overs).
From approximately 230 balls (in the 39 th over) till the end of their innings,
Pakistan’s Win Percentage went below 50% (and below 40% mostly after the 42 nd
over) and Sri Lanka became favourites even though the hosts managed to pick a
flurry of wickets towards the death overs.
This is because Sri Lanka had already crossed the expected average winning total
and also scored at a healthy run rate despite losing wickets at the death.
Pakistan had a Win Probability of just 25% at the half way stage!
The slow start (just 18 runs in the first 5 overs) to their chase meant that their
chances of a win further dipped.
But as the opening stand between Fakhar Zaman and Abid Ali grew and as they
started to get more boundaries, Pakistan’s Win Probability curve took a dramatic
upward turn (post the 9 th over) and had crossed 50% post the 14 th over.
It had reached as high as 80% just before the first wicket fell.
A few quiet overs meant that the required rate soared and Pakistan’s chances of
winning came down to around 55%.
Babar Azam and Zaman then up the ante and Pakistan were again firm favourites
with the win probability touching almost 85% till both of them fell in quick
succession and with that their chances reduced by almost half to around 45%.
The next few overs saw volatile up and down movement in the curve – with a few
overs remaining in the chase this was expected as chances of win changed almost
after every delivery. A few boundaries meant that it went up dramatically but then
a few dot balls meant that it rapidly went down.
As Sarfaraz Ahmed and Haris Sohail forged together a stand for the 4 th wicket and
inched closer to the target, Pakistan’s win chance hovered around the 60% mark
but the fall of the captain in the 42 nd over reduced it to 45%.
A few quiet overs meant that there was a sharp fall in the curve and Pakistan’s
chances fell to below 30% before 18 runs of the 46 th over took it to above 90% –
such is the degree of variance in the last few overs!
From here on, It was Pakistan’s match and they went on to register a convincing 5
wicket win with 10 balls to spare.
If we reverse the logic we get Sri Lanka’s Predictive Plot (below):
SOME OBSERVATIONS FROM THE T20I SERIES
– Sri Lanka were ‘Expected’ to score 120 in their allotted 20 overs with a Win
Percentage of 37% at the start of the first T20I in Lahore. They exceeded
expectations and scored 165 for 5. This increased their Win Percentage to 88% at
the half-way stage.
– The Sri Lankan openers – Danushka Gunathilaka and Avishka Fernando – had put
together 84 in the first match of the series but CricAlgorithmics predicted that both
would fail and score below 25 in the second T20I. As it turned out they were
dismissed for 15 and 8 respectively.
– Sri Lanka had reached 57 for 2 after the first 40 deliveries in the second T20I.
The partnership between Bhanuka Rajapaksa and Shehan Jayasuriya had just put
16 for the third-wicket. CricAlgorithmics’ Predictive Plot depicted that Pakistan’s
Win Percentage never exceeded 50% at any stage from hereon in the entire match!
– Sri Lanka had scored 41 for 2 after 5 overs and according to CricAlgorithmics
had a Win Probability of just 36%.The magnificent 94-run stand between
Rajapaksa and Jayasuriya off just 64 deliveries increased their Win Percentage to
80% 15.2 overs into their innings.
– CricAlgorithmics got its prediction bang on for the Sri Lankan top 4 in the third
T20I. It predicted that all of them would be dismissed for below 25. As it turned
out they scored 8, 12, 3 and 13.
THE BABAR AZAM FACTOR
Babar Azam has the highest average (51.37) in T20I cricket (min. 10 innings and
400 runs) since 1 st January, 2018.
He has been the backbone behind Pakistan’s rise to the number 1 ranking in the
format – they have won 18 of the 26 matches and have the best win-loss ratio
amongst all major teams in the world in this period.
Furthermore, he scored 28.12% of the total runs scored by Pakistan during this
period (only considering the matches he played).
6 of his 8 fifties in this period resulted in a Pakistani win.
Babar was dismissed (13 and 3) within the powerplay in the first two matches
severely reducing any chances that Pakistan had of overhauling Sri Lanka’s total in