Innovative ranking strategy for IPL team formation

Arnabi Mitra; Debayan Ganguly; Kingshuk Chatterjee; Ritajit Majumdar; Saptarshi Banerjee

REVIEW 5 major objections 5 minor 8 references

Innovative ranking strategy for IPL team formation

T0 review · 5 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read A role-weighted heuristic ranking plus a greedy credit-budget algorithm claims to select the best possible IPL team and the best alternate player for an unavailable target.

desk verdict A clearly-written heuristic ranking and greedy team-selection prototype for IPL, but the 'best possible team' claim is unsupported and the validation is circular. read the letter →

arxiv 1908.01725 v1 pith:A7AZTEBQ submitted 2019-08-05 cs.SI

Saptarshi Banerjee , Arnabi Mitra , Debayan Ganguly , Ritajit Majumdar , Kingshuk Chatterjee This is my paper · ORCID

classification cs.SI

keywords IPLcricketteamselectionheuristicrankinggreedyalgorithmT20playerevaluationcreditpoints

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper tries to prove that a fixed IPL auction budget can be turned into a concrete 15-player squad by a transparent arithmetic recipe: score every batsman and bowler on a handful of standard and derived statistics, weight those statistics differently according to the player's role (opener, middle-order, finisher, or bowler), combine career performance with current-season form and an experience factor, then run a greedy algorithm that fills each position bucket under a credit-point cap. The paper claims that this ranking reproduces the generally acknowledged hierarchy of IPL players and that the same algorithm names the best replacement when a target player has been bought by another franchise. A reader would care because IPL franchise decisions are currently made by human think-tanks under time pressure, and an automated, interpretable ranking plus greedy selection offers a reproducible way to spend a fixed budget.

What carries the argument

The load-bearing object is the position-specific weighted score; for example an opener's score is $30 \cdot cost_{SR} + 30 \cdot cost_{Avg} + 20 \cdot (hc/innings) + 10 \cdot cost_{RunWicket} + 10 \cdot cost_{HardHitting}$, where each cost feature is the player's raw value divided by the maximum over all IPL players. The final rank for a cluster is $\text{career score} \times cost_{xfact} \times (\text{current-year score}/\text{cluster mean}) + \text{current-year score}$, which is what balances long-term performance against current form. The greedy algorithms then partition each cluster into credit-point groups, assign each bucket a credit cap equal to the team value times the bucket size divided by four, fill positions from the highest rank downward, and backtrack by lowering a previously assigned credit when the remaining cap cannot cover any player in the pool.

What would settle it

Rerun the ranking with each opener weight shifted by ±5 points while holding the other weights fixed; if the top-five opener list changes by more than one position, the ranking is an artifact of the exact weights. Alternatively, assemble the 135-credit squad from the 2018 pool with Algorithm 2 and compare its players' total actual fantasy points over the 2018 season against every other credit-matched squad; if the algorithm's squad is not at or near the top of that distribution, the 'best possible team' claim is contradicted.

Watch

Extended reading notes

Core claim

The central claim is that a weighted-sum score, built from normalised career statistics, normalised current-season statistics, an experience factor, and position-specific weights that sum to 100, yields an IPL player ranking that agrees with expert judgement, and that two greedy algorithms can convert that ranking into a budget-feasible team of 15 players and into a next-best substitute if a chosen player is unavailable. The paper works this out for openers, middle-order batsmen, finishers, and bowlers; it exhibits the top five in each cluster and one full 135-credit squad chosen by its first algorithm, and it shows how the second algorithm avoids stacking two finishers in the wicketkeeper slots. The method is heuristic: the weights are set by hand so that the resulting tables match the authors' known player ranking, and the credit groups (10, 9, 8, 7 in the example) stand in for auction prices.

Load-bearing premise

Everything downstream depends on the hand-set feature weights — for example openers get 30 for strike rate, 30 for average, 20 for half-centuries, and 10 each for running and hard hitting — so if those weights do not capture what actually wins IPL matches, the 'best team' is only the best team according to the designer's guess.

Editorial extensions

If this is right

A franchise can fix its bucket sizes and total budget, then read off a complete 15-player squad directly from the rank-ordered, credit-grouped player lists.
If a target player is taken by another team, the next player of the same credit point (or the highest affordable credit point) in that bucket is the designated replacement, so an auction miss becomes an immediate substitution decision.
A player who scores highly in more than one cluster is labelled with both roles, so the algorithm can shift players between buckets when one position is oversubscribed.
The credit-point grouping and bucket sizes are user-set parameters; changing the team value from 150 to 135, say, changes the squad without altering the underlying ranking.
The second greedy algorithm can be extended with extra restrictions, such as requiring one wicketkeeper to be an opener, illustrating that the framework tolerates team-specific constraints.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The same scoring-and-greedy machinery should transfer to other T20 leagues (or to later IPL seasons) as long as the credit-point groups and bucket sizes are reset, because nothing in the method depends on IPL-specific rules beyond the auction budget.
If the hand-set weights were replaced by weights learned from match outcomes, the framework would become a proper predictive model rather than an expert-opinion mirror; a natural check is whether learned weights reproduce the same top-five tables.
The greedy backtracking may select a few very highly ranked players alongside several low-ranked ones, as the paper itself notes; comparing its total squad rank-score against a dynamic-programming solution over the same credit groups would show how much optimality costs.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Here's my take on the IPL ranking paper. The most useful thing to know is that the central "best possible team" claim does not survive contact with the paper's own algorithm. Algorithm 1 is a greedy heuristic with no optimality argument, and the paper's conclusion explicitly concedes that it may pair high-ranking players with low-ranking ones. So the abstract overpromises.

What is actually new: the two derived features (RunWicket and HardHitting), the role-specific weighted scoring formulas, and the two greedy algorithms with the credit-group bucketing. None of these are deep, but they are concrete and the authors are unusually honest about the heuristics. The feature definitions are sensible, and the clustering of batsmen into opener/middle/finisher with different weights is a reasonable practical move. The second algorithm's constraint on wicketkeeper buckets is a nice touch.

The soft spots are serious. The feature weights in Sections 3 and 4 are hand-set, and the only validation is that the resulting rankings "conform with our known player ranking" — the same ranking that guided the weights. That's circular in effect. No data or code are released, so the tables can't be independently checked. The alternate-player logic in Section 5.3 only looks for a replacement with the same credit value; it can miss a better player available at slightly lower credit. These are not manufactured flaws; they are in the text.

That said, the paper is not a waste. As a prototype for a decision-support tool, it is clearly written and the pieces hang together. The authors acknowledge the greedy limitation and point to dynamic programming as future work. If the claim were softened to "a reasonable heuristic team selection," the paper would be credible. As it stands, the "best possible team" wording is unjustified.

I would not cite this in my own work, but I would send it to a serious referee if the venue is a sports-analytics workshop or an applied journal. There is enough concrete method to critique, and the limitations are honestly stated. The right outcome is either reject with a clear message about the optimality claim and validation, or major revision where the authors either prove optimality (unlikely) or recast the contribution as a heuristic and add a real baseline comparison. For a reading group, it's a useful case study of circular validation, but not a paper that changes how I think.

Referee Report

5 major / 5 minor

Summary. The paper proposes an automated IPL team selection system built from a heuristic player ranking and greedy team assembly. It defines traditional and derived performance features for batsmen and bowlers, clusters batsmen into openers, middle-order batsmen, and finishers, assigns hand-set weights to those features to obtain player scores, converts ranks into credit points, and then uses two greedy algorithms to select a 15-player team under a fixed total credit. It also sketches a replacement logic for unavailable target players and presents a worked example team. The abstract claims that the proposed scheme and algorithm determine the best possible team and the best alternate player if a target player is unavailable.

Significance. The problem addressed is practical and relevant: few studies automate complete IPL team construction under a budget, as opposed to ranking individual players. The paper has concrete strengths: it uses real IPL data, gives explicit scoring formulas, provides a complete 15-player example, and is transparent that the feature weights are heuristic and that the greedy algorithm has risks. If the 'best possible team' claim were supported, the work would be a useful decision-support tool for franchises. However, the central claim is not established. No team-level objective function is defined, no optimality proof or comparison to an exact optimum is given, the feature weights are validated only by agreement with an informal 'known ranking', and the worked example depends on an inconsistency in the credit-unit calculation. The paper is best read as a heuristic proposal rather than as a method that provably determines an optimal team.

major comments (5)

[Abstract and Section 5.2] The claim that Algorithm 1 outputs the 'best possible team' is not supported. The algorithm fills each bucket position with the highest remaining credit point and backtracks only when the remaining credit is below the bucket's minimum; it never compares alternative teams with equal or lower credit but higher total rank, and no team-level objective function is ever defined. The paper's own Section 6 concedes that the greedy algorithm 'may select some high ranking players with some very low ranking ones', which directly contradicts the abstract's unqualified optimality claim. An optimality proof or a comparison with an exact optimization is required before this claim can stand.
[Section 5.2, Algorithm 1, line 1] The statement 'unit← value/5' is inconsistent with the worked example, which sets value=135 and unit=9. Since the example has 15 players and five buckets, the intended per-player unit is value/15 = 9, not value/5 = 27. Taken literally, the total cap over all five buckets would be 27×15 = 405 credits for a team valued at 135, exceeding the budget by a factor of three. This is a concrete numerical error that prevents the algorithm from being run as written and must be corrected.
[Sections 3 and 4] The feature weights in the ranking formulas are heuristic and are validated only by qualitative agreement with a 'known player ranking'. Section 3 states that the weight distribution is chosen so that the rankings 'conform with our known player ranking', and Section 6 repeats that the ranking is 'in acceptance with the known player rankings'. Because the same known ranking is used both to select the weights and as evidence of success, the validation is circular. No independent test against match outcomes, a holdout ranking, or an external rating is provided, so the ranking that Algorithm 1 optimizes is not established as a valid proxy for team quality.
[Section 5.3] The 'best alternate player' claim is not supported. The replacement logic only considers another player with the same credit point (e.g., Gayle for Warner at credit 10) and then, if none exists, drops to the next credit value. It does not search over all available players for the one that maximizes the team's total rank under the budget constraint, so a lower-credit player with a higher rank than another same-credit player would be missed. The procedure therefore cannot determine the best alternate player in general.
[Section 3.3] The finisher ranking formula uses 'not out(i)' and 'not out[i]' as features, but this feature is not defined in Section 2 or listed in Table 1. In addition, the text states that 'Mean ﬁnisher is the average score of all the middle order batsmen', which appears to be an error: the mean should presumably be over finishers. If taken literally, every finisher's rank would be normalized by the middle-order mean, changing the finisher ranking used in the team selection.

minor comments (5)

[Section 4] There are several typographical errors: 'bolwer' appears twice, and 'ﬁnal bolwer' should be 'final bowler'.
[Table 5] The spelling 'P . Chawala' should be corrected (commonly 'P. Chawla'), and Table 6's 'K. Jadav' is inconsistent with Table 5's 'R. Jadeja' for the same player.
[Section 5.1] The description 'each group contains cn/cp players' assumes that cn is divisible by cp; the handling of remainder players when this is not the case should be specified.
[Section 5.2, Table 6 caption] The table caption says 'selected using Algorithm 1' but the team contains two wicketkeepers who are also finishers; the relationship between the five buckets in Algorithm 1 and the role clusters (opener, middle-order, finisher, bowler) should be clarified so that a player selected in one bucket is not implicitly double-counted in another.
[Section 3.1 and Section 5.3] Notation f(i) versus f[i] is easy to confuse; consider renaming the season-specific values to something like f_career and f_recent, and the backward step in Algorithm 1 should be explained in plain language because the pseudocode's 'credit at j' manipulation does not describe how the player assigned to position j is changed.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the ranking and team-selection construction is self-contained, and the known-ranking agreement is a stated heuristic sanity check rather than a fitted input.

full rationale

The paper's derivation chain is explicit: (i) define traditional and derived features from career and last-season statistics; (ii) assign hand-set heuristic weights per role (openers, middle order, finishers, bowlers); (iii) compute weighted scores and ranks; (iv) bucket players into credit groups based on those ranks; and (v) run a greedy algorithm that fills each position with the highest-credit affordable player. Each step is computed from the preceding quantities by the equations given in Sections 2–5. No step is defined in terms of the final output, and no parameter is fitted to a target result and then reported as a prediction. The only place where an external ranking appears is in the statements that the resulting rankings 'conform with our known player ranking' (Section 3) and 'is in acceptance with the known player rankings' (Section 6). That is a qualitative sanity check, not a fitted input: the weights are introduced as heuristics 'so that it models the skill requirements' and no numerical fitting to the known ranking is described. Thus the validation loop is weak but not circular. The paper's own concession that the greedy algorithm 'may select some high ranking players with some very low ranking ones' (Section 6) undermines the unproven 'best possible team' wording, but that is a correctness or optimality gap, not a circularity. There are no self-citations used as load-bearing evidence, no imported uniqueness theorem, and no renaming of a known result as a new derivation. The central construction is therefore self-contained, so the circularity score is 0.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The rankings and teams are generated entirely from hand-set weights, user-defined budget parameters, and unvalidated assumptions. No external benchmark is used, and the only agreement check is the authors' subjective 'known player ranking'. This makes the contribution heavily dependent on choices that a reader did not pay for and that are not tested.

free parameters (3)

Role feature weights for scoring formulas (opener, middle-order, finisher, bowler) = opener: SR 30, Avg 30, HC/innings 20, RunWicket 10, HardHitting 10; middle: SR 20, Avg 30, HC/innings 10, RunWicket…
Chosen by hand in Sections 3 and 4 to reflect expected skill requirements and to make top-five lists match the authors' known rankings; no optimization or external validation is performed.
Credit point group valuations and number of groups = 4 groups with credits 10, 9, 8, 7 in the example
Section 5.1 states the number of groups and the valuations can be varied; the choice directly controls which players fit under the budget.
Team budget value and bucket sizes = value 135; wicketkeepers 2, openers 2, middle-order 3, finishers 2, bowlers 6
User-chosen example in Section 5.2; the greedy output depends directly on these numbers and the division rule unit = value/5.

assumptions (4)

domain assumption The IPL statistics database created by the authors is accurate and complete.
The database is described in Section 2, but no source, schema, or release is provided, so it cannot be independently checked.
domain assumption A linear weighted sum of normalized aggregate statistics is a valid proxy for a player's value in a role.
All ranking formulas in Sections 3 and 4 are weighted sums; the paper provides no empirical test of this proxy against match outcomes.
domain assumption A batsman's 2018 season alone is sufficient to represent current form.
Section 2 states that only 2018 season values are used for current form; there is no evidence that one season is stable or predictive.
domain assumption The authors' 'known player ranking' is a valid ground truth for judging the heuristic.
Section 3 uses agreement with known rankings as evidence of correctness, but that target is subjective and was informally used to guide the choice of weights.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Innovative ranking strategy for IPL team formation." pith.science (2026). https://pith.science/paper/A7AZTEBQ

@misc{pith2026190801725,
  author       = {Pith},
  title        = {Pith review of: Innovative ranking strategy for IPL team formation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/A7AZTEBQ}},
  note         = {Machine review of arXiv:1908.01725}
}

read the original abstract

Indian Premier League (IPL) is a tournament of twenty over cricket matches. Teams of this tournament are selected via an auction from a pool of players. Each team employs a think-tank to build the best possible team. Few studies have been performed to automate the process of team selection. However, those studies mostly concentrate either on the current form of the players, or their long term performance. In this paper, we have (i) selected traditional features as well as determined some derived features, which are generated from the traditional features, for batsmen and bowlers, (ii) formulated heuristics for clustering batsmen into openers, middle-order batsmen and finishers, (iii) formulated heuristics for relative ranking of batsmen and bowlers considering the current performance as well as the experience of each player, and (iv) have proposed two greedy algorithms for team selection where the total credit point of the team and the number of players in each cluster is fixed. Our proposed ranking scheme and algorithm not only determines the best possible team but can also determine the best alternate player if one of the target players is unavailable.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

8 extracted references · 8 canonical work pages

[1]

”Net - work centrality based team formation: A case study on T-20 cricket.” Applied computing and informatics 13, no

Dey, Paramita, Maitreyee Ganguly, and Sarbani Roy. ”Net - work centrality based team formation: A case study on T-20 cricket.” Applied computing and informatics 13, no. 2 (2017 ): 161-168

work page 2017
[2]

Prakash, C. Deep, C. Patvardhan, and Sushobhit Singh. ”A new Machine Learning based Deep Performance Index for Ranking IPL T20 Cricketers.” International Journal of Com- puter Applications 137, no. 10 (2016): 42-49

work page 2016
[3]

”A study on performance of cricket players using factor anal - ysis approach.” International Journal of Advanced Researc h in Computer Science 8, no

Shah, Sricharan, Partha Jyoti Hazarika, and Jiten Hazar ika. ”A study on performance of cricket players using factor anal - ysis approach.” International Journal of Advanced Researc h in Computer Science 8, no. 3 (2017)

work page 2017
[4]

”A factor analysis approach in per - formance analysis of T-20 cricket.” Journal of Reliability and Statistical Studies 6, no

Sharma, Sujeet Kumar. ”A factor analysis approach in per - formance analysis of T-20 cricket.” Journal of Reliability and Statistical Studies 6, no. 1 (2013): 69-76

work page 2013
[5]

Petersen, Carl, D. B. Pyne, M. J. Portus, and B. Dawson. ”Analysis of Twenty /20 cricket performance during the 2008 Indian Premier League.” International Journal of Performa nce Analysis in Sport 8, no. 3 (2008): 63-69. IPL T eam F ormation 7

work page 2008
[6]

David Turner, and A

Moore, Andrew, J. David Turner, and A. James Johnstone. ”A preliminary analysis of team performance in English ﬁrst - class Twenty-Twenty (T20) cricket.” International Journa l of Performance Analysis in Sport 12, no. 1 (2012): 188-207

work page 2012
[7]

Perera, Harsha, Jack Davis, and Tim B. Swartz. ”Optimal l ine- ups in Twenty20 cricket.” Journal of Statistical Computati on and Simulation 86, no. 14 (2016): 2888-2900

work page 2016
[8]

The best batsmen and bowlers in one-day cricket

Beaudoin, David. The best batsmen and bowlers in one-day cricket. Simon Fraser University, 2003

work page 2003

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.

[1] [1]

”Net - work centrality based team formation: A case study on T-20 cricket.” Applied computing and informatics 13, no

Dey, Paramita, Maitreyee Ganguly, and Sarbani Roy. ”Net - work centrality based team formation: A case study on T-20 cricket.” Applied computing and informatics 13, no. 2 (2017 ): 161-168

work page 2017

[2] [2]

Prakash, C. Deep, C. Patvardhan, and Sushobhit Singh. ”A new Machine Learning based Deep Performance Index for Ranking IPL T20 Cricketers.” International Journal of Com- puter Applications 137, no. 10 (2016): 42-49

work page 2016

[3] [3]

”A study on performance of cricket players using factor anal - ysis approach.” International Journal of Advanced Researc h in Computer Science 8, no

Shah, Sricharan, Partha Jyoti Hazarika, and Jiten Hazar ika. ”A study on performance of cricket players using factor anal - ysis approach.” International Journal of Advanced Researc h in Computer Science 8, no. 3 (2017)

work page 2017

[4] [4]

”A factor analysis approach in per - formance analysis of T-20 cricket.” Journal of Reliability and Statistical Studies 6, no

Sharma, Sujeet Kumar. ”A factor analysis approach in per - formance analysis of T-20 cricket.” Journal of Reliability and Statistical Studies 6, no. 1 (2013): 69-76

work page 2013

[5] [5]

Petersen, Carl, D. B. Pyne, M. J. Portus, and B. Dawson. ”Analysis of Twenty /20 cricket performance during the 2008 Indian Premier League.” International Journal of Performa nce Analysis in Sport 8, no. 3 (2008): 63-69. IPL T eam F ormation 7

work page 2008

[6] [6]

David Turner, and A

Moore, Andrew, J. David Turner, and A. James Johnstone. ”A preliminary analysis of team performance in English ﬁrst - class Twenty-Twenty (T20) cricket.” International Journa l of Performance Analysis in Sport 12, no. 1 (2012): 188-207

work page 2012

[7] [7]

Perera, Harsha, Jack Davis, and Tim B. Swartz. ”Optimal l ine- ups in Twenty20 cricket.” Journal of Statistical Computati on and Simulation 86, no. 14 (2016): 2888-2900

work page 2016

[8] [8]

The best batsmen and bowlers in one-day cricket

Beaudoin, David. The best batsmen and bowlers in one-day cricket. Simon Fraser University, 2003

work page 2003