Say Elo to the IPL

The IPL, about to enter its 10th season, pits teams that are somewhat evenly matched and compete in more or less homogenous conditions.  This coupled with player drafts every three years should result in a competitive and exciting tournament every season.  Six different winners in nine seasons suggests this has largely been the case.  Imran Khan looks deeper into how certain teams have performed throughout their history by considering their Elo rating – a system that evaluates teams purely on their results.

Elo ratings introduction

A team’s Elo rating indicates its relative strength compared to other teams.  When a team wins a game it gets transferred a certain number of points from the other team i.e. the total points for both teams stays the same.  This ensures the average across all the teams stays roughly constant.  Additionally, a stronger team will gain fewer points when beating a weaker team than the other way around.  At the very beginning, every team is assumed to be of equal strength so start off with the same Elo rating (in our case it will be 1500).

Let’s say we have a match between Team A and Team B.  Team A is stronger and has an Elo rating of 1600 compared to 1400 for Team B.  The probabilities of the teams winning can be calculated to be 76% and 24% respectively (assuming there are no ties).  If Team A does indeed win, their Elo rating will go up to 1604 and Team B down to 1396.  However, if Team B causes a small upset and wins its Elo rating will go up 11 points to 1411 while Team A goes down to 1589.  The intuition is that we would like to reward underdogs more for winning than we reward favourites for winning.  Note that the total number of points is the same before and after the game.

A brief history of the IPL

Of the 584 scheduled IPL matches to date, there have been 568 outright winners, 6 ties and 10 no results.  I have counted the winners of the super over after a tie as the overall winner and given half a win each to the teams involved in no results.  The plot below shows the Elo ratings for each of the 13 IPL teams to have existed over the course of every season.

Although this looks quite cluttered at first, we can distinguish some general trends.  The league was fairly closely contested in the first four seasons with 2010 the most tightly packed.  No team reached an Elo rating of beyond 50 points from the average of 1500 in that season.  In fact, 4 points separated 6 teams in the final standings with even bottom-placed Kings XI Punjab taking points off both eventual finalists.  From 2012, things started to spread out a bit more as Chennai Super Kings dominated and some teams, in particular Delhi Daredevils, started to fall away.

Chennai Super Kings

The Super Kings, the most successful IPL team so far, won titles in 2010 and 2011, have a win rate of 61% and have made the playoffs in every season in which they participated.

During CSK’s first title win in 2010, the Elo ratings suggest they were far from the best team for much of the season.  They scraped into a playoff spot on net run rate.  They also won as many games as they lost in the group stages beating teams that eventually finished in the top 4 on only two occasions.  In seasons 2013-2015, the Elo ratings suggest that there was a significant gulf between them and most of the other teams.  The finished top of the group in 2013 and 2015 but ultimately stumbled at the playoff stage.  What the graph above illustrates quite well is their performance peaking in the middle of each of the 2013-2015 seasons as opposed to at the end in 2010/2011.

CSK enjoyed the highest Elo rating of all time when they beat Kolkata Knight Riders by 2 runs in the middle of the 2015 season to take their rating up to 1609.  We can conceivably say that this CSK team was the best of all time during that period.  However, it was not to continue having staggered through the latter half of the 2015 season and losing in the final to the Mumbai Indians.

Daredevils and Warriors

In contrast, the Daredevils have been in a steady decline since 2010 after finishing 4th, 3rd and 5th in the first three seasons.  Apart from in 2012 where they exhibited a brief resurgence, Delhi have finished at or near the bottom in every season since.

Their lowest point occurred at the beginning of the 2015 season.  However, this was not the lowest of all time; the dubious honour being claimed by the now defunct Pune Warriors India.

The Warriors finished in the bottom two in each season of their fleeting existence winning only 27% of their games.  A run of 9 consecutive defeats culminated in an Elo rating trough of 1378 points when they lost to the Mumbai Indians near the end of the 2013 season.

Biggest upsets

We can use the Elo ratings to compare the relative strengths of teams before each match.  Of the 570 matches in which there was a result, 55.4% of favourites, according to Elo, won the game.  If a team with a lower rating beat a team with a higher rating, we can quantify how much of an upset this was.

The table above shows the ten biggest upsets defined by the difference between the ratings of the two teams.  The top match on the list was a surprise in more ways than one.  Chennai were coming off the back of their peak Elo rating and Delhi were near the bottom of theirs.  In that match, Delhi restricted CSK to just a run a ball and knocked them off with ease.  The Daredevils also feature in a further six of the matches in that list highlighting how, in the last few seasons, any victory was a seen as a shock.

Biggest contests

We can also consider which games were of the highest quality defined by the sum of the Elo ratings for both teams.

According to the system, the final of the 2015 season between Mumbai and Chennai was the highest rated match of all.  Both teams had quite similar ratings.  But as the graph shows below, Mumbai were the form team going into the final while Chennai were riding on their early-season performances.  It’s probably no surprise that Mumbai won by a pretty hefty margin.

The Mumbai/Chennai rivalry takes up eight of the ten slots in the list.  Mumbai follow a similar pattern to Chennai in terms of the Elo ratings although they are slightly out of phase.  They tend to get off the mark slowly at the start of the season then surge to peak near the end.  This is most noticeable in seasons 2008 and 2013-2015.

Improvements to Elo

The Elo ratings are based on a simple concept – a team is credited for winning and penalised for losing, while underdogs are credited more for winning etc.  The ratings can be refined by considering home advantage, major team changes after auctions for example and the margin of victory.  We can also dynamically change the importance of certain matches. For example, playoff matches may offer greater payoffs in Elo points while a team’s matches from several seasons ago can be discounted in value.

Imran Khan, @cricketsavant

3 replies
  1. Today Match Prediction
    Today Match Prediction says:

    I was more than happy to uncover this great site. I need to to thank you for your time due to this fantastic read!! I definitely enjoyed every bit of it and I have you bookmarked to see new information on your blog.

    Reply
  2. Islamic Products in Pakistan
    Islamic Products in Pakistan says:

    Due to advancement in science and Technology, many Islamic products in Pakistan
    are available in Pakistan through online shopping services. These products are easy to use and latest in technology. You can now have a Digital Pen Quran Reader at a very reasonable price Telebrand.pk The children today are much more demanding and want new ways of learning. For this purpose Islamic Products in Pakistan , which provides all the basic information regarding our religion to the child. We provide these products by simple process and that is to click our Telebrand.pk.
    Telebrand Pakistan
    Digital pen Quran Availble all Varity 8 Gb ,12 Gb .16 GB With cash on Delivery All Pakistan

    Reply

Leave a Reply

Want to join the discussion?
Feel free to contribute!

Leave a Reply

Your email address will not be published. Required fields are marked *