Tuesday, November 15, 2005

Introducing...The ICC

The Concept...

Though there are various tables and numbers to rank a cricket team, there is no such concept as a chart or a plot of a team’s performance over time. The purpose of such a plot would be to evaluate the consistency of a cricket team: for example, even if India seem to be playing well against Sri Lanka right now, they have historically been a very inconsistent team. Alternately, even though Australia lost the Ashes to England this time, they have, over the last 5 - 6 years been the most consistent team in cricket.

The ICC, which stands for Indian Cricket Curve (or Indian Consistency Calculator) is a method I've developed to plot the performance of the Indian Cricket team, match-by-match. As of now, it is set up only for One-Day games. Four factors are taken into account while evaluating the ICCn, which is the ICC-number. They are:

1. Result – Whether the match was won or lost.

2. Margin of Victory/Defeat

The Net Run Rate or N.R.R. reflects both the above factors (Remember that it is positive for a victory and negative for a defeat). And in calculating the ICCn, instead of cumulating the NRR over an entire series, it is calculated for each individual match.

3. Relative Ranking Of the Opponent, which is defined as (Ranking of the opponent - Ranking of India), the ranking being evaluated according to the ICC rankings available here
. The absolute value of the ranking is used in actual calculations as will be explained later.

4. Home/Away game: Whether the game was played in India or away.

This number is then plotted with time i.e. match number. I've started from the current India- Sri Lanka series, so the first match there is match #1.



The Formula...

The formula used to calculate the ICCn is designed to reflect the true nature of the result, as follows:

1. If India win against a lower ranked opponent, it is less valuable than a win against a higher ranked opponent; conversely, a loss against a lower ranked opponent causes more damage than a loss against a higher ranked opponent.

2. A win at home is worth less than a win away; conversely, a loss at home causes more damage than a loss away.


The exact formula depends on the relative ranking and nature of the result. There are four scenarios in this case. We start off with a number: 10 x NRR, and then modify it based on the relative ranking (REL, which is the absolute value) and on whether it’s a home or away game.


1. WIN against an opponent ranked LOWER than India, or LOSE to an opponent ranked HIGHER than India

- We want the ICCn to indicate that this result is not that that big a deal since India were expected to win (or lose, respectively) anyway. Therefore, the ICCn is reduced by a factor as follows:

ICCn = [10 x NRR] / [(REL + 1)/2]



2. WIN against an opponent ranked HIGHER than India or LOSE to an opponent ranked LOWER than India

- Not only is this a win (or a loss, respectively) but it’s a win (or a loss) against (or to) a higher (or lower) ranked opponent, making it a really GOOD (or BAD) result. The factor, 10 x NRR, is therefore further amplified:

ICCn = [10 x NRR] x [(REL + 1)/2]


And finally, to reflect home ground advantage, a loss at home or a win away multiplies the ICCn by a factor of 2. Otherwise, the ICCn stays as it is.

As an example, consider the following situation: As of now, Sri Lanka are ranked 4th, India 7th and Bangladesh 10th.

1. India WIN against Sri Lanka (REL = 3) with a NRR = 1.4:

Home Game: ICCn = [10 x 1.4] x [(3 + 1)/2] = 28

Away Game: ICCn = 28 x 2 = 56


2. India WIN against Bangladesh (REL = 3) with a NRR = 1.4:

Home Game: ICCn = [10 x 1.4] / [(3 + 1)/2] = 7

Away Game: ICCn = 7 x 2 = 14


3. India LOSE to Sri Lanka (REL = 3) with a NRR = - 1.4:

Away Game: ICCn = [10 x -1.4] / [(3 + 1)/2] = -7

Home Game: ICCn = -7 x 2 = -14


4. India LOSE to Bangladesh (REL = 3) with a NRR = - 1.4:

Away Game: ICCn = [10 x -1.4] x [(3 + 1)/2] = -28

Home Game: ICCn = -28 x 2 = -56


So, in the given situation, losing at home to Bangladesh is the worst that could happen (ICCn = -56) while winning away against Sri Lanka is the best (ICCn = 56).



The Curve...

All said and done, here's the ICC so far, calculated for the India - Sri Lanka series.


As you can see, the number dips below zero only once, when India lost in that 5th game, but even then, it wasn't a very bad loss. The first two games were absolute slaughter by India of a team ranked 3 places higher in the ICC rankings; hence the curve takes off!



The Future...

The curve will (hopefully) have more meaning as more and more results are added to it. I'll certainly continue to update this curve as I follow the exploits of India as they take on South Africa, who are ranked 5 places above India, next.

A few other things: even though the formula seems to have worked well in the Ind-SL series, I'd definitely modify it if it seems to give 'un-representative' numbers for other matches. We'll see! (Already, it's easy to spot a shortcoming in the formula, in that it gives ICCn = 0 for a tied game which doesn't seem correct. If India tie with Bangladesh, for example, it is a BAD performance. But those are rare circumstances and probably won't affect the overall curve much).

Future updates may include a curve for test matches and a Cumulative ICC (which seems more applicable but I haven't had time to come up with a suitable formula for it). And if anyone has the time or inclination to apply this curve for other countries, feel free to do so! Just refer this blog wherever you publish it.

Comments and suggestions welcome!

6 Comments:

Anonymous Anonymous said...

Ok machi,first of all great analysis ,you have taken into account all factors just like the icc ranking does (and in spite of what people say its the best ranking system,rediff comes close but sometimes i get the feeling its biased to suit india),however
there is potentially one more bug in this method of calculation

In rain affected matches NRR is never taken into account.For eg if duckworth lewis system is applied ,the NRR of that match is not counted towards the series NRR.Infact if you try doing that,in some rain affected matches,NRR might give a positive rate for team losing.
That is one of the reasons,most ranking systems dont take into account NRR.
Btw why did u take into account NRR when all you wanted to measure was how consistently the team performs with regards to winning and losing?

Sumant

10:57 PM  
Blogger Santosh Sankar said...

Hmm, good spot about rain affected matches. Will have to think that one through.

The reason I use NRR is to factor in the margin of victory. If India just squeak past Bangladesh, for example, even though the result is good because we won, it will give a very very low ICCn, reflecting the fact that it wasn't a very good performance.

7:18 PM  
Anonymous Anonymous said...

Actually if u see ,ICC made a stupid ruling with regards to the power plays.Do you know they still havent got in a ruling for how are the power play rules to be changed in case of rain affected.Does the block of 5 overs get reduced to something say 2 or 3 etc?

One more major thing is,DL system is based on the old funda of 15 over limit and thats how their table is computed.With these new changes DL also has to be modified.

Fair enuf for the NRR thing.So for ur curve is a team said to play consistently bad or good,if the curve is below/above zero for most part?

Sumant

8:49 PM  
Blogger Santosh Sankar said...

Hmm, it'll be interesting to see what does happen when it rains! Actually, since the matches have been during the night (for me), I haven't been able to follow how the powerplay and supersub options work :(

Yes, about the consistency thing. If the curve is sinusoidal, for example, you can conclude that the team in wildly inconsistent. If it is just above zero for the most part, the team is consistently pulling off some close victories. And so on.

Of course, I'm curious to see how this curve develops in the long run. That will tell if the formula is good and if it indeed gives us a good picture of consistency

9:13 PM  
Blogger Anantha said...

Santosh, looks like the ABAQUS analyses runs are giving you enough time to come up with these concepts. Good work. And I think Sumant has made some valid points. But a good start on your end. Keep massaging it till it becomes perfect

3:32 AM  
Blogger Santosh Sankar said...

Don't mention ABAQUS to me :)

6:26 AM  

Post a Comment

<< Home