COW (Chance of Winning)

COW (Chance of Winning) is a program that predicts the outcome of an ODI or T20 cricket match from any given situation. More on that here.

Initial thoughts

When I started writing for HoldingWilley, I was new to COW and could not understand how it helped me understand an ongoing match better. It was fascinating at first, to see all the factors from ground conditions to auto-adjusting weightages; but, at the end of the day, they were just a bunch of numbers which kept changing with changes in the score. I didn’t get many insights before the match started. A predictor, according to me, should have told me the results well in advance, whereas COW gave me 100% for a team only at the end of a match and 90% only when nearly all was lost for the other team.

We know that a team is going to lose if they need to hit 50 runs off 10 balls. Why do we need COW?

Then, on the other hand, I liked the concept of having a percentage tell me how the game was proceeding. For example, an Irfan Pathan boundary halfway through an innings never told me anything before I started following COW. With this, I knew it increased India’s chances by 1% (why do we need to know that? read on, my friend…read on).

On more gruelling and otherwise dry situations in the match (read ‘scoring by singles’ when 200 runs are needed off 180 balls), I didn’t know if the team had to go for runs risking their wickets or play safe like they did and accelerate at the end. COW told me exactly where their chances stood, how much it went down (or up) by playing that way. There were other even more interesting things that COW was able to tell me, but I’ll come to them in a while, after talking a little more about these ‘Level-1’ insights.

The COW ‘Level-1’ Insights

Following COW is like following the score, but at another level. COW does a lot of math I don’t even want to think about, leave alone sit and calculate, and tells me which team has the upper hand at the moment. When we have two partnerships, say, A being 40 (20) and B being 75 (60), I would usually think that both were crucial and leave it at that. Now, I’ll know which one of them increased their chances more; A by 12% and B by 10%.

Another interesting thing you can derive from COW is the effectiveness of the bowlers. The figures of two bowlers could read the same “4-0-40-1” but that doesn’t tell us one bit about the bowling performance. No, it doesn’t. I can hear you asking ‘how’ already, so I am directly jumping to a scenario so that you understand it better.

Take two bowlers Sree and Zee from a twenty-twenty match, and let’s say they had the same 40 runs in 4 overs figures against, say Australia. Consider the following scenario.

Sree bowled overs 14, 16, 18 and 20 when the average required run rate was 12 (that would be the average of the required run rates for the four overs he bowled). He took the wicket of a tail ender.

Zee bowled overs 2, 4, 6 and 19 when the average required run rate was 7. He took the wicket of the most dangerous batsman in the opposition for a duck.

While one guy maintained his economy rate below the required run rate, the other guy stopped a batsman from decimating the target set by the bowling team. Can’t say who bowled better (or worse)? COW can. It’s incredibly simple for the user to calculate the overall effectiveness of the bowler just by seeing the total increase / decrease in the team’s chances during his overs. We can find the most impactful moments in a match using COW in a similar way.

Who is winning?  – How to predict the outcome of a current match using COW

The one final question in everyone’s mind, isn’t it?

I would now refer to two phrases that we so often hear in the commentary every day, “Confidence of a team” and “Pressure on a team”. Statisticians have, for long, been unable to predict the outcomes of matches to an agreeable percentage because they only took the career potential and recent results of a team and not the huge factors the aforementioned terms have been. COW, if you look more closely, quantifies the current mood in the match to a certain extent.

Let us assume that we have a T20 team “AA” playing in a specific tournament against a team “BB”. By looking at the COW records of the teams from the previous matches in the tourney, we can find out a few things about the teams quantitatively.

At any position in the match, Team AA could have, say, a 44% chance of winning the match (and hence 56% for Team BB). This means that, with the team’s potential and the available resources, BB have a better chance of winning the match. BB’s players would be confident of taking the match away and AA’s players would feel the pressure breathing down their necks…the quantity being 6% of their chances.

Interesting to measure such complicated terms in terms of chances? Yes? There are several other ways in which you can look at the COW numbers too, but that will require a comprehensive guide rather than an article. So, I’ll get back to the original point.

While this by itself could give us a few Level-1 insights, we have another level of insights (Level-2) which helps us predict the outcomes. This would be checking the current numbers with the data available from the team’s previous matches (or history, to put it in another way). If Team AA have gone on to lose every match in their chances went below 45% (usually much lower) without any recovery beyond that, then the 45% mark would be their pressure-point. Oppositions should work on taking it beyond that as quickly as possible and leave the crumble-under-pressure job to AA.

Similarly, we can find out that Team BB has always won a match once their chances have crossed, urm… 55%. Comparing this with the current data, we can ‘predict’ that team AA would lose the match. Though the COW percentage is 44% (pretty much near 50%) and the team would have an achievable target at that point of time, team AA would go on to lose the match at least 90 times out of 100. This percentage can change during a tournament and would become steady and highly reliable near the halfway stage of long tournaments like the IPL. However, each tournament, if with a majorly different team composition, is a different entity and the same numbers should not be used across them.

Of course, COW isn’t 100% accurate yet, nor can any program ever be...but I can say this from the numbers I see every day; it is almost there.

So if you find something wrong, something interesting, if you like it, or if you hate it, if you feel it could be better, if you feel there is something you like which is missing, drop us a mail at editor@holdingwilley.com.

The chances of us taking what you have to say seriously is 100% :)