What is WinViz?

(Download the app) WinViz answers the first question anyone new to the game asks, the same calculation that any fan performs mentally when they look at a scorecard mid-match: “Who’s winning?”

How does WinViz work?

CricViz has built a mathematical model of cricket. This model takes the career records of the players involved, historical data from the venue and country where the match is played, and the current match situation, and simulates the remainder of the match.

It then produces two things –

WinViz gives the probability of each result.

PredictViz shows the most likely path of the remainder of the match, showing likely innings totals, declarations, chases and margin of victory.

Can you tell me more about how it does it?

The model uses something called a Monte-Carlo Simulator.

In maths, if a problem is too hard to solve simply, you can estimate the answer by simulating the situation a very large number of times and counting how many times the different outcomes occur. If, for example, someone gave you some oddly–shaped, irregular dice and asked you to find the probability of the dice landing on each of its faces then you could approach this puzzle in two ways.   You could try to calculate the answer using dimensions, centres of mass, stable and unstable equilibria and so on. Or, when that proved too difficult, you could just throw the dice a few hundred times and record how many times each face came up. This is essentially what a Monte-Carlo Simulator does.

The mathematics of a batsman’s scores is fairly robust. They obey a common pattern known as a geometric distribution. If you have a good idea what his underlying average score will be in a certain innings then you can very accurately predict the probability of him making any other given score.

For Test cricket the system takes a batsman’s record against fast and slow bowlers and adjusts these for the relative difficulties of the conditions, and the strength of the opposition bowlers (based on their past records). So, at high scoring venues averages will be adjusted upwards, against a strong bowling attack they will be reduced. This is used to predict what the batsman’s average score will be in the innings and therefore what the probability of him making each possible score is. We can then simulate the game after repeating this for each player in the match, randomly generating innings for each batsman, adjusting scores to mimic how teams behave in certain situations, and giving us a result.

We then repeat the process 10,000 times, and count the various results. If Team A has won 3,000 of the 10,000 simulations then we predict that they have a 30% chance of winning the match.

In ODI and T20 cricket the model takes into account the same factors (match situation, venue data and career statistics) and produces probabilities based on a resource-related measure. The expected and actual performance of players combine to project the WinViz outcomes.

How often is it right?

This is not really a case of being right or wrong. If we say England have a 70% chance of winning, we aren’t saying they are going to win, or even that we think they are going to win. We are saying that we think that if they played this match 100 times, they would win about 70 of them.

Likewise, if we had to predict whether a random card pulled from a deck of cards was going to be a spade, we would say it was 25%. We aren’t saying it won’t be a spade, but that it will be a spade one time in four.

How accurate are the WinViz probabilities?

WinViz has been tested against the last 500 Test matches played and across these the probabilities have been accurate in absolute terms to within a few percentage points, in most conditions.

In relative terms, however, it is more accurate than that. By which we mean, if your team’s WinViz score improves, then it is almost certain that their chances of winning have improved.