Eddie Cowan is right - we need better stats in cricket. But which ones?

I want to make at least a preliminary attempt at taking up the challenge. I've set out below a handful of statistics that I think would be extremely useful but that can be easily collected without the use of hawkeye or other elite technologies.

In putting these together I've considered the sorts of stats we already have, but also the traits that we are told make good players, or the outcomes we told that teams should aim to produce. The point of these stats is not necessary to define the 'best' players, but to offer a means of finding players with particular advantages over their peers.

I add the caveat that I think these would be most useful in test and first class cricket, but could be useful in any format.

I would welcome more suggestions in the comments!

We hear a lot that great batsmen can 'turn a game around' and this metric should measure it. It's very easy to calculate: divide the team's runs by the wickets fallen when the batsman comes to the crease and subtract from that the team's runs by wickets fallen when they get out. It should provide a fairly good measure of whether the batsman turned the game around while they were at the crease.

One limitation is that if a batsman is at the other end while a team mate makes match-changing runs, they would get credit. But if you think about it that is worth crediting, and since a player would be unlikely to consistently do that without making runs themselves, it's not such an issue.

Another limitation is how to measure situations where a batsman is not out. Not 100% sure how to solve that one, but it's not a big deal. 

Same idea as above. This would be interesting in its ability to capture whether a batsman's coming to the crease corresponds with the team scoring more or less quickly. I think this stat is neutral (except in Twenty20) but highly explanatory.

Edit: This stat needs to be tweaked because it's harder to change the RPO later in the innings (i.e. 10 runs in an over in the second over impacts the RPO more than 10 runs in the 100th over). Perhaps measure RPO of runs at the crease against RPO before coming to the crease.

We hear that bowling maidens is crucial to building pressure for the bowling side. So it makes sense to try to find out which batsmen concede the most. Perhaps this is best expressed by taking balls faced divided by 6 (i.e. overs faced) and giving us maidens conceded as a percentage of that (i.e. what percentage of overs faced are maidens).

This is related to, but I think more useful thank a measure of dot balls as a percentage of balls faced.

It is often said that some batsmen bring out the best in their batting partners. Usman Khawaja said that batting with Ricky Ponting in Johannesburg was invaluable. It was said that Steve Waugh brought out the best in tail enders by refusing to farm the strike while he was batting with them.

These sorts of claims would be very easy to quantify by measuring the average of the other batsmen while that batsman was at the crease. Just take the runs scored by those at the other end and divide by the wickets that fall while the batsman is at the crease.

We can broadly say that playing scoring shots in cricket involves more risk than blocking or leaving the ball. And further we can broadly say that a better batsman will find bigger gaps and hit them harder than a lesser batsman. 

So we should measure how many runs a batsman scores on average from shots that they score from. In other words, if it's not a dot ball, what does the batsman score on average?

We often hear that some bowlers might not take wickets, but that they help the team take wickets by creating pressure or chances at the other end. This isn't too hard to measure. We currently measure the bowler's strike rate - how many balls on average they need to take a wicket. So if we want to see whether they let others cash in at the other end, we should just take the strike rate of the overs immediately after those bowled by the bowler. 

For that matter we could take the average, economy rate etc of the overs immediately following as well. 

If maidens are so important, we should see how many of a bowler's overs are maidens as a percentage. Too easy.

We can add to that the percentage of consecutive maidens and triple consecutive maidens (which John Buchanan found massively increase the chance of a wicket).

The inverse of the scoring shot efficiency metric for batsmen. Look at how many runs a bowler concedes on balls from which runs are scored. In shorter forms, where economy rates are equal, you would probably prefer this number to be higher. That would mean batsmen are facing more dot balls and then hitting bigger shots - which broadly would mean taking bigger chances. In longer forms you'd probably prefer a lower number, meaning that there are fewer pressure-relieving boundaries.

Commentators are fond of saying of Graeme Swann that one of his strengths is his ability to take wickets in the first over of a spell. This is easy to measure. Just take the wickets in first overs of spells generate a strike rate from it. Again this could be done to give an average and economy rate in first overs of spells.

The result of this metric would be similar to an average but would better capture situations where someone leads an attack even if it doesn't produce good career stats.

The way to measure this is to take the average runs per wicket for each innings a bowler bowls and subtract it from the runs per wicket for the whole innings. Then sum those differences for a player's career.

For instance, if someone takes 2/50 in an innings of 300 then their difference is 5 (300/10 = 30  minus 50/2 = 25). Compare that to someone who takes 2/50 in an innings of 200, where the difference is -5.

Edit: this needs to be weighted by wickets (for average) or overs (for economy rate and strike rate). It should be weighted by the percentage (i.e. 10 overs in 100 over innings is the difference by .1, 3 wickets in 9 wicket innings is difference *.33).

So what this would capture is whether bowlers consistently over perform or under perform their team mates. In particular it would help to correct for bowlers who play a lot of their cricket in conditions that are friendly or unfriendly to bowlers. 

Again, this could be replicated for economy rate and strike rate.

These stats are probably not going to be that useful in finding out which players are the greatest. But I think they would do a huge amount to split players with similar surface stats. In situations where selectors are tossing up between lots of similarly-credentialled players - as they were before the first test - a discussion of these stats should illuminate important differences.