"We need to work on our defence," said John Wright, Derbyshire coach, after losing in the T20 Blast quarter-final last year. So many coaches now use the term defence when talking about bowling in T20. And if bowlers are defenders in this form of the game, it's important to look at what kind of defenders.

Football has two extremes with its defenders - lockdown players who barely leave the defensive half, and attacking defenders who help create goals as well as stop them. For years it didn't seem to matter which you were - both were paid far less than midfielders and strikers. That has slowly changed. But even a cursory look at the top-paid defenders in football shows you that defenders who can create or score goals while still being quality stoppers are worth the most of the men in the back half.

In cricket, you can see a similar pattern emerging. Lockdown T20 bowlers are great, as they make scoring almost impossible. But bowlers who can take wickets at the top might be even more important. You need both - five of one kind or the other of these bowlers and your attack will probably fail a lot. It's crucial to have a balance, but the effect of top-order wickets can't be overstated.

In the fifth round of the recent CPL draft, St Lucia Stars selected Rumman Raees with the 25th pick. Of the first seven live picks, Raees was the first bowler actually chosen in the draft (others had been retained earlier by their franchises), and he went for US$70,000. CricViz had him as one of their six best picks from the draft.

In top-level T20 (T20I and premier domestic tournaments) Raees has bowled 348 balls in Powerplay overs the last five years. In those, he has conceded only 5.1 runs an over, which is 2.4 runs less than the world average, and second-best anywhere after Scotland's Alasdair Evans. Raees strikes every 19 balls (five balls fewer than the world average), is hit for a boundary every 9.4 balls (the global figure is 5.3 balls), and he averages a staggering 16 runs per wicket (13 down on the world mark).

Of the bowlers who have delivered 50 or more overs in the Powerplay in the last five years, Raees, by his figures, is about as good as it gets. To be that frugal and still be striking is incredible. But of players on that list, Raees comes in at No. 124 for the percentage of his overs he bowls in the Powerplay. So though he is one of the best Powerplay bowlers on the planet, he bowls only 40% of his overs in that period.

Ishant Sharma averages 50 in the Powerplay, has an economy of over eight, takes a wicket every 37 balls, and yet bowls 60% of his overs in the Powerplay.

One of the most common inefficiencies in cricket is when bowlers are used. It is easy to rate Ishant as one of the worst T20 bowlers in the Powerplay. But he's as bad, if not worse, during the rest of the match. There is no right time to bowl him in T20. In the last two years, according to Mainuddin Ahmad Jonas' runs-above-average metric, Ishant is worth -10.6 runs a game compared to the average player in the IPL. (Jonas is a Bangladeshi computer-science student who is part of a new wave of amateur cricket analysts who are using ball-by-ball data and algorithms to work out who the best players are). If you have to use Ishant, up top is perhaps the best of all the bad options you have.

"One of the most common inefficiencies in cricket is when bowlers are used. It is easy to rate Ishant Sharma as one of the worst T20 bowlers in the Powerplay. But he's as bad, if not worse, during the rest of the match"

Most bowlers who have lasted a long time in T20 are good in at least one of the main three phases: Powerplay, middle, and death overs. But while you'd always like to bowl someone when they're most useful, that's not always possible. Raees is often one of his team's best two bowlers. So it's hard to use him all at the front. He has to bowl at the death because his economy there, in the PSL, is 7.7, which is extraordinary. Often, depending on match situations, he also bowls in the early-middle overs: in the PSL, his economy between overs seven and 12 is 4.6.

Only the best bowlers, however, are as adaptable as Raees. Most are suited to one - and if lucky, two - parts of a T20 match. But there is a reason that a bowler like Raees, a demon in the first six overs, should bowl there more: because that is when he can do the most damage. The new-ball phase is the most critical time in T20 cricket. Joe Harris from the White Ball Analytics blog wrote a piece on optimising the first over. It contains a graph that shows how important a wicket is in the first over. A first-over wicket brings down the expected first-innings total by over 12 runs. A wicket in the last over brings down the total by 1.8 runs, which makes it virtually only as valuable as a dot ball.

That's without even acknowledging that opening batsmen are the most significant players in T20. Over half (52%) of all T20 runs are scored by the top three. Nine out of ten of the biggest individual scores are by openers. The most prominent stars - Chris Gayle, Virat Kohli, Chris Lynn, Brendon McCullum, and David Warner - all open the batting.

Some of this is down to cricket's innate conservatism. Almost no T20 sides are using the most of their resources (the average number of wickets that fall per game in winning sides is dropping), and teams still coast in the middle overs far too often. But that makes the Powerplay more crucial.

When you take three wickets in the Powerplay of a T20 match, you have a 67% chance of winning. In the PSL that's 76%. The PSL is a bowler's league. Overall scoring rates in the BBL and IPL are 8.2 and 8.4 runs per over respectively, and the PSL is at 7.6. The global average economy rate for Powerplay overs is 7.5, but in the PSL it is 6.8.

There are only five bowlers in the last five years with an economy under 5.5 in the Powerplay overs. Two of them are Associate players, and the other three are PSL players (Raees, Mohammad Hafeez and Mohammad Amir). Over a third of the top 17 economy rates are of PSL bowlers. However, of the bowlers with the best strike rates, from Pakistan only Raees and Junaid Khan (striking every 18 balls) do better than once every 20 balls. That's not to say that there is a dearth of Pakistani early strikers, just that here are no more from Pakistan than there are from other places.

Bowlers like Wahab Riaz, who often get referred to as strike bowlers in T20, often aren't. Wahab has a strike rate of 15 in the PSL, which looks great, until you dig deeper. He takes 59% of his wickets in the last four overs, where they matter the least and where bowlers take them the most. The world average is a wicket every 11.5 balls in the death overs and Wahab's is 10. Wahab's excellent strike rate is down to the fact that he bowls 40% of his overs at the death (the last four overs), and is superb at it, but the strikes there just aren't worth much.

Mohammad Sami strikes every 17.5 balls in the Powerplay in the PSL (globally he's far higher). Raees strikes only every 22 balls, far slower than his overall record, but his economy is 4.7, so it's possible teams are trying to block him out. And so they should; players like Raees can ruin your day.

There is a twin effect in T20 when you lose early wickets: you lose a resource and batsmen try to consolidate, wasting deliveries. A bowler like Jason Behrendorff can ruin your match before the field spreads. Behrendorff might be the format's best strike bowler, with a wicket every 15.5 balls (his economy is 5.8), and he is also among those who have been used best (he bowls 71% of his overs in the Powerplay). By the time he comes back at the death, where he goes for two runs more per over than the average, he has often done the damage.

A lot of this comes down to who you bowl with. Behrendorff plays for Perth Scorchers, who have the best seam-bowling attack in T20. They have express pace in Nathan Coulter-Nile and Mitchell Johnson, the change-up king Andrew Tye, and quality spin from the Ashtons, Turner and Agar. They can use Behrendorff at the top and still be okay. Many teams don't have that luxury. Most have to survive with what they have.

Look at Worcestershire's Joe Leach. He has bowled 52 overs in the Powerplay in the last five years, and those have gone for 508 runs. And yet he has still bowled 37% of all his overs in the Powerplay. And, as bad as 9.8 runs an over may sound, Leach is probably not the worst.

Mohammed Shami has a Test bowling average of 28, and considering almost half his Tests are in Asia, that is amazing. In the IPL he has delivered 81 Powerplay overs, where his economy is nine per over, where he strikes every 54 balls, where he is hit for a boundary every four balls, and where he averages 81. The real questionable thing about Shami is that he still bowls 49% of his overs in the Powerplay. He is not an excellent bowler in the rest of the match, but he's a far better bowler in the middle overs.

Two teams in the IPL bid him up to US$462,000 before Delhi played their right-to-match card. His price was nearly double that of Behrendorff. It's not like for like (Shami is fighting for one of seven spots, Behrendorff for one of four as an overseas player in the IPL), but Behrendorff has never played in a league outside Australia, or internationals, despite being a first-six-overs destroyer for years. Some of that can be explained by injury, most of it can be explained as cricket teams not identifying what is important.

There are lots of theories about which bowlers are best at the top. Until recently, Samuel Badree aside, legspinners weren't used there much, and now they seem to be the new fad. Old-school cricket brains like tall bowlers, as they think the extra bounce causes more top edges with the new ball. CricViz analysis suggests the average release point for seamers is 198cm. Bowlers above that mark strike every 23 balls in the Powerplay and go at 8.23 an over; bowlers below it go at 7.93 and strike every 24 balls. And while the ball may not swing a lot in T20, any and all swing is gold. Behrendorff is tall but then so is Ishant. Raees swings it but so does Shami. The two successful bowlers of those four are left-arm seamers, but both arms have similar numbers in the Powerplay overall - left-armers are 0.05 runs cheaper per ball, right-armers strike 0.16 of a ball more often.

No one kind of bowler does well in the Powerplay, but by checking over splits, we can at least work out which individual bowlers do it best, or do it more often. With the effect of early wickets so massive, cricket needs to think a bit more about its defence.

With statistical inputs from Shiva Jayaraman