Osmosis of Batting Goodness to the T20 Format
Does a good Test batter make a good T20 batter? What aspects of a T20 game do Test batters, White ball batters and T20 specialists bring value to?
Note: I have used to the terms batter and player here to refer to male players. I look forward to eventually add in female player data to my analyses some time soon as well.
TL;DR: There is some de-husking of talent that happens in the funnel to get to Test level and as a result, all-format players provide a lower risk, but also potentially a lower reward option than really good T20 specialists. Also, ODI scoring rates are better indicators than Test goodness indicators in determining T20 success which is largely scoring rate based. But things are more polychromatic than a tl;dr can capture.
When we talk about contemporary batting greats, we often talk about the stalwarts of Test cricket — the Smiths, the Kohlis the Williamsons, the Azams. Sometimes, we bring in top ODI performers into the picture and talk about the Stokes-es, the Sharmas & the Dhonis. OK, one can contest this, and yes, there are several caveats to this, but you get the point. We expect these folks to also be the best T20 batters. Are they?
Back in April, when this year’s IPL (well, the first part of it) was on, I had responded to a tweet with an opinion that someone like Smith or Kohli who is a Test great, may have decent numbers and a volume of runs in T20 cricket, but aren’t necessarily as impactful as they are in Test cricket — That red ball Cricket & white ball Cricket are essentially two different sports and we shouldn’t presuppose that greatness will permeate across them. (Harsha’s tweet may have been more about the aesthetics of the innings as I cover later.)
I wanted to examine this. But honestly, I went in trying to gather evidence for my argument. However, as we’ll see here, things aren’t monochromatic.
I did two things: One, examine whether and to what extent Test & ODI batting performances translate to T20 performances, and two, explore different areas in which multi-format players and T20 specialists shine. I also talk a bit about selection below.
Let’s look at players who play all formats. There are 117 players who played at least 10 test innings at 6 of above, 10 ODI innings at any position and 10 league or international T20 innings since 2012.
The Batting Average
In Tests, the average is a stat most widely and often considered to be indicative of how good a batter is. So let’s stick to that and see how Test averages translate to T20 averages. Yes, averages don’t carry the same weight in T20s but let’s start here. We’ll add more T20-like stats to the picture later. Let’s throw in ODIs as well since these players have played all three formats.
So the correlation of T20 averages to ODI averages is a bit stronger than that to Test averages for these batters. The correlation to Test averages, isn’t all that bad though. A batter with a Test batting average of 20 has a T20 league average that is roughly 5 runs lower than a batter with a Test average of 50 — although there is a significant amount of variability here to make strong predictive claims. Note, that the shorter a game, the higher the value of a run. The value of a T20 run is higher than the value of an ODI run, which is higher than the value of a Test run as far as impact is concerned.
Beyond the Average to shorter format measures
Let’s switch from average score to what matters more in T20s — the speed at which runs are scored. Let’s use RPO for this. Also, let’s still continue to factor in the anchor role and the length of an inning in general. Let’s use Survival Factor for this. This is the likelihood of an average batter in the same circumstances having gotten out before the batter in question did. The higher this number is, the more unusually long the player in question tends to bat.
For these and other T20 numbers, I’ll use just the last 3yrs to factor out any trends in the overall approach to T20 innings.
The lack of correlation between Test averages & T20 RPO indicates that the top metric for batter evaluation in the two formats are not closely related. This is in line with our original hypothesis. But the difference in how ODI & Test averages correlate to the Survival Factor is quite interesting!
Let’s see how this translates to WPA (win probability added), both per innings (while the batter is on either end of the wicket) as well as per ball faced, relative to an average batter in the same circumstances.
The difference in survival factor correlation shows up in inning level WPA, but per delivery faced, both Test & ODI averages barely make a player more impactful. This is because averages (mostly for ODIs) are correlated more with the longevity of T20 innings than they are with RPO. This could also indicate that it is the other end that is actively adding a lot of the win probability i.e. the better players are playing anchor more effectively. The necessity for an anchor is another debate that I explored here.
Let’s look at approach to innings, using RPO in the non-T20 formats.
Unsurprisingly, ODI RPO is a good predictor of T20 league RPO. The slopes for ODI & Test relationships are similar, but the variation is much higher for Tests. The variation in the relationship for Survival Factor seems intuitively plausible.
To recap, we’ve seen that a) Better Test batters don’t make faster T20 batters nor necessarily T20 batters playing longer innings. b) ODI averages are correlated with longer T20 innings & ODI RPOs are correlated with T20 RPOs. c) Faster Test players also tend to be faster T20 players.
There is a bit of a thinking trap in c) — Test averages are correlated with Test RPOs (Better Test batters play faster in Tests). Test RPOs are correlated with T20 RPOs (Faster Test batters are faster T20 batters). But Test averages are not correlated with T20 RPOs (Better Test batters are not faster T20 batters). Just because milk prices are correlated to milkshake prices and sugar prices are correlated to milkshake prices, milk & sugar prices don’t have to be correlated.
In case it wasn’t clear, we have only looked at all-format, or essentially Test players so far, even when we looked at ODI stats.
Bringing White Ball Specialists into the Frame
Let’s look at exclusively white ball players — Players that have played fewer than 10 Test innings since 2012, but have the same criteria for ODIs and T20s played as before. There have been 157 white ball batters since 2012.
We see an overall similar relationship between the ODI averages of white ball specialists & all-format batters and their overall T20 metrics. However, white ball specialists have a lot more variance in this relationship, or in some cases as one might even argue, the lack of a relationship.
One interesting thing here is that all-format players have a slightly more significant relationship between ODI average and Survival Factor than white ball specialists, but this doesn’t translate to a more significant WPA per innings relationship as we saw earlier (when looking at all-format players’ Test & ODI averages). The per ball contribution to WPA of neither group has anything to do with their ODI average. So all-format players who are better at ODIs survive longer than better white ball players, but they don’t have more impact than better white ball players while they are out there. This is relevant to our expectations from the Kohlis and the Smiths.
So far we have seen indication that batter approaches in Tests & ODIs translate best to T20 performance indicators. None of this is counter-intuitive.
- Better Test batters don’t bat faster nor significantly longer in T20s.
- Faster ODI batters can be expected to bat faster in T20s. They also survive longer.
- Better ODI batters who have also played Test cricket tend to bat longer than better ODI players who are mainly white ball specialists.
And finally, enter T20 Specialists
Let’s now bring in the T20 specialists here (<10 Test innings, <10 ODI innings, ≥10 T20 innings played since 2012). After all, these folks should reign supreme in the format. India have had by far the most number of T20 specialists, thanks to the IPL.
Filtering this by batters in more recent times however, we see England dominate the scene with 2x as many batters than Australia, India & Pakistan. We’ll see this again in a bit.
For each group of batters, let’s look at the top line T20 stats we have been looking at and see how they stack up.
The all format players stack up at the top for every metric. So, are we saying that not only do better Test cricketers make somewhat more impactful T20 players but also that Test cricketers also make better T20 players in general?
We’ve already seen that ODI performance of Test cricketers The catch is that a lot of the best T20 batters happen to be international players and since the international game is dominated by Test & ODI cricket, these players don’t make it to the T20 specialist list. T20 specialists largely are domestic talent that has played some List-A cricket and maybe some domestic First-class cricket.
If we take the top 40 batters in the last 3yrs based on Addl. RPO over the expectation, Survival Factor, Addl. WPA/Ball or WPA/Inn (“or” not “and” — they don’t have to be top at everything), we get a list of 98 batters. Of these, 57 are T20 specialists, 19 are white ball batters and 22 are all format players. This is a highly skewed result, since there are 15x as many T20 specialists as all-format players under consideration.
Batters like Finch, Warner, Guptill, Gayle & M. Ali make up the all format group in the top list. Nabi, Pooran, H. Pandya, Pollard and Shoaib Malik make up the white ball group in the list and talents like Andre Russel, Philip Salt, Dan Christian, Azam Khan & SK Yadav make up the T20 specialists list.
A Bit on Selection
This shows that T20 specialists are risky picks and generally speaking, a small proportion of them are really good T20 batters while being a proven international player brings with it some reduced level of risk.
There is a self-perpetuating cycle here: Teams don’t have great tools to pick T20 specialists. So, a safe approach is to pick a tried and tested, but possibly sub-optimal international long format player. On the other hand, picking the next Glenn Maxwell (just a random top international pick that came to mind) equivalent domestic player requires a lot of amazing scouting and analytical talent, and luck. Picking an overseas T20 specialist comes with even more unknowns and cultural biases. Moreover, as the England & KKR analyst Nathan Leamon pointed out to me, “sample sizes are always too small (read — no great tools to pick T20 specialists) and so you get a bleed over from one format to another in terms of the inferences selectors are forced to make.”
The above narrative can make it seem like the reputation of international cricketers precedes them in T20 selection. This is not necessarily the case. Teams are taking risks in selection. Whether these are well understood or not, is another story. I can’t do this topic justice without digressing too much and so I will table this for now and follow up in another piece. But, of the 40 odd T20 batters who have debuted in T20 leagues & one of the 3 international formats since 2012, not a single one has started with Tests. A quarter started with ODIs and more than half, 27, started with T20 leagues.
There are generally also more non-international slots to fill — few overseas players allowed and a limited number of domestic players who play internationally. In the IPL, an average team has to pick about 0.2% of available overseas talent & about 12.5% of available domestic players who have played international cricket (this part is not picked but distributed exhaustively because there are so few of them) making up 60% of the playing XI. The other 40% comes from domestic T20 specialists. There are about 4x as many international long format batters who also play T20s as there are domestic T20 specialists. So, there is roughly 3–4x less competition in the selection of these specialists as there is in the selection of an overseas player. As a result, your average domestic T20 specialist can’t match up to the average all format batter (overseas batter, really, but a good number of them are all-format as I explained above). This is based on a back-of-the-envelope calculation from the IPL.
Differentiation of Batters based on T20 Nuances
If we focus on risk that paid off from our T20 specialist pool and take the top 90-100 T20 specialist batters to get about the same number of batters as we have for the all format and white ball groups, we find that they compete very well with the other two groups. Here is what the comparison looks like for the metric from above where they do the worst — survival factor.
We started off trying to see if good/great Test & ODI cricketers can be expected to make good T20 cricketers. We found that in fact, some of the better T20 cricketers out there are Test (all-format) cricketers. They also enjoy a higher success rate in T20 cricket than players who only play the short format. However, this has more to do with the de-husking that Test Cricket does for us than the actual skills employed/valued within the group of Test Cricketers.
So far, we have seen that:
a) Test batters represent quality and are low-risk choices for T20 games as well. However, better Test players don’t necessarily make better T20 players.
b) Test batters who are also better ODI batters do slightly better in T20s than pure white ball batters. However, they don’t have a higher contribution to wins.
c) Domestic T20 specialists face less competition (IPL example) and as a result their overall performance defies the “specialist” tag. However, the cream of the crop (including overseas specialists) competes well against international Test and ODI players.
We have looked at some high level aggregate metrics so far for our comparisons and analysis. T20 is a short game and several aspects can matter within a game. Let’s dive into these and see what differentiates the three groups. Then, we can look into those areas to guide us around strengths and weaknesses of the different groups. For this, I built a model to try and distinguish top T20 specialist batters, all format players and white ball batters from each other using the following T20 measures.
The model does 40–60% better than randomly guessing a given batter to be from one of the 3 groups . I took the model apart and looked under the hood to understand how these features are working to distinguish the player groups. Here are some key insights based on this effort.
A) T20 specialists are much faster starters while Test batters are careful and safer.
B) The top T20 specialists differentiate themselves a lot more against pace than they do against spin in terms of scoring rate. They do this via boundaries. Against spin, all groups have similar scoring rates, but contrasting approaches. This wouldn’t be as obvious if we looked at raw numbers instead of looking at the metric relative to expectation.
Against spin, the white ball specialists have a much higher dot rate which they make up for in boundaries. T20 specialists and all-format batters rotate a lot more. The all-format folks rotate the best ending up with a similar RPO even though they have the lowest boundary rate.
The dominance of those T20 specialists against pace is especially true against right-arm pace, but this could be showing up because there is more right-arm pace out there.
C) The differences in approach that we saw in B, also show up across phases of the innings. All format batters look to build in the middle overs, while T20 specialists look to attack in the PowerPlay.
The high RPO for the T20 folks in the PP also comes from a risky game. We saw this when we looked at the first 5 deliveries. Here it is again in terms of comfort and intent.
The T20 specialists usually play a riskier game in all situations. In the death, we already saw that all-format players catch up with others in terms of boundary rate. They are also the ones with the lowest dismissal rate. They really take advantage of the launch pad that they tend to create earlier in the innings. There may be an opportunity to move things along a little bit more for the all-format batters in the middle overs. They have a significant gap to the other groups and it isn’t clear that they make up for this in the death, nor that this is necessary (ask the WI). Here, we are just looking at what the strengths and weaknesses of each group might be, and so I am not going to say much more on sub-optimality.
D) The best T20 specialists differentiate themselves in scoring rate in chases. While setting a target, a small proportion of this group does better than others, but for chases the entire group stands out.
Circling Back to Steve Smith
If you were interested only in the questions that we were trying to poke at around how batting goodness permeates formats, you can stop here. If you are interested in looking at Smith’s innings mentioned in Harsha’s tweet given what we have just seen, read on.
Now that we’ve done all that digging, let’s circle back to the Steve Smith question in light of all this. Here are some of Smith’s advanced T20 Stats in the last 3 yrs. He has a below average RPO and a below average boundary rate, but also has one of the lowest dot rates — 1.69 dots/over. He is a good rotator. His WPA contribution isn’t exceptionally low, just about average. But his survival factor is very high and this translates to a WPA/Inn in the top 25%. So overall, OK, but not great. All of this is aligned with what we have seen above for the all-format group.
On that day — Apr 25, 2021 against Hyderabad at the Chepauk, Smith scored almost as fast as Pant on a sluggish wicket, 0.6 RPO faster than his usual self & ended with 4 runs above his average. This, while staying under his usual boundary rate, but also under his usual dot rate. In other words he rotated more than usual. Overall, an average batter in those circumstances would have scored 2.3 more runs than Smith did for the stint that Smith had & Smith played much longer (almost 2x longer) than the average batter. Smith also did not add anywhere near his usual 5+% to win probability. He added almost nothing.
Smith started faster than usual with a boundary taking him to 9.6 RPO in his first 5.
Smith usually has a higher RPO against pace (via a higher boundary rate) than spin. He rotates against spin — he has a lower boundary rate, but also a lower dot rate.
However he only faced 8 of his 25 balls against spin and scored only 6 runs from these. He would normally score around 9 against 8 balls of spin. This also means that he scored 28 off 17 against pace with 4 boundaries. He would normally score 23 with 3 boundaries in 17 against pace. So, Smith had an above average outing against pace and a below average one against spin.
He faced Khaleel Ahmed for 7 of his deliveries against pace and scored 16. On the other hand he faced spin largely from one of the best in T20 cricket, Rashid Khan, for 7 balls and took only one double, 3 singles and 3 dots. Rashid Khan on average would concede a couple of additional runs and about 0.46 fewer dots. The overall matchup against Khaleel Ahmed was +4.12 additional runs while the against Rashid Khan it was -4.66 runs. The third bowler that he also faced 7 balls off was Vijay Shankar. He scored 9 off these 7 with one boundary and one dot (-0.16).
The numbers and margins here are small. You can see how a single boundary off Khan can could’ve changed the narrative. But, for Smith, on that day, it did not come. This isn’t surprising. Smith would score ≤ 5 runs against leg break bowlers about 26% of the time. Khan would have ≤ 5 taken off him against RHB 42% of the time. Moreover, before this match, Smith had faced Rashid Khan in T20s for 26 deliveries and not scored a single boundary. He has 9 dots including a wicket in those 26 deliveries.
The takeaway here, is that given the data and the context, Smith’s approach to the innings or the result shouldn’t have surprised us. If anything, he over-performed in the first 5 against Khaleel. Looking back at it though, Harsha might actually have been commenting on the aesthetics of Smith’s innings. Smith faced about 10 slower deliveries and cutters in all given the wicket, and managed a six off one and as many as 5–6 nicks and misses. One of his boundaries was off a free hit and every other boundary was off an uncontrolled shot. He also gave a sitter that was dropped. Rashid Khan mostly bowled shorter sliders and googlies to Smith which he either picked too late and defended or tried to slog against — all in all, uncomfortably dealt with.
If you enjoyed this piece, check out more of my work at Boundary Line and follow along here & on twitter @amol_desai
