Also, I would like to see a comparison using mirror images for left-handen players.
Also, it's often better to just take the easy 3 pins from the greek church (especially after a strike) instead of risking a complete miss by trying to cover it. Covering the 3 after a strike is +6...missing completely is +0.
On the other hand, if you are behind with only a few turns to go, taking the risk of taking on the more difficult throw will be the right choice.
The issue is that the distribution of plays that good players get on their second throw is different from that that bad players get. Let's exaggerate to show why:
Collect all throws on a random bowling alley for a year. I predict that the 1-2-3-4-5-6-7-8-9-10 configuration as the first throw has a higher success rate than as a second throw.
Reason? Players who end up with that second shot must have completely missed the first shot. Good players will occasionally get there (if you wait long enough you will even find a pro throwing it, maybe because he suffers a stroke during play), but bad players will get there disproportionally more often. Of all such second shots, 90+% may be played by the bottom 50% of players.
Similarly, I think the 7-8-9-10 that isn't in this dataset isn't impossible. I would guess children who are barely strong enough to throw the ball will occasionally throw it.
Thinking of it, bowling would get way more interesting if one could get more points for difficult combinations. Let's say you announce "I'll leave the 7 and 10 pins standing and take them out with my second ball", and go on to do that. Surely, that's worth more than a dull strike?
No pro will ever leave this though. Not possible throwing at normal speeds.
Edit: Or are you trying to say they aren't going for the conversion, just for knocking down as many pins as they can, when they get there? Does the article's data include whether the bowler was attempting the conversion? My expectation is no, the data covers all frames that include that split, without excluding ones where the bowler has indicated that they're not attempting the conversion.
Very few make it.
The 7-10 split occurred 3069 times in comparison to the 785 times of 4-6-7-9-10. The mirrored one only occurred 149 times but was converted the same number of times.
Is there really enough occurrences of these to determine which is hardest?
The other thing is that pros don't leave something like the greek church very often. It was left 785 times in 447k frames. The data on picking it up just isn't that significant. It also means it isn't going to get practiced nearly as much as something more common.
Also, there are more right handed people than left, and it shows in the data. mirror image shots aren't "easier," they are just easier for a certain handed player. As most are right handed, the shots that are easier for them will show up as easier in the stats.