Imagine a game of go as the two players taking turns making moves of different simple values, and these will add up to give the final result. For instance, letâs say we have moves of values 5, 3, 2 and 1 points under Japanese rules. Black goes first and takes the 5-point move, then white takes the 3-point move, black the 2-point and finally white gets the 1-point move. The score from blackâs perspective is:
+5 -3 +2 -1 = 3
Now letâs say we look at the same game under Chinese rules. Here each of the values will be 1 bigger, so we get:
+6 -4 +3 -2 = 3
Each of the individual moves were 1 point bigger, but the final score is the same, because black and white made the same number of moves! If you retry it with an odd number of moves, youâll find a 1-point difference between Japanese and Chinese rules.
The swing value for Chinese rules you calculated in your first seki example was +7 - (-15) = 22. But both the 7 and the 15 has one of those hidden +1âs inside of them, which adds up to 2 extra points. This is what I meant by âan extra point in both directionsâ. Hopefully by the numeric example above itâs clear that throughout a whole game, all these +1âs in both directions basically cancel out.
As for the âhow big a difference between the counting methods is possible in local scoring?â question - if you have a bigger local tally you can get a larger difference in swing values.
For instance, letâs say we try to calculate the value of this ko:
Either black takes and then fills the ko, or white just fills the ko.
Under Japanese rules, the swing between the two outcomes is 1 point.
Under Chinese rules, the swing is 4 points.
Where did the 3-point discrepancy come from? In the first variation, black makes two additional moves. In the second variation, white makes one additional move. This adds up to a local tally of 3.
So if you want to keep making that number bigger, you can look at multi-stage koâs like this one:
(here the difference between the two extreme results is a local tally of 5: black could take twice then fill, or white could take once then fill)
However, the larger swing value doesnât correspond to anything meaningful in terms of gameplay! You donât usually need to prioritize endgame moves differently depending on the rulesets, not until you reach the very late endgame and the parity of dame starts mattering.
What is meaningful in terms of gameplay is the miai value. Those are consistently 1 point bigger under Chinese rules, no more and no less (ignoring seki and other weirdness for the time being). But this means the relative importance of different moves will stay consistent across the rulesets.
In summary: divide by the local tally and everything works out more nicely 
If dividing by 2 still feels a bit arbitrary for your examples, hereâs another way of looking at it, starting with a simpler situation:
Black can play A2 to make 2 points (letâs use Japanese rules for now). But itâs equally likely that white will get to play A2 and black will make 0 points. If we wanted to count how many points black has on the full board, giving black either 0 or 2 points is inaccurate, because we donât know which one it will be. Itâs better to say that black will get 1 point on average.
(If there happens to be two of these situations on the board, black will get one, white will get the other, so black will get 2+0 points, which is the same as the 1+1 we would have counted on average without knowing who gets what.)
So in the current position black already âhasâ 1 point, even without having played A2. Then by playing A2, blackâs score will go up to 2 points - a gain of 1 point. Similarly, if white plays A2, she will take away 1 point from black.
The current score is +1. Black can move it up to +2, or white could move it down to 0. The swing between the two outcomes is 2 points, but it turns out to be more generally useful to think of how far each individual move changes the score from the previous average - in this case, 1 point per move.
Thus under miai counting, A2 is a 1-point move (not a 2-point move as we might otherwise think!). The advantage of this way of counting is that it letâs us compare values between more different situations - for instance the multi-stage ko above. The total swing between the two extremes of black/white winning the multi-stage ko is quite large, but either player needs to spend several moves to get that full value. Meanwhile, the other player is profiting elsewhere. So in an actual game, what we care about is how much value we can get per move.
To bring it back to your first example, we could say that black already has 10 points on average (Japanese rules) here:
Then one of two things will happen: Black can play the triangle and gain an additional 10 points, or white can play the triangle and steal blackâs 10 points.
For Chinese rules, the average for black is 4 points, and from there black can move it up to 15 points or white can move it down to -7 points (sorry about switching your sign convention 
). The individual moves have value 11.
Hope that makes these 20/2=10 and 22/2=11 values seem a bit more natural, rather than something I just pulled out of a hat 
.
