So, you're half right, which is more than usual! If you look at BJA3, p. 522, you'll see the betting effects of removal for games with various combinations of rules. For S17, NDAS, the ace is -.5944 and the ten is -.5081. So, although we're counting both as -1 in Hi-Lo, the ace ought to be counted slightly higher.
For the H17 NDAS game, the ace EOR is -.5311 (less than for the S17, as you mentioned), BUT the ten EOR is now -.5470, MORE than it was for the S17 and virtually identical to the EOR of the ace.
So, if anything, someone might claim that Hi-Lo is BETTER correlated for H17 games because the important tens and aces are both counted identically as -1, which correlates very nicely to their almost identical EORs, whereas for S17, the EORs are about 17% different. And, this suspicion is confirmed in the p. 522 Table D18, below, where the Hi-Lo overall betting correlation for the S17 game is 0.9652 while the BC for the H17 game is 0.9685.
Bottom line: your statement that "Hilo fits better for S17 games than for h17" is just plain wrong.
Don
Bookmarks