Comments/Ratings for a Single Item

ID seems to undervalue Butterflies.

Note that jN is ambiguous, and I don't think I have implemented an expansion to all possible shortest paths like in the nN case. The XBetza sandbox doesn't seem to be able to handle it at all; it highlights no moves at all, even if there are adjacent pieces. Even for nN2 the expansion would not allow different path in the two steps; to would be a compound of a Maorider and a Moarider, but not mix Mao and Moa steps.

Play Test Applet does highlight the moves though.

Play Test Applet does highlight the moves though.

Indeed it does. The difference appears to be that the PTA uses betza.js, while the XBetza sandbox uses betzaNew.js. The former uses a separate (old) move generator for highlighting and for the AI; in betzaNew.js the highlighting is also done with the latter. It is the AI move generator that cannot handle the jN. So even though the highlighting seems OK, the AI would still choke on it.

The Betza n and j modifiers are still sloppily implemented in both move generators. The old one merely interpolates the leap at 1/3 and 2/3 of the length (rounded to the nearest integer), and tests if at least one of these two squares is occupied, or both are empty. This works for ADGH, but not for longer leaps (where some of the intermediate squares are not tested). And for N it would test both the W and the F square, which is good for jN, but would require both to be empty for nN. But the latter problem can no longer occur now that nN is recognized by the pre-processor, and replaced by afsK.

Th AI's move generator tries to calculate a unit step in each dimension as (5*total_step+8)/16 rounded down. This only works for a total step of (plus or minus) 2, 3, and 4, while for 0 and 1 it would give 0. Which is OK for a step of 0 (i.e. purely orthogonal leaps). But the step derived from (2,1) would be (1,0), and using that step to scan the path it would miss the target square, and get stuck in an infinite loop. Even if the condition for termination during the scan of the path would be made more robust, it would only examine the W square (jumping Mao). Since jN is ambiguous anyway, simply forbidding its use seemed the easiest solution to this problem. That n and j do not work on orthogonal or diagonal leaps of 5 or more is still an open problem, though. (But so far no one complained about that, and development of the Interactive Diagram occurs pretty much on 'on-demand' basis...)

The editor Bn Em has revised this page.