I was goofing around with my program, and had loaded a problem that was supposedly a win for Red, and set it searching. After searching to thirty or so ply, Milhouse was still convinced the position was drawn. I then tried examining it in Cake to 10 ply deeper, and it was still looking like a draw. I wondered what the analysis of the position was, so I tried to see if I could find a copy of Gould's book online. I found it first on archive.org, but the quality of the scan is pretty uneven: some of the pages are misaligned and cutoff. But luckily, I found that Google Books had digitized it.
For some odd reason, downloading the PDF seems to result in a file which has all the diagrams stripped out, but if you page through it using their online viewer, it works just fine. The PDF might include some elements which require real Adobe Acrobat to render properly, but the overall quality is quite good. Here's the particular puzzle I was looking at:
Here's a link to the page that contains the analysis. I'll try to present my "refutation" after I work through it a bit more.
I've been having a bit of difficulty with the Chinook endgame database, so I thought that since Martin Fierz was kind enough to release his endgame database as well as the code for accessing it, I thought I'd give it a try by making the necessary adapters to make it work for Milhouse. Martin's code is pretty good, but included a few extra Windows dependences that I didn't like, so it took me about an hour to get it to the point where it compiles, links, and initializes the database. We'll see how it works. Eventually, I'll replace it with my own implementation, probably including a DTC database when I do so, but it's obvious to me that I still have some problems with the core program, so this will serve as an interim measure. The bonus from this is that I also end up with 8 piece databases, instead of just 6 pieces.