Dear Losing Chess mailing-list,

I am pleased to report that 1. e3 b6 has now been proven to be a win for
White under FICS/International Rules in Losing Chess, and so 1. e3 wins
(recall that 1. e3 c5 was solved in Feb 2015).

Perhaps the main novelty of this proof is that 2. a4! ended up being the
correct route. I tried for awhile with 2. Ba6 lines, but they never worked
out (you can read some garbled history at my webpages).

I had first looked at 2. a4 in May 2015, and by Aug 2015 I had solved the
difficult 1. e3 b6 2. a4 b5 line, and the very difficult 1. e3 b6 2. a4 e6
3. Ra3 Bxa3 4. Nxa3 b5 (the largest in the end), and was only left with the
move 4... Qh4 in this latter tree. However, I managed to convince myself
(at first wrongly, due to using International-only TBs, though I then found
a FICS proof after a queen chase ending turned out well) that 5. h3 was the
way to go. If only I had heeded my own advice....  in fact, tonight in only
a couple of hours (128 cores) I was able to show that 5. a5 bxa5 6. Qh5! wins,
as the ensuing queen-races end up in pleasant endgames for White. However,
their scores are not *immediately* that good, so I never looked further. :(
The other Black try of 5... Qxf2 is also just an exercise of an hour or two.
As always, in retrospect, the 14-odd months were "wasted" on irrelevancies.

The final e3b6 tree size (counting transposition pointers) is 491933802,
compared to 228501054 for 1. e3 c5, and 211327499 for everything else.

Everything is available from my website:

I am looking into hosting a browseable version of the proof (including
4-piece TBs, which WinningGUI currently doesn't link to), but have no plans
for that yet.

As stated on the e3b6 page, I will also likely clean up some of the longer
reversible move sequences, as these should be unnecessary. I will also make
"lines" files (these are actually quite time-consuming with large trees),
and write up something for submission to the ICGA Journal in the next weeks.

Mark Watkins