This page is for improvements to the initial proofs.

Also, the corrections from fixing the double en passant bug are included.

It is not easy to say exactly what the criteria for improvements are.
In general, reducing the size is good, but also reducing the MAXREV
(maximum reversible move sequence) or maximum depth can be important too.

easy12.rev.bz2 993510 + 13191 nodes (1.1MB) f19e562db9e077b926c3998cbeae5e3f (2016 Oct 25)
e3c6.rev2.bz2 2013313 + 32724 nodes (2.2MB) 40e0976ed0e76ad8196ca9b201c0cda9 (2019 Apr 15)
e3Nc6.rev2.bz2 9107578 + 269736 nodes (10MB) 562e77c035d12318e6422278229dee90 (2019 Apr 15)
e3Nh6.rev2.bz2 15256430 + 433321 nodes (17MB) bf0c2562b2a7ddd3d6edcbb543d45a63 (2019 Apr 15)

29 Sep 2020: Klaas Steenhuis has found a simplification to part of
the 1. e3 c5 proof, using a fortuitous (yet now curiously irreproducible)
computer-based hint to consider 7. a3 instead of 7. Ne2 in the main Nh6 line,
reducing the relevant subtree from 38 million to 16 million. I then further
reduced it to about 12.5 million.
e3c5.rev3.bz2 168541428 + 8317374 nodes (220MB) 91583c03786117f6aa235bc11f43232f (2020 Oct 09)

Jul 2018: Vladica Andrejic has simplified some aspects of 1. e3 g5 that are
now included. Principally, 1. e3 g5 2. Ba6 Na6 3. Qh5 Bg7 4. Qh7 Bb2 5. Bb2!
is now under 19 million nodes, compared to around 30 million for Qxh8.
e3g5.rev3.bz2 23192197 + 1916326 nodes (38MB) e1e4fae6bbe09d09de2883a5e3a3d026 (2019 Apr 15)

Re-solving 1. e3 e6 is a nightmare. Over a week now of wandering through
6-unit tablebases, and no sign of nearness to completion. I don't like the
final result in some of the endgames, but improving it looks onerous.
e3e6.rev2.bz2 38694358 + 3152521 nodes (56MB) cec66f5f8c72bcdcd6db3674da3cb468 (2017 Feb 7)

With b5, I spent a long time on 10. Nge2 for Ba6/Qb8, but ultimately this went nowhere.
e3b5.rev2.bz2 69187717 + 2964725 nodes (85MB) bbf1215f5b459ee68c7d1600127726b6 (2019 Apr 15)
e3b6.no2e6.rev2.bz2 113985726 + 7152364 nodes (153MB) f482884d52f3786f13c258c9a1e35b9d (2019 Apr 15)
e3b6.2a4e6.rev2.bz2 257027998 + 27905361 nodes (453MB) 04cd282c1389c3e339c535739fe330bd (2019 Apr 15)

e3wins.rev4.bz2 693653102 + 53819992 nodes (1.02GB) 0cb847ae2371465d1708400eaac77966 (2020 Oct 09)

VERIFY WIN-TREE FAULT LOOPS
Loading win-tree PROOFS/e3wins.rev4
Tree size is 747473094 [5702mb]
With TBs
================================================
tb4 19579890 tb5 66004177 tb6 74234738
nomoves 28590994 = nopieces 14399817 + patt 14191177
trans 53819992 terminal 48170884 internal 645482218 forced 323080243
MAXREV 50 maxdp 250 (wtm 313733637 btm 287919646/331748581 internal)
OK