Backgammon Boot Camp: Errors

Walter Trice began work on what would become Backgammon Boot Camp in 2001, using the best technology available: Snowie 3.  Times have changed, and so have bots.  The following table shows all twenty-six positions where the current best bot, eXtreme Gammon 2, contradicts the old result.

Note that fifteen of the twenty-six errors concern backgames (Chapter 5).  XG2 is significantly stronger in this phase of the game than any prior bot.  Eight of the fifteen backgame positions are cube actions, and in all eight cases, the backgame player is stronger than previously thought.  Four initial cubes against backgames turn out to be no-doubles.  Three backgame passes turn out to be takes.  And one take after the backgamer has hit a shot turns out to be a big pass.  

Only ten errors felt significant enough to warrant additions to the original text.  This judgment turned out to correspond to an equity difference of .04 or greater.  These table entries are in bold.

This table is new to the fourth printing of Boot Camp.  If you own an earlier edition, print out a formatted copy to paste in your book.         

Pos # Best Play Error Size Note
1-01 10/6 9/4. 0.006 Better for the most likely variation: a close race.
1-02 13/10 6/4 0.016 No need to strip the 8 point.
1-17 Bar 23* 17/12 0.016 6/1 is great if you get hit on the ace; bad if you get hit on the deuce.
1-21 6/1 5/1 0.045 More costly to expose blots when approaching a cube-turn. Close decision if already cubed.
2-05 20/15(2) 6/1*(2) 0.037 Buriel is not so bad against an advanced anchor. Where else can your stacked checkers go?
2-15b 5/4 5/2 0.012 4/3*/0 buys only 3 additional gammons for 3 additional losses. Not worthwhile for money.
2-25 21/16 0.049 Duplication is wrong theme, as Black does not mind being hit. But he does mind being hit with 55.
4-03 Pass 0.010 Bare take reference is actually bare pass.
5-01 No Double 0.035 Close No Double against backgame
5-04 No Double 0.038 Close No Double against backgame
5-07 (31) 13/9 0.030 The 2-3 backgame better than 2-4. The midpoint is not so important
5-07 (55) 23/18(2) 13/3 0.006 Keeping the 8 is more valuable than keeping the midpoint.
5-09 B/23 8/4* 6/4 0.023 Not yet comitted to backgame.
5-20 B/20 13/9 0.020 Most pure play for building prime.
5-21 B/22 24/22 0.058 The triple anchor is stronger than was previosuly understood.
5-22 No Double 0.017 Close No Double against backgame
5-25 Take 0.078 2-3 backgame with timing can Take
5-26 No Double 0.014 Close No Double against backgame
5-28 Take 0.052 1-4 backgame with timing can Take
5-38 No Double 0.095 Volitility is super low. Since this is actually a take, no reason to double.
5-46 Pass 0.169 Trice was right to find the Snowie rollout Take "very surprising."
5-48 6/4 6/1 0.088 The prime is White's strongest asset.
5-51 16/8 0.065 No need to break anchor. Don't make the ace!
6-09 Redouble 0.118 Huge market loss after a 6.
7-08 Take 0.001 Close Pass becomes close Take
7-10 Too Good 0.022 Close Cash becomes close Too Good.