Search This Blog

January 11, 2015

Reply Roll Errors

Readers may know I record some of my live matches.  The main reasons to date have been to:

1) serve as yardstick to my PR
2) find mistakes and study them

Expanding on #2:

Normally I run through my matches quickly and stop on large errors or smaller mistakes of interest.  I may ponder a specific error a bit, maybe write up my thoughts as a blog post or maybe not, and then move on.

I recorded about 50 such matches in 2014. I thought it might be useful to take a deeper dive on these errors.  To this end, I started yet another project.  I captured all large errors (whoppers) as well as smaller technical errors from the early and late game.  I further attempted to categorize these errors.  The intention is to target mistakes from specific phases of the game.

This is an awful lot of work and I am not sure what I getting out of it so this may end up as another uncompleted project.  At any rate, here are the first fruits from this effort.

Below are reply roll mistakes I made from live play in 2014.  These reply roll blunders are particularly irritating since this is one of the few aspects of the game I have actually studied.

*******************
#1. 21$-21?

is Greg Merriman

score: 0
pip: 164
7 point match
pip: 167
score: 0

is Bill
XGID=-b----E-C---dEa--c-da---B-:0:0:1:21:0:0:0:7:10
to play 21

1.XG Roller++24/23 13/11eq: -0.263
Player:
Opponent:
44.03% (G:10.96% B:0.44%)
55.97% (G:16.59% B:0.73%)
2.XG Roller++13/11 6/5eq: -0.268 (-0.005)
Player:
Opponent:
44.40% (G:11.59% B:0.51%)
55.60% (G:17.61% B:1.18%)
3.XG Roller++24/21eq: -0.300 (-0.036)
Player:
Opponent:
43.47% (G:10.00% B:0.43%)
56.53% (G:16.82% B:0.69%)
eXtreme Gammon Version: 2.10, MET: Kazaross XG2


I went for the big split with 24/21.  I have a vague memory of Stick stating that you need to think of the five point as basically being covered (somewhere in neighborhood  of 30 of 36 rolls cover).  Therefore splitting is urgent to fight the budding prime.

I must have confused myself as I am misapplying this nugget of wisdom.  Probably this applies to a roll like 32 or 52 but not this.  Instead I need to make a play which makes progress on both sides of the board.  Therefore either 24/23, 13/11 or 13/11, 6/5 are reasonable and appear to be essentially tied for money.

24/23, 13/11 splits where he doesn't want to attack you while preparing to do something good next roll.  The blot on the 11 is probably immune next turn since opponent would prefer to cover his 5 point.

13/11, 6/5 is more ambitious.  It unstacks both heavy points while slotting your own 5 point, and keeps the anchor for good measure which may be a bit safer.  Even though the numbers are diversified a bit I find the counterslot play more to my liking.

I have a rule of thumb that was to counterslot with aces if opponent slots and you miss.  Not sure how accurate that is but perhaps I simply forgot or got confused on this occasion.

***************

#2. 41$-51 at 4a 7a

is Tuvya F

score: 2
pip: 162
9 point match
pip: 167
score: 5

is Bill
XGID=-b----E-C---dE--ac-da---B-:0:0:1:51:5:2:0:9:10
to play 51

1.XG Roller+24/23 13/8eq: -0.359
Player:
Opponent:
43.83% (G:11.33% B:0.53%)
56.17% (G:17.17% B:0.70%)
2.XG Roller+13/8 6/5eq: -0.371 (-0.013)
Player:
Opponent:
43.89% (G:11.86% B:0.57%)
56.11% (G:17.67% B:0.98%)
3.3-ply24/18eq: -0.427 (-0.068)
Player:
Opponent:
42.87% (G:9.93% B:0.52%)
57.13% (G:18.71% B:0.70%)
eXtreme Gammon Version: 2.10, MET: Kazaross XG2

This time I went for the counterslot and this is apparently wrong.  This follows my rule of thumb and is probably why I played it.

Note that for money the plays are basically tied. So probably the score is influencing this decision.  The counterslot 13/8, 6/5 gets gammoned a tiny bit more which at opponent's elevated gammon value may be enough to swing the decision from a money score.

The computer likes 24/23, 13/8.  While the difference is tiny, the concept may be important.  That concept, I think, is to strive for an advanced anchor early when ahead in the match to try and keep the gammons down.  Mixing it up by counterslotting is probably the wrong strategy.

****************

#3.  41S-43?


is Ralph

score: 1
pip: 162
7 point match
pip: 167
score: 2

is Bill Calton
XGID=-aa---E-C---dE--ac-e----B-:0:0:1:43:2:1:0:7:10
to play 43

1.XG Roller++24/21 6/2* eq: -0.204
Player:
Opponent:
46.95% (G:11.82% B:0.46%)
53.05% (G:14.83% B:0.68%)
2.XG Roller++24/21 13/9eq: -0.242 (-0.037)
Player:
Opponent:
46.36% (G:11.89% B:0.50%)
53.64% (G:15.85% B:0.73%)
3.XG Roller+24/20 13/10eq: -0.244 (-0.039)
Player:
Opponent:
46.36% (G:11.59% B:0.52%)
53.64% (G:15.53% B:0.66%)
eXtreme Gammon Version: 2.10, MET: Kazaross XG2

I split with 24/21, 13/9.  The problem with this play is the power of the opponent's superbuilder on the 9 point.  Left unmolested, the man on the 9 point has too many rolls which make the bar, 5 point, or 4 point next time. 

Better was 24/21, 6/2*.  The idea is to distract from the power of the 9 point superbuilder. Hitting loose takes away half the roll while splitting for the anchor. It is a little confusing as to when to apply this concept and when not to.  Note for money on 54 the hit and split (24/20, 6/1) is OK and tied with the split (24/20, 13/8). So 54 and 43 appear to be the only ones where you hit and split facing the 9 point superbuilder.

******************

#4.  41S-51?

is Dan Smith

score: 3
pip: 162
9 point match
pip: 167
score: 2

is Bill
XGID=-aa---E-C---dE--ac-e----B-:0:0:1:51:2:3:0:9:10
to play 51

1.XG Roller++24/23 13/8eq: -0.147
Player:
Opponent:
46.60% (G:11.60% B:0.44%)
53.40% (G:15.34% B:0.61%)
2.XG Roller++24/23 6/1* eq: -0.201 (-0.054)
Player:
Opponent:
45.33% (G:11.14% B:0.36%)
54.67% (G:15.74% B:0.74%)
3.3-ply24/18eq: -0.181 (-0.035)
Player:
Opponent:
46.18% (G:10.49% B:0.45%)
53.82% (G:16.48% B:0.55%)
eXtreme Gammon Version: 2.10, MET: Kazaross XG2

I told you this was confusing. This time I sniffed out the 9 point superbuilder, combined with my brick of a roll, annd mistakenly went for the hit and split 24/23, 6/1*. While this idea has some merit, it is wrong in this specific position. As noted above, it looks like you only hit and split with certain 4's (54 and 43).

Best was the pederstiran 24/23, 13/8. This play isn't great, but has the virtue of not panicking and hoping for a better roll next time.

**************

#5. 65R-51 at 8a 8a?

is Felix Yen

score: 1
pip: 156
9 point match
pip: 167
score: 1

is Bill
XGID=-a----E-C---fE---c-e----B-:0:0:1:51:1:1:0:9:10
to play 51

1.XG Roller+24/23 13/8eq: -0.229
Player:
Opponent:
43.99% (G:10.18% B:0.29%)
56.01% (G:12.02% B:0.49%)
2.XG Roller+13/8 6/5eq: -0.242 (-0.013)
Player:
Opponent:
43.97% (G:10.90% B:0.32%)
56.03% (G:13.31% B:0.65%)
3.XG Roller+24/18eq: -0.273 (-0.043)
Player:
Opponent:
43.69% (G:9.03% B:0.30%)
56.31% (G:13.65% B:0.51%)
eXtreme Gammon Version: 2.10, MET: Kazaross XG2

I'm not actually convinced this one is even an error but the computer flagged it so I guess I should think about it.

I slotted with 13/8, 6/5. This is one of my rules of thumb, which is when opponent runs with 65 and you roll an ace, slot it.

Note that 13/8, 6/5 slotting is essentially tied for money with the splitting play 24/23, 13/8.

Perhaps a rollout would change XG's mind.  Or perhaps the slightly elevated gammon value at this score (.55) when combined with the extra gammon losses that slotting involves is enough to swing slotting to a small error.



No comments:

Post a Comment