PWC: Progress Report 2

My DMP PR is 5.15 after 700 1-point matches.

This is down from 5.19 in my last report. I'm not sure how meaningful an improvement this is, however. Over my last 100 matches I've seen my PR swing pretty wildly, from 5.03 on the low end to 5.28 on the high. Though my sense is that I'm making progress, with that much variance, it's hard to know what to make of it. I'm cautiously optimistic that this is the beginning of a long term trend.

I've come to group my errors into four rough categories: 1) small conceptual errors; 2) large conceptual errors; 3) technical errors; and 3) mental errors.

I've been spending extra time on the early game when reviewing my matches, and that's where I'm trying to clean up some of my small conceptual errors -- knowing when to favor splitting over building over running, for example. These are smallish errors (~0.020), but they add up since we face them at the beginning of every game. I'm also in the camp that believes a thorough understanding of the first several moves of the game provides the conceptual basis for much of what goes on in the middle game. I think I'm making some progress here.

Large conceptual errors are cases where I lack a basic understanding of what's going on in a position. Errors here can be quite large (0.80 or more) and they can add up quickly if the position is complex and doesn't resolve itself quickly. Emerging backgames come to mind. I'm not yet studying these in any systematic way and I'll  have to figure out an approach to doing so.

I'm not too worried about technical errors at the moment. They tend to be small and less frequent -- subtleties in the bear-off, for example -- and I feel I can make greater progress by cleaning up my larger, more frequent errors.

Mental errors are the most frustrating. A missed hit, failure to make a key point when you roll the perfecta -- these are just problems of awareness and focus, but they are typically quite costly. This is a particular weakness of mine due to health related issues, and it's an area I must clean up if I'm to make any real progress. To a large extent, it's about creating good habits: slowing down, looking for alternatives, noting good numbers before rolling, etc. It takes vigilance and discipline, and this is an area I feel is ripe for progress, since it doesn't require any backgammon knowledge I don't already have. I think most of my progress has been in this area.

Position 3: Hotspots

is Player 2 score: 0 pip: 175

1 point match

pip: 162 score: 0 is Player 1

XGID=-b-a--E-C---cE--bc-dA--A--:0:0:1:53:0:0:0:1:10

to play 53

eXtreme Gammon Version: 2.19.211.pre-release

I had a choice of making my 3pt on Brown's head or making his 5pt and bringing a fourth checker down to my 8pt.

Over the board, I felt my position was inflexible after 13/8 23/20. I imagined leaving blots in my outfield or inner board and ending up with three or more checkers behind Brown's main army. I saw that 8/3* 6/3 does two good things -- putting Brown on the bar and making a new inner board point -- and I wasn't too afraid of Brown entering and attacking me on the 20pt. I'd have the stronger board, and we'd be fighting on his side of the table.

I think that analysis is fairly sound, as far as it goes, but it lead me to the wrong play. Making the 20pt is clear, and in retrospect it seems obvious. The hotspot in this position is Brown's 5pt. The rest of the board is pretty quiet. I should be looking to resolve the hotspot, to either make the point or get off of it. 13/8 23/20 solves the problem, directly and permanently. 8/3* 6/3 perpetuates the problem by addressing it only indirectly and and partially.

The rollout:

is Player 2 score: 0 pip: 175

1 point match

pip: 162 score: 0 is Player 1

XGID=-b-a--E-C---cE--bc-dA--A--:0:0:1:53:0:0:0:1:10

to play 53

1.	Rollout1	23/20 13/8	eq: +0.096
	Player: Opponent: 54.82% (G:18.23% B:1.63%) 45.18% (G:10.69% B:1.04%) Conf.: ± 0.006 (+0.091...+0.102) - [100.0%] Duration: 6 minutes 09 seconds
2.	Rollout1	8/3* 6/3	eq: +0.048 (-0.048)
	Player: Opponent: 52.42% (G:23.77% B:4.27%) 47.58% (G:12.34% B:1.06%) Conf.: ± 0.007 (+0.042...+0.055) - [0.0%] Duration: 6 minutes 43 seconds
1 1296 Games rolled with Variance Reduction. Dice Seed: 29771930 Moves: 3-ply, cube decisions: XG Roller

eXtreme Gammon Version: 2.19.211.pre-release, MET: Kazaross XG2

Position 2

is Player 2 score: 0 pip: 132

1 point match

pip: 161 score: 0 is Player 1

XGID=-a---AD-D---bD---cde----B-:0:0:1:53:0:0:0:1:10

to play 53

eXtreme Gammon Version: 2.19.211.pre-release

XG made a clever move here that I would have missed. The normal follow up to a missed 5pt-slot would be 13/5. Here, however, the normal play is a big mistake.

The key feature of the position is Brown's offense, which is coiled like a spring and ready to take advantage of a great many rolls. 13/5 does nothing to address that and leaves Brown free to improve. After covering with the 3, the best 5 is 6/1*, taking away half of Brown's upcoming roll and preventing him from making offensive progress unless he rolls doubles. White's position is fine, but he needs the tempo hit to ensure that he's able to stay competitive over the next several turns.

Backgammon is largely a game of pattern recognition. But those patterns can march us into trouble when we follow them blindly. I find that many of my early game errors are due to my following normally sound patterns when the game has gone in an unusual direction.

The rollout:

is Player 1 score: 0 pip: 132

1 point match

pip: 161 score: 0 is Player 2

XGID=-a---AD-D---bD---cde----B-:0:0:1:53:0:0:0:1:10

to play 53

1.	Rollout1	8/5 6/1*	eq: -0.148
	Player: Opponent: 42.62% (G:12.59% B:0.85%) 57.38% (G:22.47% B:5.08%) Conf.: ± 0.006 (-0.153...-0.142) - [100.0%] Duration: 4 minutes 32 seconds
2.	Rollout1	13/5	eq: -0.214 (-0.066)
	Player: Opponent: 39.32% (G:10.27% B:1.09%) 60.68% (G:22.62% B:4.96%) Conf.: ± 0.005 (-0.219...-0.208) - [0.0%] Duration: 3 minutes 54 seconds
1 1296 Games rolled with Variance Reduction. Dice Seed: 2 Moves: 3-ply, cube decisions: XG Roller

eXtreme Gammon Version: 2.19.211.pre-release, MET: Kazaross XG2

DMP Opening Moves

I'm going to be looking at replies to opening moves (i.e. the second move of the game) at DMP soon. My approach to memorizing replies will be to identify a "default" move which is right most of the time, then learn and study the exceptions to understand when adjustments need to be made and why. Typically, the default reply will be the same as the best opening move, so I thought I'd throw up a quick post to review the best opening moves (per XG's Opening Book) at DMP.

Here's a diagram of the opening position for visual reference:

is Player 2 score: 0 pip: 167

1 point match

pip: 167 score: 0 is Player 1

XGID=-b----E-C---eE---c-e----B-:0:0:1:00:0:0:0:1:10

on roll, cube action?

eXtreme Gammon Version: 2.19.211.pre-release

DMP openings fall pretty neatly into four categories: 

1) The Pointers: 61, 31, 42, 53

No surprises here. Make the 7, 5, 4, and 3 points, respectively.

2) The Slotters: 51, 41, 21

Play down from the midpoint with the large die; slot the 5pt with the 1.

Note that 51 plays differently than for money, where it is correct to split.

3) The Runners: 65, 64

Run one back checker as far as possible. Note that 64 plays differently than for money, where most players prefer to split.

4) The Splitters: Everything else (i.e. 63, 62, 54, 52, 43, 32)

For 63, 62, and 32, the major split (large die up from the 24pt; small die down from the midpoint) is best.

For 54 and 52 the major split is not possible, so the minor split (large die down from midpoint; small die up from the 24pt) is "forced."

43 is the exception: the major split is possible, but the minor split (13/9 24/21) is best.

---

That's all there is to it. In future posts I'll be looking at adjustments that need to be made to these moves if your opponent plays first.

Update: I changed my mind. I had intended to do a post for each responding roll and discuss how it should be played at DMP. I've since realized that it's simpler to use XG to memorize the replies. Just set up a position with Op's opening move, then cycle through the dice and check the evaluation. Easy peasy. Once I get things under my belt, I may post some analysis of replies that are close or interesting, especially once I start contrasting DMP and money play.

Position 1

is Player 2 score: 0 pip: 105

1 point match

pip: 117 score: 0 is Player 1

XGID=---CCbC-CB-------cbbcc--A-:0:0:-1:62:0:0:0:1:10

to play 62

eXtreme Gammon Version: 2.19.211.pre-release

White is behind in the race, has a better board (no 5pt gap) and prime. My reflex is to make my 2pt. Why risk breaking the anchor and leave myself open to attack? Turns out there are a number of reasons.

First, making the 2pt weakens my offensive position. I have a nice prime, but I only have three spares to work with. I'll need those to make my 3pt, or to hit loose there if Brown steps up. Meanwhile, the 2pt doesn't do much for me. It hardly hinders Brown from stepping up, and while it might come in handy if I were able to put Brown on the bar, I won't have enough material to pull off an attack without compromising my prime, which is my strongest asset.

On the other side of the board, there is some danger in breaking the anchor. I could be attacked and distracted long enough for Brown to get over my prime. But there is plenty of risk for Brown too. As noted, I do have the better board and the better prime, so I shouldn't be overly fearful of mixing it up. And even Brown comes out on top in a hitting war, he will still have the problem of escaping his straggler before he crunches.

Brown's clearest path to victory, and therefore my biggest concern, is to simply hop my prime and run for it. Obviously, his 62 joker hangs over me like a guillotine. And making my 2pt does nothing whatsoever to address that.

The anchor breaking play (20/14 20/18) does, however. It would leave me a direct 5 shot (in addition to 33 and 63) at the fleeing checker. As a bonus, there is meaningful duplication of Brown's 2s, which he would need both to hit and to step up. Best of all, if Brown does nothing special, the game is practically gin.

I don't think this is a particularly difficult position, but I would have blown it (It was XG's move) by making the obvious play and failing to consider whether I was actually addressing the needs of the position.

By the way, if we change the position by advancing Brown's rear checker to White's 2pt so that White can point on his head, the anchor breaking play is still clearly best. Brown simply isn't much of a threat unless he gets to the edge of the prime or over it, and those are the possibilities White should preparing to counter.

Here are rollouts of both the original position and the variation mentioned above:

is Player 2 score: 0 pip: 105

1 point match

pip: 117 score: 0 is Player 1

XGID=---CCbC-CB-------cbbcc--A-:0:0:-1:62:0:0:0:1:10

to play 62

1.	Rollout1	20/18 20/14	eq: +0.544
	Player: Opponent: 77.21% (G:13.07% B:0.48%) 22.79% (G:4.56% B:0.18%) Conf.: ± 0.003 (+0.541...+0.547) - [100.0%] Duration: 1 minute 37 seconds
2.	Rollout1	8/2 4/2	eq: +0.468 (-0.076)
	Player: Opponent: 73.42% (G:13.83% B:0.75%) 26.58% (G:5.23% B:0.18%) Conf.: ± 0.003 (+0.465...+0.472) - [0.0%] Duration: 1 minute 19 seconds
1 1296 Games rolled with Variance Reduction. Moves: 3-ply, cube decisions: XG Roller

eXtreme Gammon Version: 2.19.211.pre-release, MET: Kazaross XG2

---

is Player 2 score: 0 pip: 104

1 point match

pip: 117 score: 0 is Player 1

XGID=---CCbC-CB-------cbbcc-A--:0:0:-1:62:0:0:0:1:10

to play 62

1.	Rollout1	20/18 20/14	eq: +0.553
	Player: Opponent: 77.66% (G:13.74% B:0.75%) 22.34% (G:4.98% B:0.28%) Conf.: ± 0.004 (+0.550...+0.557) - [100.0%] Duration: 1 minute 32 seconds
2.	Rollout1	8/2* 4/2	eq: +0.499 (-0.054)
	Player: Opponent: 74.96% (G:14.68% B:0.42%) 25.04% (G:4.54% B:0.32%) Conf.: ± 0.003 (+0.496...+0.502) - [0.0%] Duration: 1 minute 20 seconds
1 1296 Games rolled with Variance Reduction. Moves: 3-ply, cube decisions: XG Roller

eXtreme Gammon Version: 2.19.211.pre-release, MET: Kazaross XG2

Glossary

As I wrote my first post, I realized that I'll be using a lot of abbreviations that some may not know. To help with that, I've added a simple Glossary to the top menu. I'll try to keep it updated as I go.

Project World Class: Progress Report 1

This is my first post about Project World Class (PWC), but I've been working at it for some time. I'm currently in Phase 1, which means my focus is on getting my DMP PR down to 5.0.

My PR is currently 5.19 after 599 1-point matches against XG. I recall it being around 7 when I started this project a few years ago, but that's misleading since I stopped playing for a couple of years and only began again recently.

Since I resumed my focus on DMP, my PR has dropped fairly rapidly from the mid-6s to where it is now. In the past couple of weeks, progress has slowed noticeably and I think this last bit is going to be a bit of grind. My initial goal of reading a PR of 5.0 seems like a good one -- attainable, but requiring some effort.

My study consists primarily of playing and reviewing a lot of 1-point matches against XG. Now that I have the blog up, I hope to post some analysis of positions I have trouble with or find instructive. I also need to nail down my DMP openings and responses.

Opening play is something I've been lazy about. I've absorbed a lot of it through reviewing my XG matches, but you can go a long time between seeing a given opening/response and there are still a handful that stump me when they come up. I hope to take a more systematic approach to openings here on the blog, but that is the kind of tedious work that I have trouble following through on, so we'll see how it goes.