Game 5, move 1 : Herd 6-5

[diane]

After Herds move, Linus rolled 3-1, and made the 5 point - as I just couldn't see him doing anything else at that point.

Herd now up again with this 6-5.

[Julia_H]

Spoiler

Both running moves look good, but 22/11 both moves the one more vulnerable to attack, and puts an extra builder in our outfield. Seems clear to me...

[close]

[KDP]

Spoiler

i dont think there is a huge difference between either plays.  at this point in the match we just need to win this game.  im more inclined to move the guy off the 24.  it safety's a checker and moves the man which is farther away than the other, which is often the hardest to get home. one  blot is better than 2  for me. 

[close]

[KDP]

one more thing, even though 13-2 is a candidate and legal play i cant honestly see anyone doing this.  the difference between 24-13 and 22-11 is minimal.

[ah_clem]

Spoiler

I've been dinged repeatedly by XG and gnu for playing lover's leap in similar positions.  22/11 is typically ~.05 better than 24/13.

The reason is that the checker on the 22 is more likely to be attacked, we don't really want another checker on the bloated midpoint, and the 11 makes a nice builder.  Add 'em up and you get something better than the auto-pilot play.

[close]

[Zorba]

Spoiler

This is an interesting choice with an early 6-5 that comes up every now and then with split backcheckers. Just run one man to complete safety with 24/13, leaving the other backman closer to the mid but also more in danger of being attacked; or run the more advanced backman around the mid, vulnerable to a few indirect shots but also useful as a builder, while the other backman sits safely back on 24, but a bit farther away from the mid?

The bots sure taught me to move the more advanced backchecker more often. So what are important factors that come into play?

• Number of indirect shots at the run-around checker: here it's only 6-4 for 2 shots.
• Usefulness as builder: all we got is an extra builder on 8 so far, so the builder on 11 is very useful
• Danger of advanced backchecker if playing 24/13: Linus already made his fivepoint, he'll be very eager to attack. The gammon-go-ish matchscore is an extra incentive to attack.
• How do specific numbers play for opponent? Strong duplication can be a factor favouring one play over an other. It doesn't seem very relevant here, attacking plays at our 22 backchecker are not really duplicated, nor is the 6-4 hit at our 11pt checker duplicated with the other play.

That looks like 3 factors clearly in favour of 22/11 and one neutral, so I'm pretty confident that 22/11 is best and 24/13 an error (but not a blunder).

[close]

[ah_clem]

No cubes this game, so we should be able to do rollouts immediately.  I'll hold off it the position looks "static" i.e. we may be faced with a simolar decision next roll.

rollout

Spoiler

3-ply analysis says that Lover's leap is about a .02 error, but a full rollout shows a much smaller difference - after 1296 trials it's still within the margin of error.   22/11 comes out on top, but not by much.






XGID=-b----E-D---eD---b-db-A-A-:0:0:1:65:4:1:1:5:10

    1. Rollout¹    22/11                        eq:-0.208
      Player:   48.27% (G:15.13% B:2.06%)
      Opponent: 51.73% (G:16.90% B:0.60%)
      Confidence: ±0.006 (-0.215..-0.202) - [90.7%]
      Duration: 18 minutes 39 seconds

    2. Rollout¹    24/13                        eq:-0.214 (-0.006)
      Player:   48.66% (G:15.47% B:2.10%)
      Opponent: 51.34% (G:18.24% B:0.58%)
      Confidence: ±0.006 (-0.220..-0.208) - [9.3%]
      Duration: 19 minutes 38 seconds


¹  1296 Games rolled with Variance Reduction.
   Moves: 3-ply, cube decisions: XG Roller


eXtreme Gammon Version: 2.10, MET: Kazaross XG2

[close]