Hello everybody,
I am trying to improve my game and playing against gnubg.
The problem is there are play which the programs marks as bad or very bad and I can't see the reasons why my play is inferior.
Below are positions I gathered from a few days of playing which I don't understand,
I was wondering if someone could take a look at them and explain the right play.
I also have a general question about the program:
In some of the positions, after I view the list of the best plays, they are sorted by equity.
I sometimes find that plays with lower equity than the best play have a higher winning chances according to the program. Why is that?
I thought It could be because maybe they have higher gammon chances,
but in some of the position the gammon and backgammon winning ratio of
the other plays were also higher than the play with the highest equity.

Here are the positions:











And two screenshots showing my other question



Any help would be appreciated, thanks.

Wow -- quite a collection. I'd attach some numbers to problems in the future, just to make it easier for people to refer to.

I'll post some comments over the next few days. Let's start with #1 (Black has a 2-1 to play).

If Black owned the cube, the play would be 23/22 5/3, giving him a chance to get out and win the race if White doesn't roll a 1 or a 2, making his 6-prime.

The cube in the center means that White will double Black out next turn no matter what Black plays. GNU will therefore evaluate all plays as equal, I would think. (GNU has a lot of weird bugs, so you never really know what it will do. Most players rely on Extreme Gammon as far and away the best of the bots.)

As to the equity question, you're getting cubed out no matter what you play in those positions, which is why the equity is -1 on all of them. The winning chances are only if you incorrectly take.

Thanks, I disabled the doubling cube for now so things won't get complicated too much
for me, but I see now that it's there (only for the first position).
In the screenshots I posted with the equity there is no doubling cube involved,
so i assume it is just a bug that gnu thinks the computer will turn the cube on me and
therefore all the equities are -1?

It's a feature, not a bug.

If you want the cube truly ignored, you need to set it up as match play where both players have only one point to go to win the match. This is called double match point or DMP. I don't know if gnu can alter an existing position in that way though.

Position #2 (Black to play 3-2 in a 1-2 back game):

White would normally have doubled this long ago, so I'll put the cube on Black's side.

Black's game is pretty hopeless here. As White moves his last two checkers around, Black's home board will collapse and he'll be playing to get off the gammon. His best winning chances (and a very long shot at that) come when he makes a small block and White rolls some awkward medium doubles. I'd play 6/4 and 8/5, after which a 4-4 or a 3-3 by White actually give Black some chances.

The problem with 6/4 and 5/2* (which I'm guessing was your play) is that hitting slows White down and accelerates your collapse.

This would be a very good position to play out 10 or 20 times and just watch what happens.

Position #3: Black to play 2-1. Cube should be in the middle.

13/11 with the 2 is for sure. Unstacks and puts the checkers in a relatively safe place. Then the ace is either 24/23 or 11/10. I like 24/23 as it attacks the checker on the 17-point and gives Black extra chances to escape if White make his 5-point or 4-point. It also duplicates White's 6-4 everywhere.

Not as sure about this as with the first two. 11/10 is usually a small mistake but it's occasionally right. This might be one of those times.

Hey Bill, black is on the bar in position 3. That reduces the options

To me it is a choice between anchor on the deuce point or unstack the midpoint.

Robertie, in Position #2, would it be a mistake to play 4/1 5/3? This avoids slowing white down and doesn't hinder our chances of escaping the gammon?

Also in Position #3, I think you have missed the piece on the bar.

In position #4: 10/5 looks to be the best move. You remove a blot without losing a builder - and that builder is better placed to make a point on your home board.

24/21 10/8 looks to be an alternative, but this leaves you very vulnerable to a 6/2 6/4 2/2 4/4 5/5, which could be very costly.

Position #3: I did miss the piece on the bar. My apologies. Just play Bar/23 24/23. A solid position with a better anchor.

I like 13/10 6/4 in Position 4. All of White's threes make a strong point on the other side of the board, so the cost of slotting is very little.

3-2 doesn't look that strong for White in position 4, or am I missing something?

I almost always get wrong the positions where you have to slot in your home board.
These and position when you roll double and have to move a point in your home board lower in order to hit again.
It's black who is playing in all of the positions, or did I misunderstood you?

Also pay attention that from position #7 black is playing clockwise.

Position 5: (Black to play 4-4).

The cube could be anywhere here, so for argument's sake I'll leave it in the middle.

The pip count is key. Black will lead by 114-122 after the roll, and since he's winning the race he needs to start to disengage. Let's play 20/16 (2) with the first two, then look around. He could play completely safe with 20/16 with the third checker and something like 6/2 with the fourth, but that doesn't look quite right. White has two exposed blots and Black may well get a shot next turn. I like 20/16(2) and 8/4(2). Moving the 8-point instead of the 9-point should get an extra shot or two, and with the cube in the middle Black just need to hit to win.

This is fun. I'll take a crack at #6.

White has probably doubled already, so I will assume black holds the cube here. The choice with 54 is basically to hit or not. Hitting risks getting gammoned, but also gives chances to win.

Let's say black hits with 15/10* 15/11. The immediate return hits are all 3s plus 21 and 55, total 14 rolls, which are all probable gammons.

If white does not hit (22 rolls), how good are black's chances? This is trickier to judge. White still has some chances to escape, and black may need to leave another return shot in the containment effort (= more gammon chances for white). I guess black will win about 2/3 of games where white misses his first shot, which leaves gammons and wins close to even.

Alternatively black could hit with 15/10*/6, giving more cover numbers for the 3 point, and a chance to pick up another blot when white dances, albeit with duplicated 1s. However this also leaves white three additional shots.

What about not hitting? White escapes on half his rolls, and these are immediate losses. If white does not escape, black gets another chance to hit. There is the problem of what non-hitting move to actually play: covering the blot on the 3 point leaves no good 4, so black may have to leave that blot there, putting him back in a similar situation with 1s still duplicated.

Overall I will hit with 15/10* 15/11.

Bill Robertie mentioned that all White's threes were making strong points on the other side of the board (after Black's roll). I noticed that almost all White's threes indeed looked strong, but 3-2 in particular didn't look that strong.

I have to admit that it might look like nitpicking, but I wanted to be sure I hadn't missed anything.

Prroblem 6 -- Black to play 5-4.

Black should own the cube here, so he's got a basic question. Should he play safe with something like 11/6 11/7, or go for the win with 15/10* 15/11?

Black's trailing by 46 pips in the race before the roll. If he plays safe he'll almost never win the race, although he does have a chance to get another shot. (11, 33, 55, 21, 51, 61, 32, 52, and 53 all leave a shot next turn -- 15 rolls.) The problem is that 15 shots only translate into about 5 hit, and many of the hits will still leave winning return shots.

If Black hits now (15/10* 15/11) White will have 14 return shots from the bar (all 3s plus 21 and 55) and most of those will lead to gammons.

So, does Black play to win and take a big chance of losing a gammon, or does he play safe and mostly just concede the game?

The general rule here is that you play to win, because turning a loss into a win gains twice as much as you would lose by turning a loss into a gammon loss. When you turn a loss into a win, you gain 4 points, -2 to +2. When you turn a loss into a gammon loss, you lose only 2 points, -2 to -4. If the play seems somehow close, play to win. That looks like the case here.

#7 is fun because of all the choices. Make the 5pt or not seems the major choice. If you don't make it 13/9 24/16 15/11 looks pretty, duplicates some 4s and races when ahead in the race. If you make the 5pt. you can move a third man to the 5pt., move both blots or even make the 2pt. I think I'll forgo the 5pt nad make the pretty play.