So I've been playing gnubg on its grandmaster level, as many of you suggested, and I've learned a ton as a result. At first I was really struggling, but now I'm definitely holding my own in cash game sessions, and I've found my cube decisions are getting much better as well. Gnubg has been rating my cubing at expert and supernatural levels, and I'm seeing the results translate well into my games against human opponents.

But once in a while, gnubg still pulls some crazy moves that I just can't really wrap my head around yet, and sometimes I start doubting that a human player would ever make such a play. For instance, I find gnubg playing to his 1 point way earlier than I think he should, and trying to hit blots on the 1 before it's really necessary. Here's a recent play that had me a bit stumped:



Here gnubg is white and rolls 32. His play here was 19/22*/24*. I just hated this play, but not because he hit me twice. Why bury that checker so early? It's obviously a good thing that he hit me twice, but not only am I a big favorite to bring both checkers in with one roll, but I might actually now come in on his blot, or come in with better position like doubles or 54 or something similar, and his checker on the 1 is now almost useless. So it seems that by making this play, nothing really changes with the board, except that now I have a chance to freeroll my back checkers to better position, while he effectively took a checker out of play and exposed himself to a return shot.

I know that white doesn't have many ultra-productive 32s to play here, but I would think that another option could be a little more productive. He could hit me on 22 and make the 14 point. He could make the 14 and slot his 5. He could break his back men and safety the checker on the 14. To me these are all other options, some better than others, that would make more sense than to double hit here and play to the 1.

What do you think? Am I completely off base here? Is this just a desperate play to try and get something going because my board is more developed than his? Is he trying to maybe stop me from doubling soon?

I'm looking at it through White's eyes, and my first instinct is 12/14 17/20. But then I see how far ahead Red is in this game, and how the only hope White has is to slow Red down and contain, and pray for a good building roll to follow. Just looking at that board from White's POV makes me feel like I'm back on the ropes, with very little built, two checkers locked up in Red's home board, and very few opportunities to hit. I can't argue why this is good from a mathematical perspective, but I can "feel" why hitting your checks is okay here, not to mention it's only slotting one silly checker in the White home board anyway.

I don't "hate" hitting two here, but I would also have played 11 3* (ie. 14 22*).

Once you go to the 3point that's already a little bit deeper in your board than you would like so going to the ace point to put an extra man in the air is not that much of a concession compared to the advantage of taking away your whole roll instead of just half. Another factor is hindering the development of your prime and buying some time to get those back checkers moving.


But a point is still a point and by just hitting one you can't do much about offensively anyway so I think, that I slightly prefer to be a bit more constructive.

I have noticed that the bots like these plays when you make either small split (24/23 or 24/22) followed by 41 and 32 respectively. I guess the idea is to just take a roll away from the opponent, and try to create some counterplay before the prime is overwhelming.I have always avoided these plays, sacrificing some equity because I just hate playing the resultant positions. they make me feel dirty.

The double hit is a great play for White right now. Consider:

• White's back men are in severe danger of being trapped. In most cases, Black should trade his bar point to make the five, and that means there are 10 rolls (11, 22, 33, 44, 12, 13, 23) that allow Black to make the five on his next turn if he is left unmolested.
  
  
• Since White cannot make a strong blocking point with this roll, he is not sacrificing an asset to double hit.
  
  
• White's distribution is such that his six point is overstacked. That is the point from which White wants to play.
  
  
• This kind of double hit, where you attack a minor split, is much different than a double hit after a major split. For instance, if your opponent splits to your five point, and then you roll 41, often it is correct to hit and slot the five point -- a point you want -- and to pass up the double hit.
  
  
• If White chooses not to double hit, he has an urgent need to play his two by splitting his own back men. Otherwise, it may be impossible later. So, if he splits, where's the three? Hitting 6/3* would be my choice, I suppose, but it will leave White with four blots and only his 6 point made in his home board.
  
  
• If Black hits White on the one point, White will have an easier time making an advanced anchor.

One thing I think you have right, or at least you imply, is that if Black misses, White -- whether as a human or as GNUBG -- will often wish to cover on the one point!

Hi,

I agree with much of the above, there is only one argument i missed so far:
With the black prime building white would love to have an anchor on the 5-point or the 3-point.
Right now only D2 or D4 will make this for white (chance is 2/36)
But if black hits the white blot on the ace point white will have 54 and 32 to make a better anchor:
chance is 4/36.
On the next move white might also get the opportunity to hit and split,
for example split with a 4 and hit another checker.
Another thing could be that white covers and hits next move
for example cover the one with a 5 and hit another checker.
After this both players have two inner board points and white has a blitz initiative.
Getting a 3- or a 5-anchor or blitzing are the only two game plans for white right now.

I think there are two very important factors that have been brought up. The first is that white is in bad shape, and needs some help somewhere. He's not in a position to build a lot of home points, and so he's got to get something going somewhere. Making his 1 is better than nothing, and hitting 2 black checkers in the slotting process just makes it better. If black rolls a 6 and leaves one dancing, white actually may have a small sliver of hope to complete the 1, and maybe even hit the one checker that did make it in. Or maybe he could hit his long shot double to make an advanced anchor. I think white knows he's a pretty big dog here, and so there's not much risk in making it a little worse to give himself a shot at digging out.

That brings me to the second part. White could be volunteering a shot knowing that an extra checker in the back would actually help his cause there. As was pointed out, those back white checkers are going to have a really hard time getting out unless they hit their longshot doubles to advance a bit. But having an extra builder/attacker back there might keep black a little more threatened and provide a little variety for building an anchor.

Next question: Black now has two on the bar and white offered a return shot on the 1. I like this spot for black to offer a juicy double. Agree?

I think a better question than "Why is this the right play" is simply "Is this the right play?" An XGR+ Eval favors 3* and 11, and I'm rolling out the move as played, that one, and 3* 22. Should have results in a few hours...

Hi,

Position +
Race +
Threat +

It's a double.

Is it a take?
Anchor, yes
contact, yes
5-prime, no

With no men on the bar the contact would be to small for a take
but with two on the bar it's a take.

Interesting. I think I like 3* 11 a little more than 3* 22, so I'll be interested to hear the results. Also, I titled this thread the way I did because I have gnu's settings ramped up to the highest levels, and so I just assumed it had it's deep dark reasons for making that play. But from the sounds of everyone's opinions on the subject and the potential conflicting results that you're getting, gnu may have been slightly wrong after all.

I would have played it just like GnuBG -- not because I like the play but because I like all the alternatives less. White is obviously in a lot of trouble and any move with this roll is going to suck for her. However, hitting on 3* and covering the 11 point is not better, and neither is hitting on 3* and splitting to 22.

First of all, hitting both checkers leaves only 11 return shots (any 1). Hitting on the 3 leaves 20 return shots (11, any 3, and any 2 except 62). She goes from getting hit about 30% of the time to over half the time.

Also the 11 point is not in much danger so covering it gains her VERY little. Moving to the 22 point isn't horrible since she hit you, but if White doesn't make this anchor on her next roll, which is about two thirds of the time, she will quite likely lose a gammon. Being under the gun here can get real nasty.