Hello all,

I'm far from being a great player but I play alot to try and improve my game. I analyse with gnubg and have noticed a repeated type of error in my game.

When I have the good fortune to roll doubles, I often make horrendous blunders and I don't know why - if I play them properly these great dice should allow me to capitalize this luck and to win the game.

But I regularly seem to make a hash of them - my rationale for making these moves is definitely off

Are there any concise 'rules of thumb', or tips that anyone can give me on how to play rolled doubles better, things to look out for, how to optimize their usage?

I've attached a screenshot and an example gnubg exported game with 3 of these doubles blunders. Wasn't sure which format to use so it's in *.sgf, *.mat and *.txt

Hopefully the attachments will show up as I'm a new forum member.

Any feedback is more than welcome.

Cheers!

kamikaze

Thanks Clem, this is exactly the sort of advice I was after. It's not just that particular match I posted but other matches tend to have 'blunders' whenever whenever I play doubles.

Like you said, there tend to be many more ways to play them.

Maybe it's also a symptom of playing against a computer; I think I'll just have a quick game - then click, click, click, Analyse to complete it as fast as PacMan or Space Invaders but that doesn't do much to improve my skills (apart from speed and instinct I guess).

Forcing myself to properly check and re-check all possible legal options even though there may already be a pretty decent choice right in my face is good advice. There may be something even better if you look longer!

If there are any other one-sentence tips for optimizing usage of doubles I'd really like to hear them.

Thanks again!

Wayne

An interesting and useful thread, with good advice from pck and clem. I advocate letting go of the mouse (the equivalent of putting down the dice cup) and saying "doublets" out loud, then sit on your hands! Not only do you have four moves to make instead of two, but there very often exists the possibility of switching to an entirely different strategy, e.g. converting a holding game to a race or exchanging a priming game to a blitz. The likelihood of making a blunder is much greater than usual, so tread carefully.
Some words of warning about accepting Gnu (or any other bot) analysis on its face value. If you can't immediately see why your choice is so inferior, then use the rollout feature. Bots make mistakes too and just like us, they make them more often with doublets. Also, it can sometimes happen with small doublets that the best play doesn't even make it into the list! This is because a bot analysing at say 2-ply on Gnu doesn't look at each of the dozens, sometimes hundreds, of legal plays at 2-ply. It skims through them all at 0-ply, discards any that don't get close to the "best" play and then analyses what's left. It can and does blunder at 0-ply, so the best play never even gets looked at on the higher ply and I have seen this happen several times with 1-1 and 2-2. So, use your rollout option and set it at a reasonably high level.
Clem's assertion that 24/21(2), 13/10(2) is the best way to play a 3-3 in the opening isn't true often enough to be valuable. Depending on what it is replying to and the match score, there are a number of other ways to play it that can be much better in the circumstances.
Please take a look at http://dorbeldaily.blogspot.com/ , you may find it interesting and useful, particularly as in today's post there is a 6-6 to play.

Agree completely, except for the fact that it was pck making the assertion, not me.

Like pck, I struggle with early 33s - and while 24/21(2), 13/10(2) is a reasonable way to reply to most openings it's not always best.  For instance , if you can POH, that's usually correct. Sometimes it's better to make the 5 point, and sometimes it's better to make two homeboard points.  The dilemma is that there are so many good things to do that you have to figure out which of those good things are most important.

If you look at http://bigwillgammon.site90.com/bgcardv2.pdf, which shows the "correct" replies to various openings, of the 30 usual openings 24/21(2), 13/10(2) is favored for only 11 of them.  And that's just money play - throw the match score into the mix and it's even more complicated.