Forum Archive :
I've played hundreds of matches with Jellyfish and many times I've
noticed a strong reluctance for it to get into a backgame. (I mean here
a strict backgame, rather than a holding one: holding two or more
points, keeping good timing, while building up a strong forward
position, without making inner-board points, and waiting until your
opponent is home before hitting.) The general advice, which I try to
follow, is to only accept such a strategy as a last resort when there is
no other defence. They are so much fun, however, that I suspect I am not
alone in accepting a full backgame rather sooner than I should. But the
acceptance point of Jellyfish seems vanishingly low.
Magriel says of a position demonstrating the backgame that (I'm quoting
from memory) " ... at this point it would be a mistake to adopt any
other strategy". One of Robertie's books is almost totally made up of
backgames. I remember one game on FIBS where I had a backgame going and
we ended up collaborating, discussing each move, helping each other out,
totally fascinated by the varying problems each throw gave us. It was
the most enjoyable game I've had there! All this seems to make backgames
an enjoyable and acceptable strategy to humans - and yet the bots hate
them. Indeed their play against them plays into the backgame player's
hands, hitting men regularly and allowing recycling for no reason that I
can see. Is it the way they are trained?
Given that we all trust Jellyfish and Snowie rollouts, does this mean
the backgame is dead; a faulty strategy only kept alive in human v human
games by unconcious collusion between human players?
Bob Meixner writes:
The problem is picking the appropriate situation to play for a backgame.
Some players enjoy them so much that they steer to them at every chance.
When they win it is pretty; but the number of gammons(and backgammons) they
lose more than offset the occasional glorious win. As Robertie puts it...
"Back games are the goal-line stands of backgammon. A well-executed
goal-line stand may indeed save a touchdown. But a team that plays a game
from the shadow of its goalpost is bound to be a losing team." As you
point out in your post they are a last resort.
BTW, bots(especially Jellyfish) are not especially strong at backgames. I
suspect the complexity of the positions does not lend itself well to neural
nets. In fact, I believe the "use caution" option on JF is there because
some experts were able to run the cube to extremely high values against JF
and win large point swings because JF mis-evaluated back game equities.
Don't get me wrong. There are good backgame situations but they are
relatively rare, Imo.
- After an early blitz attempt (Daniel Murphy, Apr 1997)
- But they're so much fun! (Laury Chizlett+, Oct 2000)
- Checker problem (David Montgomery+, May 1995)
- Defending against a backgame (KL Gerber+, Jan 2003)
- Defending against a backgame (Michael J. Zehr, Jan 1995)
- How many men back? (Brian Sheppard, July 1997)
- Play for a backgame from the start? (Alan Webb+, Dec 1998)
- What is a backgame? (Daniel Murphy, Apr 2001)
- When to double (David Montgomery, May 1995)
- Which anchor is best? (Kit Woolsey, July 1996)
- Which anchor to break (Brian Sheppard, May 1997)
- Which anchors are best? (sebalotek+, Jan 2012)
- Which anchors are best? (Adam Stocks, Apr 2002)
- Which anchors are best? (Mary Hickey, Mar 2001)
- Which anchors are best? (Jerry Weaver+, Apr 1998)
- Which anchors are best? (Chuck Bower, Jan 1997)
- Which anchors are best? (Marc Gray, Nov 1995)