Forum Archive :
How should we learn at backgammon?
This is a question that has been gnawing at me for over a year now,
actually triggered by a question Ilia asked me in an interview he did
for his local newsletter (Madison) which was later pubbed in GV.
You often see new (novice/beginner) players ask this, for example on
the newsgroup (rec.games.backgammon). There are a series of rote
answers from the more experienced players: Play, play, play! Read,
read, read! Read, play, read, play!! Study annotated matches (or even
just recorded matches) of the recognized masters. Record your matches
and run them through bot analysis. Watch the masters LIVE. (This was a
big recommendation back in the 70's -- go to the NY Mayfair Club.)
Play against the bots.
I'm left feeling insatiated. These answers are, at best, only part of
the solution. You can do most (probably all) of these things and not
learn. But even with serious effort you learn inefficiently. There's
got to be more to it than this.
One of the problems is that often the best players don't even know how
they do it. (Sorta like idiot savants. :) Many have a natural
propensity to playing and learning. Even those who are good teachers
probably don't tell you how they do it. (I'm not accusing them of
hiding secrets. I suspect they just don't know how they learn; it
comes second nature.)
Many here on GoL are well past the early learning stage, and some have
been past it for many years. We plod along, thinking we're making
progress. But that progress is slow (at best) and very inefficient.
There has to be a better way.
If you've ground through this message counting on a profound answer,
I'm sorry to disappoint you. I don't have it. It's out there, but I
haven't found it. Some of you may have, so please share. Even if
you're not sure, share anyway.
I've recently seen a couple of tools which are very helpful and
probably part of a good learning program. One takes quite a bit of
effort (nothing wrong with that!) and requires more than one person.
Mary mentioned this a while back. It was shown to me by Perry Gartner
at INDY and he credits Magriel with coming up with the idea.
Basically, a moderator (wiseman?) chooses from MANY games, one in
which there is a series of difficult choices. (Typically about 15
plays of which 2/3 are tough decisions.) S/he uses a bot to roll out
all decisions and then presents a sequential quiz to a student (this
can be an entire class of students, in principle). The student makes
one move at a time, is graded on that move (preferably followed by
discussion), and then is told what play was actually made at the
table. The quiz continues from that resulting position. This is kinda
like Kung Fu (TV series). The master allows the student to err, since
we learn best from our own mistakes. But the master is there to
correct, lead, and even encourage.
A simpler tool is one I used today, and recommend you do the same.
Take a position where a play you made is wrong. (You can also take a
position where a play you made was right....) Alter the position a
minimal amount until your play is correct. Why is this valuable? Why
not just walk away knowing your play was wrong and the other play was
right? The reason is that often our mistakes are based upon sound
judgement, reasoning, or instinct. There is just one (sometimes
seemingly small) reason that the choice was incorrect. By leaving
feeling that your play was just *wrong* (and the correct play was
"clear" :) then you've possibly done more damage than not having seen
the problem at all!
Casper van der Tak writes:
You already mentioned this, but I think the method of making small
changes in positions in which one is unsure about the right move, and
then rolloing it out AND playing it out by hand is THE method to gain
more understanding about backgammon. One gets to know what aspects of
the position are relevant to the decisions and what aspects are not.
This only works after one has reached a certain level and experience
(read read read play play play I am afraid).
Positions to which apply this to can be selected from actual games and
sessions with bots - any grave mistakes, anything you are unsure
about. After doing that for a large number of positions, classify,
generalize and consolidate.
Chuck Bower writes:
Actually I didn't say all of this, but it's a very interesting
technique. It combines a method of old (hand rollouts) with a method
of new (bot rollouts). In JF (other bots, too?) you can even play
"interactive rollouts" in which the bot is your opponent AND variance
reduction is included in the final results.
Some oldtimers thought you could get "the right answer" with hand
rollouts, and that's true if the competing actions were sufficiently
separated in equity. But even ignoring that, you see things developing
in the process of the game which you may not have considered when you
chose your play/cube. For example, you might put a lot of value in a
prime (yours or your opponent's) only to find, by hand/interactive
rollouts, that the prime didn't stay solid for very long and the
defending side was able to wait it out.
Michael Strato writes:
Maybe there are too many answers and one must just plough ahead
following the methods that seem best. They say that even if you quit
your day job and dedicated your life to studying the game and playing
in tournaments one will never learn it all.
I believe those good teachers do tell us everything they know and at
least why they think they know it (it should not be a "just trust me
on this") because if they didn't, there would always be some other
expert jumping on his back for having taught someone wrong (which can
There are lots of players that read everything out there, study with
Snowie and discuss the game with others. Then they go out and win or
place in tournaments. Ilia Guzei is an excellent example. In 2001, he
played in the Intermediates and made it to the money once, in 2002 he
played in the Advanced division and made it to the money twice. And
now in 2003, he plays at the Championship level and has made it to the
money four times. Holy comoly! Ironically, Chuck, the same person that
asked you the question in the interview might just have some of the
those answers you seek :-)
In my way of thinking, one must first understand the very basic stuff
and acquire a visual understanding of positions. I know some people
might not believe or like this "visual thing" but to me every position
tells a story. Look at it like trying to understand the hidden meaning
of a poem (though similarly, we will not always get it right). Some
examples of understanding positions: When to hit or offer a hit (odds
and probabilities); when to race (pipcount); when to blitz (playing
safe or priming versus bold as in attacking or blitzing); and there's
We can read and discuss these aspects forever, and it's good to do so,
but where we get the real experience is playing the game. Especially
so, say many of the top players, if you play for money because the
lessons from your errors stay more etched in your mind if you had to
pay for them. Personally, I don't believe this too strongly, if you
are truly serious about learning, the blow to your judgement, pride,
ego, whatever you want to call it, is enough to make you remember your
errors (and you get to keep your money).
The hard part is really to understand why you made an error, and if
you can weigh the pros and cons and accept why you erred, then that's
how you learn. This may require putting aside your pride and ego :-))
It's the same when discussing a position; give your viewpoint, but
learn to accept when others, especially the majority of others see
things differently. And don't shy away, come back and discuss more
positions, you don't have to apologize for getting it wrong. I think
the questions or viewpoint of a beginner or intermediate player are
what can bring out the inspirational answers or opinions of the expert
Find a system for everything. Odds and probabilities are easy to
study, there's a simple chart for those available online. So let's go
to one more topic - the pipcount. One might be confounded by the 10 or
more different pipcount methods. Joe says this one is best, Jack the
other, and they give their (no doubt) good reasons as to why it is for
them. I think my pipcount method is simple and fast (tell me if you
disagree), using math one learned back in Grade 5 or 6. I count from
the furthest checkers and keep a running total by continually
repeating the last sum I reached in my mind as I do the math on the
next point where there are checkers. Here's an example of the starting
position: 2 checkers on the 24 point, 2 x 24 = 48, I keep saying 48,
48, 48, to myself as I move to the next set of checkers, 48, 48, 48
and 5 x 13 = 65, 48, 48, 48 and 65, = 113, 113, 113, and 3 x 8 = 24,
113, 113, 113 and 24 = 137, 137, 137, and 5 x 6 = 30, 137, 137, 137
and 30 = 167. You then need to leave that 167 recorded in your mind in
a separate place or create an easy system where you use parts of your
body to which you assign numbers, or your 10 fingers (equal to all
possible digits) and then tally up your opponents pipcount.
I believe that many errors are due to playing too fast. A position
needs more than a quick glance otherwise it is too easy to find a move
that you quickly justify to be correct because you see several of its
positive features immediately - this has to be correct, right? Make
sure you do ask yourself "right?" and "why?" Because in fact those
same features can make you oblivious to the value of other moves. Once
again, compare positions by scoring them by pros and cons, every
checker has a value (even a dead checker, its value is zero) and if we
could find a system (they exist, which is the simplest?) that assigns
values to checkers in scenarios this could be another way of measuring
where we stand in the game and what moves add or subtract from it.
And too often we look at only the positions in which we made errors.
Okay, it was a -0.36 error, not so bad, let's roll it out, and voila,
it turns out to be the right move. But what about those moves that
were analyzed as being correct and the second choice was off by a
smilar amount? Study what's wrong but also study why right is right,
if it really is.
Douglas Zare writes:
> We can read and discuss these aspects forever, and it's good to do
> so, but where we get the real experience is playing the game.
> Especially so, say many of the top players, if you play for money
> because the lessons from your errors stay more etched in your mind
> if you had to pay for them.
When I don't think I am at my peak, I have no qualms about playing
against a bot, or playing for fun. Then I have an excuse for my
errors: Perhaps I am not awake, or I'm not concentrating on the game.
When I play for money, I have no excuse for playing badly. Even if I
am not fully concentrating, or if I am playing too quickly, I still
made the decision to play. Any blunder represents a real weakness in
Props that I have played are vivid for me far longer than blunders in
casual play. To get the right lessons, it is important to analyze
things later with bots, as the results of the prop are often highly
misleading, and often the player on the correct side quits.
Real playing experience is necessary, but it is good to keep in mind
that people can play badly for decades. Suppose you are trying to
decide between plays that differ by 0.020. In order to learn which is
right from experience, you would have to be willing to try both plays
(or to find opponents who do), to hit similar positions tens of
thousands of times, and to be able to collect the data correctly. It
is very hard to get proper feedback from results, and it is harder for
some people than for others. It is often easier to learn from a bot or
> I believe that many errors are due to playing too fast.
That may be true. Errors from overlooking plays can be remedied by
playing more carefully and slowly, but I'd prefer to reduce the
problem by having a better method of choosing plays quickly. If I
overlook a play, what concerns me is that the correct play didn't jump
out at me. Sometimes that is because I don't have the right tactical
possibilities in mind, but more often it comes from not having the
right game plans in mind.
- Advancing beyond intermediate (James Eibisch, July 1998)
- Beginners' mistakes (Alan Webb+, Nov 1999)
- Best way for a beginner to learn (Koyunbaba+, July 2007)
- Committing to memory (RobertFontaine+, Feb 2011)
- Getting better than "awful" (Morph+, May 2004)
- How to excel in backgammon (Max Levenstein+, Aug 2011)
- How to improve (N Merrigan, Jan 2007)
- How to improve (Albert Steg, Feb 1996)
- How to improve cube handling (RealNick+, Jan 2011)
- How to learn and improve (Hristov, Aug 2005)
- Lowering your error rate (Stick Rice+, Apr 2009)
- Maintaining your game (Robert-Jan Veldhuizen, Apr 2005)
- Matchqiz and Jellyfish (Gilles Baudrillard, May 1997)
- Missing candidate plays (Klaus Evers+, Apr 2009)
- Most efficient way to learn (Stick+, May 2007)
- Practice and preparation (Ian Shaw+, Mar 2004)
- Practice/study plan (Marcus Brooks+, Nov 1995)
- Reference positions (Chuck Bower, July 1999)
- Study Methodology (Phil Simborg, Dec 2012)
- Study method (Jason Lee+, Jan 2012)
- Study plan (Tenland+, Nov 2012)
- Taking your game up a level (CW+, Aug 2002)
- Taking your game up a level (Ron Karr, Aug 1996)
- The backgammon cake (Daniel Murphy, Nov 1997)
- The best way to learn (Chuck Bower+, Oct 2003)
- Three steps to better play (David Montgomery, July 1998)
- Using Jellyfish tutor (Stephen Hubbard, Sept 1997)
- What more can I do? (Alison Wylie+, Apr 2000)
- Zen in the art of backgammon (Robban+, Aug 2009)