Backgammon Then and Now...
Dave Hart
Recently, I have been teaching my 17 year old daughter how to play
Backgammon. Learning the "popular" opening rolls, basic cube play and
using away scores to determine some cube movements. No serious math or
trying to figure out the Janowski algorithm when it's 4 away, 3 away in
a 7 pointer... we'll move into those realms way down the road. Just
trying to make simple, solid checker plays or as MY mentor Pat Gibson
has quoted Ray Fogerlund as saying on checker play "Offense, Defense...
and Common Sense".
The other night we were discussing an opening 5-3, and I told her that
the "popular" move was to make the 3 point. She replied "...but I
thought it was important to make builders and that the order you should
TRY and make points in was 5, 4, 3..." etc. I replied that "computer
rollouts and such show that mathematically it is better to make the
three point, which rang a bell in my head (and the purpose of this
post). Remember, I do NOT want to inundate her head with serious BG
math at this point, I want her to ENJOY the game, get a decent method of
play and then let her begin to delve into the math WHEN she wants to, IF
she wants to.
I have talked to many older, seasoned players and I have heard them say
on many occasion that "X play was what I used to play years ago, but I
play Y play now", etc. etc. and that has caused me to wonder... in the
70's, certain plays were "popular" and now in the 90's they are not.
IS THE CHANGE DUE SOLELY TO THE ADVENT OF POWERFUL HOME COMPUTERS?
(Rollouts and/or analysis of large tables of games played (e.g.; Hal
Henrich's match database)) or is there anything else involved? I
understand that by using a computer you can obtain more accurate and
quick rollouts (e.g.; 12,960 games giving you a damn accurate
representation of the "future" in a given position) than doing it by
hand in the 70's. So now, when teaching my daughter and suggesting she
read Magriel, should I add the disclaimer that "read it for sound logic
on checker play as most pro players have done, but use the math results
and rollouts for the most accurate play" ????
So, based on that then is this conclusion correct: "Those who have the
better math skills will always be the better backgammon player in the
long run?". Current events lead me to conclude YES. My conclusion is
based loosely on these things:
1. Art Benjamin is one of the best backgammon players I personally
know. Art fairly kicked ass this year, taking the ABT #1 spot and
numerous other non-ABT titles. Art is a mathematics professor and even
has a book and video set on "Mathemagics" (it's wonderful, if you don't
have it, GET IT). He even does a math "magic show" at the Magic Castle
in Hollywood. When you play Art, you can see him REALLY get into the
math in his head over the board. Other popular players are also deeply
into the mathematics world professionally, Chuck Bower, Stephen Turner,
etc. Simply put, Art Benjamin can perform mathematics in his head at a
skill level obtained by few others. Art is one of the most highly rated
players in the game TODAY.
2. The most popular BG books today, "How To Play Tournament
Backgammon", "New Ideas in Backgammon" and "Can A Fish Taste Twice as
Good?" are based on serious mathematical analysis, "Tournament"
incorporating Hal Henrich's database to create Kit Woolsey's popular
equity table, "New Ideas" based on Jellyfish rollouts and "Fish" based
on equity tables. All leading me to the conclusion that "the
mathematical way is the right way".
Talking with high caliber players such as Ray Fogerlund, Steve Sax and
Joe Russell, they equate most things to evaluating match equity.
However, to accurately calculate AND USE match equity over the board,
you better be damn good at figuring out what your Game Winning Chances
(GWC) are to BEGIN with. Lending me to conclude that if you can't
accurately figure your GWC AT THE TABLE then all the other math you
perform is just a helpful indication (a guess). *OR* is the ability to
figure out your equity based on memorizing an equity table and then
performing the math with your away scores better than nothing? For
example, I can guarantee that Kit Woolsey's estimation of his GWC at the
table is better than mine, thus giving his final mathematical
calculation more value than my guess at my GWC then looking at my equity
calculation. In summary, since he has better knowledge in figuring GWC,
his calculations are more helpful, my knowledge of figuring my GWC at
the table is inferior making my calculations less helpful?
So, is my best bet in trying to become a better backgammon player to
memorize the hell out of positions to the point to where I can recall
them AND it's Jellyfish rollout #'s OVER THE TABLE, and then training
myself to be able to equate my current position to my "mental database",
thus giving me a fairly accurate GWC, THEN training myself to be able to
do the match equity calculations over the table. IS CONCENTRATING ON
THIS PROCESS THE ONLY WAY TO BECOME A CHAMPION CLASS BG PLAYER?
Anybody's thoughts are encouraged.... :)
Dave Hart
BG HUMOR: How do ya drive a BG player crazy? Put 'em in a round room
and tell them to stand in a corner, speak continously and not say the
word "JellyFish"...
Kit Woolsey
Dave Hart (dh...@niteshift.com) wrote:
: 2. The most popular BG books today, "How To Play Tournament
: Backgammon", "New Ideas in Backgammon" and "Can A Fish Taste Twice as
: Good?" are based on serious mathematical analysis, "Tournament"
: incorporating Hal Henrich's database to create Kit Woolsey's popular
: equity table, "New Ideas" based on Jellyfish rollouts and "Fish" based
: on equity tables. All leading me to the conclusion that "the
: mathematical way is the right way".
Thanks for the plug. However, I disagree with your conclusion. I
believe that for the most part backgammon cannot be attacked
"mathematically", and those players who attempt to quantify everything in
some way or other tend to lose the forest through the trees and make more
bad plays than those players who play by their gut instinct.
It is true that there is some math in match equity analysis. However the
most important figure, namely your game-winning (or gammon winning or
losing) chances cannot be mathematically calculated except for a few
simple end-game positions. There are simply too many variables. All we
can do is look at the position, take into account all the factors, weight
them as best we can, and come up with our best guess. Note that this is
exactly what the neural nets do -- only their weightings are often better
than ours.
For play decisions, what good would it do to memorize a bunch of rolled
out positions? You probably aren't going to see that exact position in
your lifetime. What is important from these positions is the priorities
which are involved. The rollouts, if they are accurate, can tell us
where our estimates of the priorities for a type of position may be in
error, so if see a similar sort of position in the future we can adjust
our personal weights and priorities and come up with a better
conclusion. This is what Hal and I tried to emphasize in our book -- not
the positions themselves, but the reasons which made the best move the
winner.
In actual play, I do very little calculating. Most of my plays are made
on my feel for the position, based on priorities and weights which I have
developed from analyzing similar positions. I believe this is true for
most of the top players.
Kit
John Goodwin
On Mon, 19 Jan 1998 07:19:59 GMT, kwoo...@netcom.com (Kit Woolsey)
wrote:
snip
>
>In actual play, I do very little calculating. Most of my plays are made
>on my feel for the position, based on priorities and weights which I have
>developed from analyzing similar positions. I believe this is true for
>most of the top players.
This brings up an interesting point.
How much is it possible to improve one's game over a short period of
time by reading, or using JellyFish?
There are fairly frequent posts from people who claim to have made big
improvements to their game very quickly, by reading certain books, but
if top players *do* play mostly by instinct, as I would have thought
they did, how much good will a book or tutor program do without a
*lot* of practice?
J.G.
Kit Woolsey
John Goodwin (J...@opticon.demon.co.uk) wrote:
: On Mon, 19 Jan 1998 07:19:59 GMT, kwoo...@netcom.com (Kit Woolsey)
: wrote:
A TON of good. How do you think the top players of past years improved?
By seeing what happened when they made their plays, and learning from
their (often bitter) experiences. Now, if you can grasp what a good book
or tutor program is telling you, you can profit from the "experience" of
the program or the author without putting in the practice yourself. Of
course, practicing and seeing for yourself what works and what doesn't is
still the best way to improve.
Kit
Unknown
>On Mon, 19 Jan 1998 07:19:59 GMT, kwoo...@netcom.com (Kit Woolsey)
>wrote:
>
>snip
>>
>>In actual play, I do very little calculating. Most of my plays are made
>>on my feel for the position, based on priorities and weights which I have
>>developed from analyzing similar positions. I believe this is true for
>>most of the top players.
>
>This brings up an interesting point.
>
>How much is it possible to improve one's game over a short period of
>time by reading, or using JellyFish?
>
>There are fairly frequent posts from people who claim to have made big
>improvements to their game very quickly, by reading certain books, but
>if top players *do* play mostly by instinct, as I would have thought
>they did, how much good will a book or tutor program do without a
>*lot* of practice?
>
>J.G.
In my experience, the thing that made the most difference was neither
JellyFish nor a book, but Messrs. Woolsey and Heinrich's Matchquiz,
which lets you play through every move of 40 or more matches between
very good players, comparing your chosen move to theirs, and then to
Woolsey's.
Having played lots of backgammon over a period of 10 years, I frankly
say that before I went through Matchqiz I was not sure "what was going
on" in a lot of positions.
While I have never made it to the top grade, I do believe that nobody
who draws against me thinks he is getting a soft match; I have put out
a lot of top players. Whatever form I have achieved, I owe to
Matchqiz.
deekay
John Goodwin
On Mon, 19 Jan 1998 22:21:39 GMT, kwoo...@netcom.com (Kit Woolsey)
wrote:
>John Goodwin (J...@opticon.demon.co.uk) wrote:
>: On Mon, 19 Jan 1998 07:19:59 GMT, kwoo...@netcom.com (Kit Woolsey)
>: wrote:
>
>: snip
>: >
>: >In actual play, I do very little calculating. Most of my plays are made
>: >on my feel for the position, based on priorities and weights which I have
>: >developed from analyzing similar positions. I believe this is true for
>: >most of the top players.
>
>: This brings up an interesting point.
>
>: How much is it possible to improve one's game over a short period of
>: time by reading, or using JellyFish?
>
>: There are fairly frequent posts from people who claim to have made big
>: improvements to their game very quickly, by reading certain books, but
>: if top players *do* play mostly by instinct, as I would have thought
>: they did, how much good will a book or tutor program do without a
>: *lot* of practice?
>
>A TON of good. How do you think the top players of past years improved?
>By seeing what happened when they made their plays, and learning from
>their (often bitter) experiences. Now, if you can grasp what a good book
>or tutor program is telling you, you can profit from the "experience" of
>the program or the author without putting in the practice yourself. Of
>course, practicing and seeing for yourself what works and what doesn't is
>still the best way to improve.
That certainly wasn't the answer I expected. From your response, I get
the impression that you thought that I was questioning the worth of
books or tutors per se.
This was not the case. What I really wanted to know was; from the
experience of others, what is a good ratio of 'book learning' to
practice?
Perhaps this ratio changes as one moves from beginner through
intermediate to advanced.
J.G.
Chuck Bower
In article <34C103...@niteshift.com>,
Dave Hart <dh...@niteshift.com> wrote:
(lots of stuff, some of which I snipped)
>Computer rollouts and such show that mathematically it is better to make
>the three point, which rang a bell in my head (and the purpose of this
>post).
>I have talked to many older, seasoned players and I have heard them say
>on many occasion that "X play was what I used to play years ago, but I
>play Y play now", etc. etc. and that has caused me to wonder... in the
>70's, certain plays were "popular" and now in the 90's they are not.
>IS THE CHANGE DUE SOLELY TO THE ADVENT OF POWERFUL HOME COMPUTERS?
>(Rollouts and/or analysis of large tables of games played (e.g.; Hal
>Henrich's match database)) or is there anything else involved?
>So, based on that then is this conclusion correct: "Those who have the
>better math skills will always be the better backgammon player in the
>long run?". Current events lead me to conclude YES. My conclusion is
>based loosely on these things:
>1. Art Benjamin is one of the best backgammon players I personally
>know. Art fairly kicked ass this year, taking the ABT #1 spot and
>numerous other non-ABT titles. Simply put, Art Benjamin can perform
>mathematics in his head at a skill level obtained by few others.
>Art is one of the most highly rated players in the game TODAY.
>2. The most popular BG books today, "How To Play Tournament
>Backgammon", "New Ideas in Backgammon" and "Can A Fish Taste Twice as
>Good?" are based on serious mathematical analysis.... All leading me
>to the conclusion that "the mathematical way is the right way".
>Talking with high caliber players such as Ray Fogerlund, Steve Sax and
>Joe Russell, they equate most things to evaluating match equity.
>However, to accurately calculate AND USE match equity over the board,
>you better be damn good at figuring out what your Game Winning Chances
>(GWC) are to BEGIN with. Lending me to conclude that if you can't
>accurately figure your GWC AT THE TABLE then all the other math you
>perform is just a helpful indication (a guess). *OR* is the ability to
>figure out your equity based on memorizing an equity table and then
>performing the math with your away scores better than nothing?
>So, is my best bet in trying to become a better backgammon player to
>memorize the hell out of positions to the point to where I can recall
>them AND it's Jellyfish rollout #'s OVER THE TABLE, and then training
>myself to be able to equate my current position to my "mental database",
>thus giving me a fairly accurate GWC, THEN training myself to be able to
>do the match equity calculations over the table. IS CONCENTRATING ON
>THIS PROCESS THE ONLY WAY TO BECOME A CHAMPION CLASS BG PLAYER?
I think this post was meant to generate a newsgourp dialogue and
I'm surprised that this hasn't occurred. Maybe everyone is as busy
as I and haven't had the time to respond. Maybe after Kit posted
people decided "What could I possilby add?"
Dave's question is a VERY OLD one, but has been cast in a somewhat
new light. I really think it has two parts: 1) How important is it
to understand the details of computer rollout results? and 2) How much
over-the-table calculation is required to be a top level player? I
don't consider myself a top player (and what's worse, neither does
anyone else!) but I still may be in a decent position to answer.
I find it convenient (for this argument, anyway) to break
backgammon activity into two parts: training and play. (This is
analogous to sports--sometimes you practive; sometimes you compete.)
During play there is a direct opponent and in tournaments there
are many indirect opponents. Matches have time constraints and
you just can't spend 4 hours playing an 11 point match! When you
are training (especially if training against a robot), you can
take all the time you want to analyze, think, experiment, etc.
You can record matches on FIBS (for example) during competition
and replay them later, studying any and every position if you like.
In training you can (and I believe, SHOULD) run positions
through Jellyfish. For play decisions, once you have significant
statistics (and this is one important thing to understand about
rollouts--when do you have sufficient statistics?) you pretty
much know the answer.
Yes, you may want to play the position out by hand or against
JF to make sure you're following a good game plan, and or to gain
confidence (or check for errors) in JF's play. If you have
time and the position is one which may recur--not necessarily
EXACTLY, but close--it may be worthwhile to spend a lot of time
on a single position (or on a group of similar positions).
Anyway, once you feel JF is playing the position correctly,
then the RELATIVE equities of the different rollouts will tell
you the correct play.
Cube decisions have extra layers of complexity. There are
two reasons for this: 1) JF rollouts are cubeless, except for
the "limited cube" level-5 rollouts, but even these require
considerable adjustment (much of the time). 2) For match cubes,
you need to understand how the match score affects the decision.
So even if JF's rollout is trustworthy (and that is usually the
case from my experience) you often still have some pencil
pushing to get to the answer. This is where the Janowski theory
and the early work of Keeler, Spencer, Zadeh, and Kobliska enter
in. (Recall, too, that they are dealing with money play only.)
So far I've been talking about backgammon TRAINING.
I think it is important to learn some of the math (which, BTW,
is no more than high school level, and MOST of it just add,
subtract, multiply, and divide)! What about during the match?
Do we need to turn into Mathemagicians?
Here I use my match playing experience. In a typical match
I probably (on average) go through one or two calculations (but obviously
sometimes it's zero and occasionally more than two). There are a lot
of shortcuts which allow you to skip calculations. For example,
early in a match (moderate to long match), small cubes are virtually
identical to money cubes, so you just have to ask "is this a money
double" or "is this a money take". Money cubes are benchmarks that
most players have a feel for.
Even late in a match, you can rely on memory to indicate whether
you should be doubling early, taking late, etc. Often a cube comes
over and you think "I would take this at money, and at this match
score I'm taking later than normal, so, no need to calculate, just
SCOOP IT UP!" I'm surprised how often this happens. Of course I
could be misevaluating, but no number of calculations can make up
for this (since I'd be calculating with bad input).
The key is that DOING YOUR HOMEWORK during training) pays off
dividends in competition. You will (without going out of your
way) end up memorizing match equity tables, drop points, doubling
points, gammon fractions (or equivalently gammon prices), etc.
Dave mentioned Art Benjamin (and SRET and me) as examples of
BG people with significant math backgrounds. Yes, it helps, but
advanced degrees are not required. I'm sure half or more of the
top level players never took math past high school (and the ones
who did probably either didn't pay attention and got C's and/or
forgot it all, just like you)!
How much of backgammon is mathematics? I could argue "all
of it!" (Setting myself up for 'out of context' misquoting.)
There are patterns and rules and probabilities and pip counts
and timing counts and... Like it or not, ALL good BG players
have mathematical ability (or they couldn't play).
How much of backgammon involves counting and calculating?
Obviously there is no simple answer, and it varies from person
to person. But 100% is the wrong answer, and so is 0%. 5%, maybe?
My recommendation (to Dave, and whoever else is still listening)
is to continue to learn through training and play. Read the GOOD
books (and ABSORB THEM). Talk to other players. Read the newsgroup
(and ABSORB the serious posts). Watch good players. Ask questions.
Study your own positions using rollouts (manual and Jellyfish).
And, most importantly, enjoy! Doing all these will answer your own
question ("How much math is req'd to play at the top level?").
Chuck
bo...@bigbang.astro.indiana.edu
c_ray on FIBS