my new backgammon book

[Robert Wachtel]

Hi All,

I used to follow this group regularly, but since the Stick forum arrived I have neglected it. Anyway, some of Tim Chow's posts there reminded me that it was still around, so I have rejoined -- which gives me the chance to do some shameless plugging of my new backgammon book, In the Game Until the End, Volume II.

It's a fairly technical (but fascinating) monograph, and so might really appeal only to the purist, but that is what most of you are here -- and since a few of you may not be USBGF members, you might not have heard about it. It's published by the Federation & you can have a good look through this link at what it's all about.

http://usbgf.org/shop/books/igue-2/

Thanks,

BobWachtel

[Tim Chow]

Bob,

Glad to see you here! I ordered your book last week and am expecting it to arrive today or tomorrow. I look forward to receiving it.

---
Tim Chow

[Tim Chow]

I forgot that I chose the cheapest mailing option so I didn't receive the book until two days ago. So far, I have only had a chance to skim through it, but I can already say that this is a really remarkable book. One way to describe it is that it is an in-depth analysis of several variants of the famous "coup classique" in which you play a last-ditch ace-point game and win by hitting two blots and closing them out. Among the positions analyzed is the so-called "Tino Road Position":

http://www.bkgm.com/rgb/rgb.cgi?view+1262

Money game: X to play

+24-23-22-21-20-19------18-17-16-15-14-13-+ O: A player
| O O O | | O O O | (Cube: 2)
| O O O | | O O O |
| | | |
| | | |
| | | |
| |BAR| |v
XX | | | |
XX | | | |
XX | | | |
XXX | X | | |
XXX | O X | X | O O |
+-1--2--3--4--5--6-------7--8--9-10-11-12-+ X: B player

According to the book, this position is the "specialty of the upbeat Australian expert Tino Lechich. For many years Tino took on all comers in this position allowing them to be Black, White, or to switch back and forth. Tino usually won from both sides!"

Before saying more, let me clarify one confusing thing about the title. The subtitle is "Volume 2," which I feel is a misnomer. It is really a totally revised and greatly expanded version of the book by the same title that Bob published back in 1993. When I see "Volume 2" it makes me think that it covers different material from Volume 1, but here the so-called "Volume 2" basically supersedes the earlier volume. All the essential backgammon content of the original book is contained in "Volume 2" so there's no real reason to buy the original version of the book except for historical interest.

In the original version, Bob tried to evaluate these difficult positions by means of hand rollouts and logic, since those were the pre-bot days. Given those limitations, he did an amazingly good job, and much of his analysis has stood the test of time.

However, as you might expect, modern bots do not agree with all of Bob's conclusions. The interesting thing, though, is that even modern bots do not understand all the intricacies of the positions in the book. Bob made a serious effort to obtain the best possible rollout information about the positions, not only using XG Roller+ for both cube and checker decisions, and a Huge search interval (i.e., much stronger settings than the default 3-ply/XGR rollouts one typically sees elsewhere), but also getting Neil Kazaross and Xavier Dufaure de Citres to help him do rollouts using the beta version of XG3. This bot is not yet generally available; it is the next generation of eXtreme Gammon, and is considerably better at containment than XG2 is. Despite all this silicon brainpower, Bob gives evidence that there are still some subtleties that the bots aren't quite fathoming correctly. At the same time, it is possible, with great care and effort, to draw some tentative conclusions. This is what Bob does in the book.

Should you buy the book? Since Bob examines only a few (very interesting) positions in excruciating detail, the book will probably not be of immediate practical value to you unless you can entice people to play the Tino Road Position (or one of the other positions in the book) with you as a proposition. However, in my opinion this is one of the few backgammon books out there that shows you what it really means to *analyze* a position. Lots of people seem to think that analyzing a position means putting it into the computer, clicking on a button, and making up an imaginary story to explain what you see. Bob's book will quickly disabuse you of that naive viewpoint, and it will also dispel any illusions you have that the bots are infallible. If you want your eyes opened to how deep backgammon can be, then you should definitely buy this book.

---
Tim Chow

[greedygammon]

Good book review. It deserves a free autographed copy (at the very least). As for the proposition position, I would pick to play the priming position. Other than 1-1 by player B eventually there will be 3 blots laying around and the key is to make damn sure B doesn't get a chance to anchor on the ace point by aggressively attacking with loose blots to pick up the remaining checkers. The position I find most challenging is how to roll home a prime when most checkers have been sent back. Back games are a fun and exciting part of the game. I look forward to reading the book.

maareyes at greedygammon.com (email)
http://greedygammon.com (home )
http://greedygammon.com/support (support discussion FAQ)

[Paul]

I don't think it's so easy for O to obtain 3 blots. Besides soon rolling a 1-1, there's another common parlay as follows. X soon rolls an unfortunate ace, exposing two blots. However, O misses the double shot and X covers with another ace later.

The fact that this is a debatable proposition suggests to me that the position is probably very even in equity terms. If we take literally the claim that the inventor of the prop actually won more than 50% of the games no matter what side, and if we assume that the opponents of the inventor were very strong players who didn't sacrifice that much equity, then a likely conclusion seems to be that most of O's cubes are passes. If many of O's cubes are takes and if the inventor really can win this position more than 50% of the time with O, then the inventor can really make a killing.

My intuition would be that O had the advantage. However, my intuition is clearly wrong since I would intuitively favour O even with the cube in the middle. However, the centered cube must give a big advantage to X else the prop wouldn't be played the way that it is.

X's position seems far easier to play, particularly at the beginning where X doesn't own the cube. Therefore, given the choice, I would choose X if I thought my opponent was stronger and O if I thought my opponent was weaker.

Also, from a money perspective, the book might be very valuable because I don't think it's difficult to find opponents for props, particularly if you let them take either side.

Paul

[Tim Chow]

On Friday, August 15, 2014 9:07:52 AM UTC-4, Paul wrote:
> My intuition would be that O had the advantage. However, my intuition is
> clearly wrong since I would intuitively favour O even with the cube in the
> middle. However, the centered cube must give a big advantage to X else the
> prop wouldn't be played the way that it is.

I can a partial spoiler from the book. Prior to consulting XG, but after consulting Tino and doing his own hand rollouts, Bob estimated that O was a slight equity favorite. I don't remember from the book (which I still haven't studied carefully) whether Bob explicitly considered a centered cube, but I think your logic is sound. Certainly it helps O to own the cube and if the position is almost level with O owning the cube, then it would surely favor X with the cube centered.

However, according to XG, O is a much bigger equity favorite than Bob or Tino suspected. I believe that at $1 per point, XG thinks the position is worth almost sixty cents to O (remembering that when O cashes she wins $2 not $1).

It's open to question whether XG's assessment is correct, but I agree with your intuition that X's position seems much easier to play, so I would expect that if anything, the position could be worth even more to O than XG thinks.

---
Tim Chow

[greedygammon]

If my "far easier" you mean "trivial" I agree. Most of the checker plays for X are forced. The cube action might get tricky. I would drop if it looked like I would get 2nd checker sent back.

"Therefore, given the choice, I would choose X if I thought my opponent was stronger and O if I thought my opponent was weaker"

I would choose the opposite. I most certainly don't want to be X against a strong player who knows how to attack and circulate checkers to pick up the remaining blots. A weak opponent would likely try to maintain the prime and not be so aggressive in trying to pick up all the blots.

[greedygammon]

Ok so I thought I would play a few matches against gnubg with this interesting position. I won the first game.. it turned into a double pass so I saved it and started another game without actually playing it out. Next game, Gnubg anchored on its own ace pt so I saved resigned and started the next game. The next game was interesting quite a few moves I doubled the bot took and I won with all three checkers on the bar and the cube at 4. A dialog popped up that I won the game I clicked OK and BAM!! my game disappeared and a new game started. WTF.. is that game saved anywhere? If I click "save" it only saves the new game that just started. I don't play gnubg much only use it for analysis. Anyway I am up 6-2 after 3 games. I am not familiar with propositions. How many games should I play before I can boast that I am crushing gnubg playing O?

[greedygammon]

Never mind about the gnubg save. I only had to click the center arrows to get back to the prev game. Score after 7 games bot is losing 18-6. sgf files available on demand

[greedygammon]

On Friday, August 15, 2014 9:09:34 PM UTC-7, greedygammon wrote:
> Never mind about the gnubg save. I only had to click the center arrows to get back to the prev game. Score after 7 games bot is losing 18-6. sgf files available on demand

Played 11 games. Here is what it looked like.

gnubg 2-2-2-2-2 total 10 points

maareyes 2-4-8-4-4-4 total 26

Here are the sgf files if anyone wants to take a look.

http://greedygammon.com/positions/tino_prop.zip

The game where I got the 8 pts gnubg says my double was a blunder (beaver) yeah.. whatever.. Im crying all the way to the bank. hehe

[Tim Chow]

On Friday, August 15, 2014 10:00:22 PM UTC-4, greedygammon wrote:
> "X's position seems far easier to play"
>
> If my "far easier" you mean "trivial" I agree. Most of the checker plays for
> X are forced. The cube action might get tricky. I would drop if it looked
> like I would get 2nd checker sent back.

One interesting decision that X can sometimes face is whether to play 2/1 with an ace, with the idea of hoping to roll another ace before O has a chance to organize an attack on the blots. This decision is discussed in detail in Bob's book.

---
Tim Chow

[Paul]

On Saturday, August 16, 2014 3:47:46 PM UTC+1, Tim Chow wrote:

> One interesting decision that X can sometimes face is whether to play 2/1 with an ace, with the idea of hoping to roll another ace before O has a chance to organize an attack on the blots. This decision is discussed in detail in Bob's book.
>

This is an interesting pay now/ pay later problem where "paying" means volunteering the double shot with 2/1*. It never occurred to me that an early payment might be correct. For example I would definitely play bar/22 with 21.

I don't like paying early here because there are two ways of avoiding payment altogether. X might roll 11 before rolling a horror ace (9%), and X might be able to avoid aces until O is forced to leave X's 1 point. The latter possibility is a reduced payment rather than a non-payment actually, because O can play to get the blots even when X plays 2/1 without the hit.

Could you give an example where you (or the author) like the idea of early payment? One thing for sure is that, from a strict financial point of view, early credit card payment is not a good idea because you simply lose money on the interest with no compensation. So you should only pay at the last possible moment where otherwise would occur fines.

I thought the Tino Road position was a credit-card-type problem.

Paul Epstein

[Tim Chow]

On Saturday, August 16, 2014 2:25:52 PM UTC-4, Paul wrote:
> This is an interesting pay now/ pay later problem where "paying" means
> volunteering the double shot with 2/1*.

I'm pretty sure that 2/1* is never correct when there's an alternative. I said 2/1, not 2/1*. This isn't an option immediately but sometimes O runs off the ace point and X has the option of playing 2/1 without immediately volunteering a shot on either of those two checkers. If I recall from the book, this is rarely correct, but if you really want to know then I recommend you buy the book.

---
Tim Chow

[Paul]

I would think the converse of your statement is true, too. If a person really wants to buy the book, then that person probably wants to know the subtleties of the Tino Road position.

I have just placed it in my to-buy-and-read list, but I'm a bit too busy to read it right now.

Paul

[Tim Chow]

On Sunday, August 17, 2014 9:46:40 AM UTC-4, Paul wrote:
> On Sunday, August 17, 2014 2:04:02 AM UTC+1, Tim Chow wrote:
>> if you really want to know then I recommend you buy the book.
>
> I would think the converse of your statement is true, too. If a person
> really wants to buy the book, then that person probably wants to know the
> subtleties of the Tino Road position.

Since you're a mathematician, you may be amused when I point out that the converse is actually, "If I recommend you buy the book, then you really want to know"!

Of course the reality is that most conditionals in natural language (including the ones above) are not material conditionals.
---
Tim Chow

[Paul]

On Saturday, August 16, 2014 3:00:22 AM UTC+1, greedygammon wrote:
...

> "X's position seems far easier to play"
>
> If my "far easier" you mean "trivial" I agree. Most of the checker plays for X are forced. The cube action might get tricky. I would drop if it looked like I would get 2nd checker sent back.

...

To be concrete, if X rolls an immediate 31 (which is forced), O is (obviously) a favourite to send a 2nd checker back. So, are you really saying that after 31, O can immediately cash?

My opinion is that O should hold after X rolls a 31. If O hits and X fails to anchor, O will have a double which could well be a pass.

Paul

[greedygammon]

that's what I would do. I don't know if it's right or wrong. I have not read the book nor studied the gnubg rollouts. As I stated earlier, I like the priming side. If playing X, I have no idea how to play the cube once I have all 3 blots exposed. There are other scenarios also one being where X anchors in O home board. That can get tricky. It is an interesting position for both sides. Maybe I underestimated the value of X needing to know how to handle cubes. Based on my very limited 11 game experience against gnubg it seems the bot likes to take and ends up losing more points but maybe with more games played the score would even out. I got tired after 11 games, it was late. And also I was giving all the first rolls to the bot so maybe that could have skewed the results in my favor.

[Paul]

On Tuesday, August 19, 2014 2:48:31 AM UTC+1, greedygammon wrote:

Paul wrote:
> > To be concrete, if X rolls an immediate 31 (which is forced), O is (obviously) a favourite to send a 2nd checker back. So, are you really saying that after 31, O can immediately cash?
>
> >
>
> >
>
> >
>
> > My opinion is that O should hold after X rolls a 31. If O hits and X fails to anchor, O will have a double which could well be a pass.
>
> >
>
> >
>
> >
>
> > Paul
>
>
>
> that's what I would do. I don't know if it's right or wrong. I have not read the book nor studied the gnubg rollouts. As I stated earlier, I like the priming side. If playing X, I have no idea how to play the cube once I have all 3 blots exposed. There are other scenarios also one being where X anchors in O home board. That can get tricky. It is an interesting position for both sides. Maybe I underestimated the value of X needing to know how to handle cubes. Based on my very limited 11 game experience against gnubg it seems the bot likes to take and ends up losing more points but maybe with more games played the score would even out. I got tired after 11 games, it was late. And also I was giving all the first rolls to the bot so maybe that could have skewed the results in my favor.

I don't understand how you can't "know if it's right or wrong." It's extremely obvious that you're wrong. It has been repeatedly stated here that the prop inventor did well in the prop playing both sides, and we can assume he played strong players. If it was so easy for O to cash, no one could do well playing the prop as X. Surely, O is a big favourite to get a position similar to the type I mentioned.

It also doesn't make sense to say that giving the first rolls to the bot may have "skewed the results". The prop says clearly that X is to play so if you were playing O, you weren't "skewing the results." 11 games is an extremely small sample, as you acknowledge.

What does the bot say about relative luck? If the bot doesn't think you've been relatively lucky, then there's either a problem with the bot's play or a problem with the prop. Props should be interesting for both sides.

Paul

[greedygammon]

And we can also assume he played some weak players who don't understand the position so I don't know what your point is with that statement

As for the first roll, I didn't bother to see if the prop specified which player should move first. I think it would still be interesting to see if putting X on roll gives an advantage to O.

"What does the bot say about relative luck?"

I posted the sgf files, look it up.

Look Paul.. you already lost me when you posted earlier
"I don't think it's so easy for O to obtain 3 blots"
That is a goofy statement when it should be clear that the only way for this prop to work is when there are 3 blots without which is an easy win for X. Maybe you should go back to posting obscure math problems LOL..

[Paul]

On Tuesday, August 19, 2014 2:01:20 PM UTC+1, greedygammon wrote:
... Maybe you should go back to posting obscure math problems...

Ok, happy to cater to your request. I've been stuck for about a year on the original elementary proof of Szemeredi's theorem on arithmetical progressions: http://matwbn.icm.edu.pl/ksiazki/aa/aa27/aa27132.pdf

I am stuck on Fact 6 (page 212) because I don't follow the "induction" argument that X' and R have non-empty intersection.

Many thanks for your help.

Paul