The Chaos of the Dice: A backgammon hustler’s quest to gain an edge (2013) | Hacker News Reader

The Chaos of the Dice: A backgammon hustler’s quest to gain an edge (2013)

newyorker.com

63 pointsphilshem34 comments

I absolutely love backgammon. I think people dismiss it as not being on the same level as chess or go because of the "chance element," but that element is probably better described as "management of risk" or "the calculation of risk," and it is remarkably deep.

And while it's true that a novice can beat a grandmaster in a single game, you can be sure that you would have been destroyed in six games by a player like Natanzon.

The other nice thing about it is that while two strong players can break each other's brains over it for hours, it's also a fantastic casual game that's just good fun with coffee and conversation (as it is throughout much of the Middle and Near East).

I'm always hoping it undergoes a resurgence one of these days. Stranger things have happened!

My (now defunct) family's software company had a backgammon game with fairly good AI. We frequently got complaints from customers accusing the AI of cheating by getting favorable dice rolls. This wasn't possible, as the dice roll algorithm literally didn't know which player it was rolling for (let alone the context of the overall game). Instead, the complainers didn't understand that a chief strategy of backgammon is to set yourself up to be able to take advantage of as many different dice rolls as possible.

Got a kick out of seeing this post. If I am not leaping to the wrong conclusions from your user name, then I know who you are and I'm the original author of that backgammon game :). I worked with your Dad to publish it back in 1990 or 1991. And yes, we got complaints all the time about die rolls favoring the computer player. At least as of a few years ago you could still find some archived usenet threads where we tried to put those to rest by discussing the architectural features you alluded to, as well as publishing some multi-million roll statistical distributions. I recall that Joe Celko brought up the situation in one of his digressions on pseudo-randomness. Thanks for the walk down memory lane.

To save anybody else from having to do the research on this rabbit hole I believe the game in question is: https://www.mobygames.com/game/mvp-backgammon

Some fun background from 1996: https://bkgm.com/rgb/rgb.cgi?view+59

Yes, that's the game, and I am the author of the usenet post you linked. Thanks for digging it up! I haven't seen that in years.

Hey Mark!

Just saw this. You indeed have the right person. Here's an archived version of the MVP site (ugh that design was terrible, even for the early 2000s) that tried to deflect some of the angry phone calls: https://web.archive.org/web/20060325184439/http://www.mvpsof...

That's awesome :). I remember writing that defense. I was very young and certain of myself, apparently. Well the die rolls may not have been "as close to absolute randomness as is possible with a PC" but I still think they were pretty good. Thanks for digging that up. It was really an unexpected pleasure to run into you here and see your post on this.

"a chief strategy of backgammon is to set yourself up to be able to take advantage of as many different dice rolls as possible"

Exactly. I have many friends who refuse to play with me because I blundered into that when I was in high school and continued to refine it. I guess it just clicked for me but seems to be hard for many people to really understand.

That and managing risks too, as others point out.

And I just realized the misplaced modifier. My family is intact (though my father has passed away), but the company is defunct...

To me a major difference (besides dice) is that Backgammon has the concept of points which a player can double the value of during a game and if the opponent refuses then they forfeit. But once you’ve doubled then the opposing player gets the next opportunity to double. This is the part of the gave that really sets apart the good players I think? Over a series of games you have to calculate your advantage at certain points. I think this element makes it more a game of skill (like chess where there is no randomness and all game knowledge is shared at all times) than people give it credit for.

I can also see why gamblers like it so much. Nothing like a mid game “more action” button.

Right, the doubling cube is what sets apart the good players.

For every possible state, there is an expected value of your outcome, from the current state averaged over all possible continuations of the game or match, if you make perfect moves and your opponent also does. (This can be denominated in either cash/points value within one game, or in expected match-win value across a match of multiple games.)

The swing of backgammon is in the imperfect moves, and in how much expected value each one costs you. It's not that hard to study checker strategy well enough that you're rarely losing more than 1% or so of a game's value; even if you make imperfect moves, they're generally still pretty good. But the doubling cube can account for much greater swings; a mistake like doubling too early or incorrectly responding to your opponent's can easily cost 20% equity or more.

Same with Magic The Gathering, Poker, etc. over a few games you can win. Long term, nope. The best players at MTG have maybe a 60-65% win rate in a professional setting. And that is considered very good. But they can do it consistently. So really the game is about how much better than 50/50 you can get. The difference in adding 5-8% to your win rate is an astonishing amount of skill. It’s part of what makes magic fun. Complete novices have a shot. Maybe that rankles some people, but that’s the game :)

We just need a new awful Netflix series about a backgammon hustler.

Or a backgammon clone of https://lichess.org

That’d make a awesome anime.

The cracking sounds of the board, rigged dice, shady characters, a 40 something NEET with nothing left to lose, hell this thing writes itself!

Not exactly this, but close: Kakegurui.

https://en.wikipedia.org/wiki/Kakegurui_%E2%80%93_Compulsive...

I was thinking more along the lines of Prince of Tennis meets Kaiji but I’ll check that one out too!

>And while it's true that a novice can beat a grandmaster in a single game

This is what makes it so good for gambling. A wealthy novice is very unlikely to gamble against a grandmaster in a game of chess because its clear they have no chance to win. Backgammon (like poker) is another matter..

> Unlike chess, backgammon is tactile, fast-moving, even loud, with checkers slammed down and tiny dice sounding like rattlesnakes as they traverse the board. Casual players who believe that they are good persist in the illusion because the element of chance obscures their deficits. At its heart, backgammon’s cruelty resides in the dramatic volatility of the dice. Even a player who builds flawless structures on the board can lose to a novice. The good players simply win more often. As a result, backgammon is often played in marathon sessions that reward physical stamina, patience, and emotional equilibrium. One notable match lasted five days, with both players getting up only for bathroom breaks. The loser fell to the floor.

> Chess players can visualize what the board might look like twenty moves ahead, but in backgammon the dice offer twenty-one random possibilities at each turn. The game must be encountered frame by frame.

---

Sadly, Matvey Natanzon (aka Falafel) passed away this year.

https://en.wikipedia.org/wiki/Matvey_Natanzon

I'm sure the rest of the statement is true but there are fast time-control variants of chess that are very tactile over the board. Bullet chess is a minute of total time per side so moves are blitzed out at lightning speed.

In this case your citation of the the Wikipedia page has no citation... any other news of what actually happened?

There's a Times obituary. It was cancer: https://www.nytimes.com/2020/03/03/nyregion/matvey-natanzon-...

It's cited at the head of the article, where his death is first mentioned.

https://www.economist.com/obituary/2020/03/19/falafel-died-o...

I'm an average player it seems, but I've had the problem of easily beating every free backgammon app on Android, that is until I discovered Backgammon NJ. It was $10 when I purchased it years ago, but it's well worth the money.

One of these days I'll have to actually try to study the game and work on strategy, but until then, I like pitting myself against the difficult computer every-so-often.

>> “I’ll give you fifty bucks right now to eat one of those pies,” Mr. Joseph said, pulling out a crisp bill.

What toxic assholes - the lot of them, all those romantic characters in the article that could well be the summary of a Tom Waits song. They just come across as a bunch of toxic fucking jerks that don't mind endangering each other's health for the pleasure of winning some money off each other.

The guy who got the implants for a year on a bet? How the fuck is that healthy? Maybe it's all good because he liked them after all - and hey, he got to meet ladies!

What the hell is wrong with these people? It sounds like they all have very severe undiagnosed mental health issues.

I read HN just so that I can stumble upon an article like that once in a while

All due respect, but you can subscribe to the New Yorker and stumble upon an article like that once a week!

You can’t know all the places to look and you don’t have the time to look in all those places, no less filter it all for the best content. This is why I and many others like aggregators. That’s not to say we don’t also have New Yorker subs. I do, through my library–I can read it on my iPad–but I don’t have time to read every issue cover-to-cover. And I definitely find less than one article per issue really resonates with me personally, for any given publication and depending on where I’m at in my life when an issue goes to press. I bet there are plenty of people who don’t even know what the New Yorker is; your suggestion doesn’t help them. HN very well could introduce them, just as it’s introduced me to plenty.

Do computers destroy top backgammon players in the same way as chess and go?

No, they’re _much_ better. Backgammon AI has achieved supremacy years before chess.

For anyone interested in neural nets Backgammon makes a fascinating example. Essentially, AI computes the value of each position. Like a dome roof, the entire height could be off by a fraction of a percent, but all that matters is the relative values of the possible next moves. These valuations converge much faster than the overall function.

Who gets to roll the dice 8)

Would you mind pointing me at some sources please? I'd be intrigued to see if any algorithms or other insights have come out of the AI stuff.

TD-Gammon from 1992 was an early success of neural network based reinforcement learning. Descendants of the technique remain standard today in RL.

Snowie, and Jellyfish.

Reply on news.ycombinator.com