Snake EyesBackgammon

Every checker on the board is a distance to travel — its point number, counted from your own perspective, is how many pips it still owes you before it's home. Add up every checker's distance and you get your pip count. Do the same for your opponent. Whoever's number is lower is ahead in the race. That's it. No feel, no vibes — just arithmetic, and it's the single most important number on the board.

Here's why it matters more than almost anything else you'll read: pip count decides your entire strategic posture. If you're ahead in the race, your job is to simplify — trade contact for safety, break away from your opponent's checkers, and let the dice walk you home. If you're behind, racing is close to a dead plan. You need to change the shape of the game — hit something, hold an anchor, build a prime — because a pure footrace you're already losing doesn't get better by playing it safe. Playing passive while behind isn't cautious, it's just losing slowly.

Learn to count pips fast, learn to read the gap, and half your strategic decisions each game get a lot easier before you've even looked at the shape of the position.

There are exactly fifteen distinct opening rolls in backgammon, and modern computer rollouts have converged on a stable, well-documented answer for almost every one of them. This isn't a matter of taste. It's close to solved, and playing an opening roll on instinct instead of on the table is just giving away equity for free.

The headline: 3-1 is universally regarded as the best opening roll in the game — 8/5, 6/5 makes your 5-point, the single most valuable point on the board, risk-free. 6-5, the "lover's leap," has one real answer too: run a back checker the full distance, 24/13, straight to the midpoint. Most other rolls make a point when they can — 6-1 makes the bar-point, 5-3 and 4-2 make the 3- and 4-points cleanly.

But the table isn't frozen in amber — the bots changed parts of it. Classical theory from the pre-computer era favored slotting the 5-point on 5-1 and 4-1; rollouts moved both away from that. 5-2 is a flat reversal — old theory liked 13/11, 13/8, modern rollouts prefer 24/22, 13/8. And on 6-4, a play human masters used to mock as too passive — 8/2, 6/2, making the 2-point — turned out to be roughly equal to the traditional running play once someone actually rolled it out. Know the table. It's free equity sitting on the very first roll of every game you'll ever play.

Strip away the noise and every mid-game backgammon position reduces to one of five recognizable shapes. Learn to name the shape you're in and the right plan usually falls out of it.

Priming game — you build a wall of consecutive points to physically trap an opposing checker behind it. Strongest when you're not badly behind in the race, because a prime buys time rather than winning outright. Blitz — an aggressive early attack, hitting blots and racing to close your home board before your opponent can re-enter. High variance, high reward, best played when the dice are already running hot for you. Backgame — holding two or more deep anchors in your opponent's home board, waiting for a late shot while they bear in. This is a fallback plan for when you're already well behind, not a first choice, and it demands real timing discipline to pull off. Holding game — a single advanced anchor, a calmer middle path between racing and backgaming. Mutual holding game — both sides are holding anchors at once, and the game turns into a waiting contest over who's forced to leave the first shot.

One more piece of hierarchy worth memorizing outright: not all anchors are equal. Your opponent's 5-point is the best anchor in the game, then the bar-point, then the 4-point. If you're choosing where to dig in, dig in there, in that order.

Here's the concept that separates players who understand primes and backgames from players who just build them and hope. Timing is whether you have spare checkers and spare moves you can make without being forced into a bad one — breaking a key point early, or stacking checkers on top of each other because you've got nothing better to do with the roll.

Good timing means you can afford to wait. Poor timing means the board is making your decisions for you, whether you like the options or not. A priming game with bad timing doesn't get to wait for the perfect moment to spring the trap — it's forced to break the prime early, on the board's schedule, not yours. A backgame with bad timing collapses before the shot it was built for ever arrives.

This is exactly why a backgame is a rescue plan and not a strategy you reach for on purpose — it demands precise timing to hold together, and precise timing is hard to engineer, easy to lose. If you only take one thing from this note: before you commit to a prime or a backgame, ask whether you'll actually still have moves left when the position calls for one. If the honest answer is no, you don't have a plan, you have a countdown.

Step 1 — Know the object. The doubling cube isn't something you roll. It's a marker with six faces: 2, 4, 8, 16, 32, 64. At the start of every game it sits in the middle of the board turned to "64" face up, which everyone just reads as "1" — the game is currently worth one point.

Step 2 — Know when you're allowed to touch it. Before you roll your dice on your turn, and only then, you can pick up the cube instead and offer it to your opponent. That's a double. You're saying: "I like my position — let's play this one for twice as much."

Step 3 — Know your opponent's two options. When you get doubled, you have exactly two choices and nothing else. Take — turn the cube to the next number up and keep playing, now for double the stakes. Pass — end the game right there and give up whatever it was worth before the double. Passing costs you 1 point. Taking risks 2.

Step 4 — Know who owns the cube after a take. Whoever takes now owns the cube. Only the owner can double again later — your opponent can't hand it straight back to you next turn. If you own it and things swing further your way, you can offer it again (a redouble), turning it to 4, then 8, and so on.

Step 5 — The newbie's gut-check for take-or-pass. You don't need a computer to get started. Ask one honest question: "If we played this exact position out a hundred times, would I win more than 25 of them?" Yes — take. No — pass. That one-in-four line isn't a rough guess, it's the real mathematical break-even point. Everything you learn after this is just refining it.

Step 6 — Two wrinkles worth knowing early. A gammon (you win before your opponent bears off a single checker) is worth double the cube. A backgammon (you win while they've still got a checker in your home board or on the bar) is worth triple. If a gammon or backgammon is a real possibility against you, lean toward passing — giving up 1 point on purpose beats risking 3 by accident.

Step 7 — The one habit that separates beginners from everyone else. Look at the cube every single turn, even when you're not going to touch it. Beginners treat it as an afterthought. Everyone who's any good treats it as a live decision on every roll of the game.

Everyone knows the chess story — a computer finally beating a reigning world champion, headline news. Backgammon got there first, and quieter, years before chess had its own version of that moment. In 1979, a program called BKG 9.8, built by Hans Berliner at Carnegie Mellon, beat reigning World Champion Luigi Villa 7-1 in an exhibition match — the first time a computer beat a reigning world champion at any board or strategy game. Berliner himself credited some of it to lucky dice, an honest caveat worth keeping, but the door was already open.

The real turning point came later. In the early 1990s, Gerald Tesauro at IBM built TD-Gammon, a neural network that learned by playing itself, over and over, with no human game data at all. Published in 1992, it was the first complex two-player game where a self-trained program reached elite human-level play — a direct ancestor of everything from AlphaGo onward. From there the lineage runs Jellyfish, Snowie, GNU Backgammon, and eXtreme Gammon, and in 1997, Jellyfish 3.0 played a 600-game match against two strong human players and it came out essentially a draw — the moment parity became undeniable.

What actually changed wasn't vague "computers got better." It was specific: slotting reversed on the 5-1 and 4-1 openings, the 5-2 opening flipped outright, and a 6-4 play human experts had mocked for decades as too passive turned out to be roughly equal to the traditional line. The lesson isn't "trust the bot blindly." It's that rollouts kept finding real equity that pen-and-paper analysis missed for fifty years — and the same rigor is available to you every time you review your own games instead of trusting your gut.