What to play solo against a computer: games with a strong AI

Playing solo against a computer isn't just a "backup for when no one's around." A computer opponent is available every second of the day, never gets tired, and — if it's built well — adjusts to your level. The only real question is how smart it is. A weak bot moves almost at random and bores you within an evening; a strong one makes you think and genuinely raises your game. Let's cover what to play against the computer so it stays interesting for the long haul.

WHY AI QUALITY IS EVERYTHING

When your opponent is a person, the interest sustains itself: everyone has their own style and their own mistakes. In a game against a computer, everything rests on the engine. If the bot is too weak, winning is dull; if it beats you every time with no way back, you give up. The ideal is an opponent with adjustable difficulty: forgiving and instructive at the start, and at the top calculating so well that every game becomes a puzzle.

WHERE THE COMPUTER HAS LONG BEATEN HUMANS

In the classics this is settled ground. In chess, engines like Stockfish and the neural-net Leela Chess Zero play above any world champion — grandmasters train with them. In Go it was long thought the search space was too big for a computer, until AlphaGo beat Lee Sedol in 2016; today open nets like KataGo are available to everyone. There are strong bots in checkers, reversi, poker. They share one thing: modern AI no longer grinds through every move in order — it "senses" the position, the way a human does, only faster.

HOW MODERN GAME AI WORKS

The breakthrough came with the AlphaZero approach: the neural net learns not from other people's games but by playing itself millions of times. It combines two skills — a "policy" (which moves look reasonable) and a "value" estimate (who stands better right now) — and on top of that runs a tree search (MCTS) that reads the most promising lines deeply. That's why such a bot plays in a human way: it builds plans rather than bludgeoning with brute force.

WHAT MAKES A VS-COMPUTER GAME WORTHWHILE

Strength alone isn't enough — the opponent should help you grow. Three things worth looking for:

1. Several difficulty levels — so there's both an easy start and a ceiling to reach for.
2. A game review — so after the game you can see where you went wrong and which move was stronger.
3. Hints and puzzles — so you can drill tactics separately, not just replay blindly.

HIGHRISE HEIST AS A MODERN EXAMPLE

If you want something of your own rather than a classic, take a look at Highrise Heist (in Russian, "Perehvat vysotki") — an abstract strategy game built precisely around a strong AI. The opponent here is an AlphaZero-class neural net (that same pairing of MCTS with a policy/value network) with five difficulty levels: from a gentle beginner setting to one where a single slip costs you. After the game an AI review kicks in, there are hints, and a daily Puzzle Rush for tactics. The rules take a minute to learn: across 10 stands you need to close 6 highrises in your color, and the signature transfer lets you snatch a nearly finished highrise at the last moment — which is where the comebacks and live endings come from.

You can play right in the browser — free, no install, no registration, in Russian and English. A good way to tell whether the game is for you is one match against the Easy AI: you'll grasp the rules immediately, and add the difficulty yourself later.