How good is chess software from 1982?
I wanted to see how I’d fare against a decades-old chess program. I got some odd surprises along the way.
Nowadays, there’s no point in trying to defeat a chess computer. No human being can even come close to the strength of Stockfish and other modern chess engines.
But that wasn’t always the case. When IBM’s Deep Blue chess computer defeated reigning world champion Garry Kasparov under tournament conditions in 1997, it caused a sensation. The playing level of chess computers increased with the available computing power.
Chess programs capable of beating amateur players must be considerably older than Deep Blue. I was given a replica of the Sinclair ZX Spectrum. The chess game Masterchess was released for this ancient home computer in 1982. I want to find out if I can beat this chess software.
The test setup
For this article, I’m playing Masterchess on the Spectrum emulator Fuse. It’d also work on The Spectrum, but it’s easier in the emulator – for example, when I want to take screenshots.
Of course, I have to play this 1982 game without a mouse and type in two square names for each move, which takes some getting used to. Can I take back a move if I make a typo? No chance. If I accidentally tap X, the game’s over immediately. Well, that’s how it was back then.
I estimate my playing strength at around 1,400 Elo. Of course, this varies from game to game. Like any amateur player, I occasionally make silly mistakes. Plus, I can’t memorise most openings, but I do know the basic principles of the early stages.
The game offers ten difficulty levels (levels 0-9). I’ll start at the easiest level and work my way up.
Level 0: the clue’s in the name
At the easiest level, I play five or six games with both White and Black. I always win, often in fewer than 25 moves. Only once does it take me a long time – 61 moves – to get to checkmate with Black.
Playing at this level’s fun. The software doesn’t make obviously stupid moves such as «I’ll position my rook on a square where it can easily be taken», but it’s relatively easy to outsmart by planning two or three moves ahead. It also doesn’t recognise checkmate threats if they’re more than one move away.
Level 1: my first loss
With White, I win in 31 moves. I think things will go on like this – but they don’t. In the rematch with Black, I get my first thrashing. After two stupid mistakes on my part in moves 21 and 24, the game’s over. It still takes a tedious 59 moves, because my opponent’s anything but determined to deliver checkmate in the endgame.
Level 2: a lucky draw
Playing White, I’m bad from the start, never managing to build up a half-decent position. The computer carves out a big material advantage and should really win. But then it keeps avoiding my perpetual check in the same way. After the third repetition of the exact same position, it’s a draw. The software obviously doesn’t know this rule and happily continues playing. I resign and chalk it up as a draw. It’s not my problem if the computer doesn’t know the rules.
I play better with Black. It’s an even game with a slight advantage for the computer. I have an extra minor piece at one point, while it has three more pawns. In the end, everything comes down to whether I can promote my pawn one square from the back rank. I can’t. So, it’s defeat after 49 moves.
Level 3: no chance
I’ve lost my motivation. I barely stood a chance at level 2. What’s the point now?
At level 3, the computer takes significantly longer to think, about a minute per move – longer than I do. A stupid mistake means I lose my queen on move 29. From move 35 onwards, there’s no hope left. I resign because I don’t want to torture myself endlessly again.
I’ll forgo a rematch. There’s no doubt about it: the chess program’s stronger than me. I concede defeat.
But since I want to know how strong Masterchess really is, I’m pitting it against Komodo. It’s a strong, modern chess program that I can access easily and for free via chess.com.
Level 9: ridiculous thinking time
I jump straight to the highest difficulty level. The program’s ridiculously slow. The first two moves are executed instantly because they’re taken from an opening database. From the third move on, Masterchess gives itself four hours of thinking time. Not for the entire game; per move. The program easily goes past the 100-hour mark. But because the hour display only has two digits (in typical 80s fashion) the timer resets to zero.
On a true ZX Spectrum, the game’s unplayable on the highest settings. Nobody wants to wait that long for their opponent, especially since the computer has to be running the whole time.
However, I can run the emulator at up to 160 times the normal speed. This reduces four hours to 1.5 minutes. That’s acceptable. Strictly speaking, though, I’m playing against the computing power of a modern machine. So, level 9 doesn’t tell us how good a chess computer from 1982 is.
I’m letting Masterchess compete against Komodo at this level. Just because I can.
Strange behaviour
Something weird happens on move five: Masterchess returns the bishop to its starting position after four hours of thinking time. This is unusual and probably bad. In the early stages, the goal’s to free the pieces from their inconvenient starting positions as quickly as possible. Retreating to H5 would have been better to maintain the threat against the queen.
On move 13, it happens again: Masterchess returns the already developed knight to the back rank. Sorry, but that’s just stupid. Chess.com agrees: the live rating’s now +3.21 for White.
In move 23, something even stranger happens: Black moves its queen from D8 to H4, but it isn’t shown on the board. I don’t know whether Masterchess still has the queen on its radar or just forgot about it.
Never mind, Black’s position was hopeless anyway. Two moves later, it’s checkmate. The analysis on chess.com gives Masterchess a playing strength of just 1,050 Elo in this game.
Illegal moves
Suddenly, I have hope again that I can beat the software on level 9. I lose twice, once with White and once with Black, but I hold my own each time. On the third attempt, I can’t move my queen on move 15 – Masterchess claims the move’s illegal every time. Presumably, the program’s forgotten another queen. This time to my disadvantage.
I can’t play like this, of course. I feel like the chess program’s overloading itself at level 9. Presumably, the extremely limited hardware can’t hold such deep analysis in memory. The Spectrum ZX only has 48 KB of RAM.
Level 7 against Komodo at 1,600 Elo
Since something’s clearly wrong at the highest difficulty level, I’ll finish with Masterchess playing Komodo at level 7. At this level, the program allots 15 to 30 minutes of thinking time per move. I’ll set Komodo to an Elo rating of 1,600.
Playing Black, Masterchess wins easily in 29 moves. Komodo isn’t playing well in this game either. The chess software varies its strength, presumably to better simulate a human opponent. After 27 moves, Masterchess displays «Mate in 2» on the screen. That’s it; I can’t even finish the game. The software doesn’t show how it delivers checkmate. Odd.
Playing White, Masterchess loses after 57 moves. The game’s level for quite some time, then both sides make some pretty serious blunders. Komodo plays surprisingly poorly in the endgame, but eventually wins.
Masterchess seems to work correctly at level 7. However, the AI isn’t much stronger than at level 2.
Verdict: beatable for good amateur players
Playing against old chess software’s an exciting challenge that I can recommend. Of course, you can also compete against modern bots that deliberately play a bit badly. But it’s not the same as beating a normal chess computer.
I couldn’t beat it: Masterchess from 1982’s better than me. Club players, though, should be able to defeat it. I estimate its strength at 1,600 to 1,800 Elo.
Masterchess never plays truly poorly and is pretty strong at even level three out of ten. I have to think two or three moves ahead to have a chance. At the higher levels, the thinking time increases exponentially, reaching ridiculous extents at the hardest level. However, the playing level doesn’t increase at the same rate. Ironically, at the hardest level, weird things happen that suggest the software’s faulty. It seems overwhelmed by its own depth of analysis, makes strange tactical errors and even forgets a queen on the board.
In summary, the old software easily reaches a moderately good level, but then struggles to surpass it – despite ridiculously long thinking time.
My interest in IT and writing landed me in tech journalism early on (2000). I want to know how we can use technology without being used. Outside of the office, I’m a keen musician who makes up for lacking talent with excessive enthusiasm.
Interesting facts about products, behind-the-scenes looks at manufacturers and deep-dives on interesting people.
Show all
