#038 - Even Wins: An Odd Duck in the Nim Family

Release Date: July 1973

Platform: Mainframe (BASIC type-in)

Genre: Puzzle (Nim)

Developer(s): Unknown

Publisher(s): Digital Equipment Corporation

History

101 BASIC Computer Games already has several variants of the ancient game of Nim. We've already seen Batnum (the 1970 version) and Nim (otherwise known as Gamnim ), and now we've arrived at the final variant of the "piles of objects" games in the original version of the book - Even Wins.

Fittingly, there's an even number of games.

This is actually a pair of games, as there are two different versions of Even Wins present in the book, creatively named EVEN and EVEN1. The 1978 version of the book calls them Even Wins and Game of Even Wins, respectively. Here, I'm looking at the former of the two, which happens to also be the original version of the game.

What's the difference between the two games? It comes down to how the AI is programmed. In the former of the two, the AI runs with a set algorithm, essentially a "set" difficulty, whereas the latter leans more into the "machine learning" style of AI that's cropped up a few times before. This "Cybernetic Version," as MobyGames calls it, implements an AI similar to that present in a game like Awari, where the AI starts of not being very good at the game, but improves its play over multiple rounds as it "learns" how you play. 101 BASIC Games states that it'll take about 20 rounds for it to become "a challenge to beat." I'll be covering this latter version of Even Wins in the next article, probably as a Gaiden article next Tuesday, as it seems redundant to dedicate a full Friday release to the same game two weeks in a row.

This and the original are the only media I could find on Even Wins.

While we know who developed the EVEN1 version of the game, Eric Peters of DEC, the author of the original game is unknown. I even scoured through several computing magazines to see if there was any mention of the game or a possible original author, but to no avail. Neither PCC nor Creative Computing even make mention of the game.

The Game

So, what makes Even Wins different to the other Nim games? Hint: it's in the name. The same gameplay loop is present, of two players taking turns removing objects from a pile - a pile of 27 objects (marbles) in this case - no more, no less. How the victor is determined is the major difference. In the typical game of Nim, it's the person who takes the last object that wins or loses, depending on the ruleset. In Even Wins, however, it's the amount of marbles the players take the determines the winner. Whoever has an even-numbered pile of marbles when the last marble is taken wins. It's why the game starts with a pile of 27 - it has to be an odd-numbered pile for the game to work. The aim of each player then becomes to force the other into ending with the odd-numbered pile.

Here's how the game explains itself.

Forcing the AI opponent into losing is easier said than done. Unlike some of the other "solved" game AIs I've dealt with in the past (OXO, Batnum '67), Even Wins' AI presents a stiff challenge, but not an unbeatable one. 101 BASIC Games' take on the difficulty is "the computer... is impossible to beat if you don't know how to play the game." So tough - but not impossible.

In practice, I found this to be a very accurate assessment of the game's challenge. Unlike a game like Batnum '67, I wanted to try and beat the game without doing any sort of research into the winning algorithms. Naturally, I got stomped pretty badly the first ten times I played because I, like the book says, had no idea what I was doing. Once I got a handle on what I needed to do, by observing the way the computer played - that's when my progress in beating the game moved in a positive direction.

How most rounds end up.

I observed that the computer liked to try and force me into a position at the end of the game where there was 5-7 marbles remaining. That seemed to me to be the red zone where the winner is always determined; whoever forces their opponent to take marbles when 5-7 are remaining will always be the victor. So I looked over the previous matches and observed what moves the computer made - what numbers it liked to leave the pile on at the end of its turn, and how it responded to my moves - and tried to incorporate those moves into my own strategy.

After about 15 minutes worth of attempts, I had a win! I was able to manipulate the pile into a position where I had it end on 11 marbles - apparently a key position, which allows making the decisive move into that 5-7 range. Admittedly, I was quite chuffed about this - the computer really made me earn that win. I was also extremely relieved that this wasn't another one of those "impossible games" situations, where the AI is programmed to be perfect, mercilessly slaughtering all comers who would dare challenge its genius.

Victory!

However, I'll briefly note here that, from all my testing a playing, that the only position the player can win from is by going first. I was unable to win going second. This is typically the case in most early Nim variants and other solved games, that winning is only viable from a certain starting point. Some, it's going second, others, like here, it's going first. It seems that going first is the position of control and dictating the flow of the game.

A Brief Word on Difficulty

There's something to be said here about games and difficulty. It's apparent to me that balancing the challenge level of a game can have a vast effect on the overall quality of the playing experience. Take Batnum '67, for example (one of my favourite punching bags as of late). That game makes it nigh-on impossible to beat the AI outside of the deliberately programmed first round. The level of challenge is far too high, and it royally sours the experience, because I know I can't beat the game. What's the point in trying? LEM suffers the same fate, but more because that game is far too complicated for its own good.

On the opposite end, there's Batnum '70 - the younger brother of Batnum '67. That version allows for near-endless customisation of the game rules and settings, so much so that I can just manipulate the game state to where I'll always win - Bomber also has the same problem, for a more recent example. This also ruins the game experience, but at the opposite end of the scale, because I can basically cheat in a game-approved way to always win. And, like the near-impossible games: what's the point in playing?

Just some brief shower-thoughts for your consideration. If you have any thoughts, considerations or counter-arguments, I'd love to hear them in the comments. With that out of the way, let's score Even Wins.

Scores

Time Played: 13 minutes

Difficulty: 6 (Challenging)
It's like the book says: impossible to beat if you don't know what you're doing. I needed to play a few rounds to get a grip on how I needed to manipulate the numbers to get into a winning position. Watching what the computer was doing in response to my moves, and where it boxed me in helped to give me an idea of what I needed to do to beat it.

Gameplay: 4/20
This reminds me of Batnum '67 but better in the sense that the computer is beatable without the game having to do some weird programming nonsense to force the AI to make a mistake in the first round. Otherwise, Even Wins is more-or-less the same game, except with the difference of how to win, which did subtly change the way I played compared to the other Nim variants. It's got that going for it at least, and the computer being beatable presents a proper challenge to overcome, which I feel like I earnt over the course of playing the game. Even Batnum '70 doesn't feel as rewarding due to the customisation options of it permitting me to manipulate the starting parameters in my favour, hence why I think Even Wins is slightly better than it.

Controls: 5/10
It does that funny thing that Cube does where you have to use binary for yes (1) and no (0), but it doesn't feel as intrusive or unintuitive here as it did there.

Visual: 1/10
Even Wins is very rudimentary on the visual front. It's only text, no other formatting tricks like what we've seen in other text-based games from 1971 - 73.

Functionality: 5/5
No bugs present during normal gameplay.

Accessibility: 3/5
Putting this as standard for a text-based game.

Fun Factor: 5/20
There's nothing Earth-shattering here, just a simple puzzle game, but I actually did enjoy trying to beat the computer somewhat. It was a stiff enough challenge that didn't feel too far out of reach, so I enjoyed the process of trying to figure out how to beat the AI.

Overall, Even Wins earns a score of 23 - much better than what I was expecting coming into the game. It's tied with Dartmouth Championship Football in terms of total score, but Even Wins has the better gameplay and fun factor, so it wins the tiebreaker of which is the superior game.

Next up I'll be doing a short side article on the Cybernetic Edition of Even Wins, which will be coming on Tuesday. After that, we have a rather oddly-titled game for Friday's full release. 

As a side, I hope everyone is enjoying the new article on the origins of video games. I had been feeling quite dissatisfied with my early work on that period of time for a while, and felt a revisit was due. I'll likely do individual pieces on all of those early games periodically, in the lead up to revisiting some of my early articles. My writing and research is much improved from two years ago, and I'd like to give those games the respect due them. Thoughts and feedback are always appreciated - I'd like to hear from you readers what you'd like to see out the blog, and what improvements I could make.

Don't forget - if you enjoy my blog, be sure to leave a comment and hit the follow button so you don't miss any updates!