We’re on the verge of releasing Let It Snow as our first game under the @thoseMetaMakers banner. We will be uploading it to the iOS app store the very moment the Apple Gods say we can!

Let it Snow is a particularly difficult game. We’re calling it hailstone hard, and it really is. The rules are very simple, but getting good at it takes some effort, and to do well in a particular game requires both high concentration for the duration of the game and a healthy dose of good luck. Just when you think you have mastered it, your concentration will drop and a run of bad balls will leave you reeling in the negative scores after minutes of play.

There’s been a mini-competition in The MetaMakers Institute to see who is the best at the game (to be the world record holder!) As the original game designer, and really the only person playing the game for a long time, I was hoping to hang on to the record. This morning I posted a time of 71 seconds, and I was really pleased at my achievement. But then later in the day, Pete posted a score of 68 seconds…


    

…which frankly took the shine off my achievement. But still, it’s nice to be a member of the two-person “Purple Snowflake Club” of people with a best time of less than 100 seconds – no one in the group has yet managed this other than Pete and me. As an aside, we were going to make the award for being in the club a golden snowflake, but – as Pete pointed out – yellow snow is never good.

The sad thing is that really neither me nor Pete are the world champions, as the AI-bot that Ed has built as part of Gamika Technologies can beat both of us easily. Or so I thought.

The thing is, we’ve made a few minor changes to the design of Let It Snow. The original game was made really quickly, and there were some minor mistakes that we’ve tidied up for the release of the game. The most important of these was not allowing snowflakes or raindrops to spawn on top of each other, as this causes a nasty-looking physics glitch. In stopping the spawning of flakes/drops so close to each other, we effectively spread the balls out more, which made the game that little bit harder.

This opened back up the question of whether the AI bot was still better than us mere humans, and the game was on. Swen did a great job today of resurrecting Ed’s AI-bot to play Let It Snow and he and I ran some tests to see just how good the bot is. And to our surprise, while it had no trouble joining the Purple Snowflake Club, its best scores were only around the 75 second mark.

I even gave the bot the ability to give up and start again: if it was on less than 10 points at 20 seconds, or less than 40 points at 40 seconds then it should just shrug its shoulders and click on restart. This meant that I could run many more playtests with the bot. After around 100 tests, I was beginning to think the bot had limitations. At around midnight, I ran one last batch of 10 tests…

Let It Snow is the game that keeps on giving – and yet again it led to a surprise when the bot stormed in at just under 60 seconds (which got rounded up). Here is the video of this momentous game:



In years to come, people will be saying: “Where were you when the Gamika bot achieved a sub-60 second Let It Snow time?”

Watching the AI master play this game, I was surprised at how things went. To do well in a particular game of Let It Snow requires a bit of luck, and the bot needs this too. I was expecting it to get the raindrops trapped at the bottom of the screen really early on, because of good luck in the spawning and a bit of fast finger work and then do pretty much nothing as the snow kept coming in a blizzard. However, the bot has to work pretty hard, and is constantly stopping the rain from clustering.

One thing to note for aspiring players out there is that in this masterclass game, the snow literally never stops falling the whole time, and there are very rarely more than 2 snowflakes on the bottom line. The bot plays differently to me, as I tend to wait until the flakes/drops have settled before I start to try and get it to snow without rain. This makes it more of a puzzle than an fast reaction game. Looking at the bot’s gameplay, however, I may have to change my strategy if I’m going to get any better at this game.

It looks like me, Pete and the rest of humanity may never beat the bot, but maybe I can take back the human Let It Snow crown for a while. There might just be light at the end of the tunnel: I just achieved a score of 70 seconds!

AI beats man in last ditch Let It Snow thriller