We are tirelessly working on creating the best human-like bots possible and our 3.1 update is the latest iteration.
What is bots 3.1, and why so soon after 3.0?
Bots 3.0 was a major change to how our bots work. We introduced a deep way of influencing how the bots play, it was a massive undertaking and changed all bots in their foundation. The Christmas challenge with Santa gifting pieces and Gingy playing reckless attacking chess on his sugar rush was our first experiment with this, and it was hugely successful and fun. So we doubled down and changed literally every bot to be able to do this. The first taste of this are the new playstyles that you can really feel, you can read about it here. But we've laid the foundation for the future and now have tools we didn't have before.
Introducing such a big change comes with some challenges, and while we did extensive testing, we have quite a few tuning issues and a couple bugs. Since then we've been working day and night with gathering feedback and tuning and fixing the bots to get them into the shape we want, and with 3.1 we believe we're there.
Did we release 3.0 too early?
To be honest, probably. We did extensive testing, but a few too many problems slipped through for us to be satisfied with release. As an old Spotify software engineer, I'm not strange to shipping fast and being prepared to quickly fix problems, that usually a lot better than being defensive and having long development cycles. In this case, we should have slowed down a bit though.
Is 3.1 perfect and no more changes needed now?
We will more than likely never stop improving the bots. There's always something we can improve. Be it more playstyles, more accurate ratings, less (or more) blunders, better personalities. The possibilities are endless, which means that so is our work. That's part of the fun though, and I really hope you're excited to join us on the ride!
So is 4.0 around the corner?
No. We have some really exciting ideas, but now it's time to slow down a bit with major changes to the bots and instead evaluate their current state and make sure they're as good as they can be. There might be a 3.2 coming in the not so distant future, but the changes will be incremental and focus on stability and making sure the bots feel trustworthy and play quality games (at their rating).
What exactly did change in 3.1?
A lot. A whole lot. We introduced a couple of major systems in 3.0 that weren't functioning as intended, these have now been fixed in 3.1 and are really noticeable. For example an extensive "shuffle-protection" which is intended to keep games exciting at all times. This should sort out issues with, especially lower rated, bots playing without a plan, aimlessly shuffling pieces.
We also have a whole new category of weaker bots at around 1200 rating and down. These bots should feel beatable for even quite new players. We've had a hard time making convincing weak bots, but we think we succeeded now, and it should still feel rewarding to beat them. Or in other words, they won't just throw away pieces to simulate being bad, they play with intention, just pretty bad intention.
There are also a few unrelated fixes that we threw in with this change. We did for example get a few reports of bots severely neglecting passed pawns at times, which is a type of blunder that is both unrealistic and pretty boring as the game basically ends on the spot if one side gets awarded a free queen for nothing. This should not happen anymore, or rather, all bots will pay realistic attention to passed pawns from now on.
And a lot more, the bots should feel quite different, in a positive way.
How about the ratings?
To be very clear, a lot changed. So the bots you knew before 3.1 are probably very different now. Our aim with bots 3.0 was to increase the ratings across the board, and we absolutely did. However, after a few bug fixes and tuning, the bots went up quite a bit in strength, while their ratings remained the same.
We've now made another attempt at increasing the ratings, so you should hopefully see an increase in your own rating, while the bots should better reflect what you could expect.
Am I better now because I can beat bot X?
Unfortunately probably not. We did increase ratings across the board, which should generally make the bots feel a bit weaker, but we also moved a few bots around to cover gaps in ratings and playstyles.
The best way to see it that we have a whole new set of challenging bots that you need to find your foot against.
Going forward
I believe the bots are in a much better spot that they've ever been in before. We really hope you feel the difference. Give it a go! And if you find something we're always eager to hear your feedback, either at support@chessiverse.com, or even better join our Discord and talk to us directly. We've had great discussions there and a lot of the changes we've made have come as suggestions directly from users there.