Does anyone use the Zillions Of Games program?

http://www.zillions-of-games.com

I'm tempted to sue them for false advertising! Zillions? Pah, more like a 50 or so. But still, it has many chess variations available. Not just xiangqi and shogi but variations on Western chess. The playing strength isn't all that strong, but then having a program that plays so many games well is definitely too good to be true. I

It also counts the games that you can implement using it. Not only the games that others happen to have implemented with it for you. So I don't think there is anything misleading here.

That being said, I don't use Zillions. The scope of Zillions is very wide (including cards and dice), and I am mostly interested in proper Chess variants. So I prefer to use WinBoard, in combination with configurable Chess-variant engines like Sjaak II or Fairy-Max. The latter alone already supports 30 western Chess variants. Sjaak II supports an even wider range of Chess variants, including those with drops, such as Shogi, and zoned boards, such as Xiangqi. These engines are in general also stonger than Zillions. And for games with very special rules (such as Chu Shogi) there often are specialized engines available.

Just curious, how strong is it in e.g. ELO terms?

That hasn't really been measured, because it doesn't support a standard protocol for chess engines (like CECP or UCI), and thus cannot play automated games. And for a reasonably accurate rating determination you need many hundreds of games.

In addition, it would not be equally strong in all variants it can play. It has a heuristic for estimating piece values, and this heuristic is known to frequently fail in predicting reasonable values. It does not seem to understand divergent pieces, (with separate moves for capture and non-capture), so for pieces that can be dropped (like in Shogi) is uses ridiculously large values.

Its search is also not very stable: in variants with too powerful pieces it suffers from search explosion. There are several implementations of Tenjiku Shogi in it, but none of those can finish the 1-ply iteration in the initial position even in 10 min, and as a consequence it then plays the first move of the static move ordering. (Which sacrifices a Bishop General on the Side Soldier.

For orthodox Chess and 10x8 Capablanca variants it probably plays reasonably well, both search-wise and evaluation-wise, but even then it tends to lose from Fairy-Max (which for normal Chess is the same as micro-Max, rated 1868 Elo at CCRL), in the few operator-mediated games these played against each other.

I wish alpha zero was open source it technically could play any variant

There does exist an open-source implementation of the Alpha-Zero algorithm, called Leela Zero (for Go) or Leela Chess Zero ('LC0') for orthodox Chess.

Of course you would have to adapt the source code if you want it to play a variant (board size and move generator, possibly other rules such as winning condition). And then train it. Which for Leela/Alpha Zero is no small effort; Google had its globe-spanning network of servers to do it, for LC0 it has been a massive group effort of people donating CPU time.

Did anyone realise how weak the engine in chessvariants.com for atomic is? It can not even see a mate in 4!

I suppose you are referring to the Interactive Diagram that goes with the article about Atomic Chess. Then it really should not come as a surprise. The I.D. there is designed as a sparring partner for readers who have just acquainted themselves with the rules, still giving them a good chance for beating it. You don't want to turn away people from a variant by clobbering them on their first experience.

So by default it is set to search 2.5 ply ahead, adjustable between 1 and 4 ply through buttons below the Diagram. And actually ply is already pretty challenging. But to see mate in 4 requires a look ahead of 9 ply. Most humans would not be able to calculate that deep, and would have no chance at all against a program that searches 9 ply.

Badeebadabba wrote:

Zillions? Pah, more like a 50 or so.

While it is true that the Zillions of Game (ZoG) program file, specifically or .1, installs a Rules folder with fifty games included in its Rules folder with the application, at the time of this comment there are 3084 games submitted by the ZoG Community of all kinds. It’ll soon be 3085 if I get my act together and submit a game of my own design, but I’m not up to speed yet on the ZRF rules file format to tackle that task.
Many are chess games. Many aren’t implemented very well by folks not totally familiar with logic programming (a form of declarative language involving LISP-like S-expressions, more data than code in a sense). Many have horrible visual and audio choices, which is one of my pet peeves about ZoG gameplay.
To suggest that there are only 50 games, limiting the scope to only those included with the Demo apps, is misleading.
Although ZoG development ended in 2003 with the release of ZoG .1, the community is still quite active today (September 5, 2025, for the record).
You can view game pages and download Zip files associated with all those games from https://zillionsofgames.com/games/index.html

The playing strength isn't all that strong, but then having a program that plays so many games well is definitely too good to be true.

It is indeed quite true that the default AI player is quite naïve. It’s also important to consider that Zillions of Games (ZoG) is built to be a general game playing system (. GGP project in the Stanford Logic Group, sadly abandonware and not currently being developed as is also the case for ZoG which hasn’t been worked on since the release in 2003). Neither of these general game playing systems are designed specifically for chess and chess variants. 
The built-in AI is quite simplistic and looks ahead as far as it can in a fixed (or infinite) allotment of time per move. And if the AI is playing a game where the rules state that players place (or move) two pieces in a turn, as is the case for the Connect6 Gomoku variant my son and I are quite fond of, then after it thinks about how to place its first stone, it thinks all over again about where to place the second stone.
That specific game is played on a 19x19 Go board with black placing one stone first, then white placing two, black placing two, and so on until someone has created 6-in-a-row along any direction, including diagonals. [And overline wins are allowed if you are familiar with what they are and how they are not considered wins in traditional Gomoku.]
Connect6 is an AI killer because the only rule for placing a stone is that the position must have previously been empty. So there are 19x19=361 possible opening moves by Dark. And then 360 possible first moves by Light, and 359 possible second moves by Light, and so on. That fans out exponentially to so many positions to be considered just to get to 2-ply or 3-ply lookahead… not sufficient to spotting possible wins… that the built-in AI plays the worst games one could imagine.
That said, the ZRF rules format, in addition to setting up the sounds and graphics of the game, and the abstract game rules using LISP-like S-expressions (similar to what Stanford’s GGP software supports), also allows hints to the AI for things to focus on and things to avoid. So, one possible minor optimization might be to say that if a piece is placed more than five positions distant from any previously placed pieces, then it’s likely a spurious move not worth spending any time thinking about. Similarly, in placing opening moves you likely don’t want to play them right up near the edges of the board where they have limited power and flexibility.
I noticed that the Connect6.zrf rules file does not contain such AI hints, while the Backgammon.zrf rules file does. I’m such a fanatic now about that specific game that I’m considering writing some AI hints once I can figure out what they are. They call it ‘forking’ when you can line up a 3-in-a-row with a 3- or 4-in-a-row, essentially the same as setting up a checkmate in the next turn. So one can set up simple rules to catalog all your pieces by how they align “in a row” with others, and to spot threats similarly by looking at the “in a row” alignments of the other player’s pieces.
Through such hints, it should be relatively straightforward to accelerate the move searches by at least one order of magnitude, and possibly two, leading to significantly deeper lookahead.
That said, the tendency to fan out to a large number of moves to consider, well exceeding that of chess, may mean that even with the hints, that the quality of the bot won’t be all that improved. But maximizing your “in a row” opportunities while minimizing those of your opponents may lead to improved choices, combined also with the simple hints to avoid spurious moves or to play too close to the edges.
I mentioned two ZRF files. Here are links to them, should anyone wish to study them. The Connect6.zrf file is quite short and straightforward to read with a little familiarity with ZRF rules syntax. It is linked to in source [1] below.
Finally, I’ll say that the owners of the ZoG site are now taking arbitrary donations to install codes to unlock all the features from the demo version of the app. I can’t recall what I donated to become legit, perhaps around USD$10, but I purchased my son’s key for a mere buck, and you might even be able to go lower, though I haven’t experimented with whether there is any minimum donation. It sounds like the website owners are supporting ZoG, but not investing in making much money for their efforts.
___________________________________________________
SOURCES 
[1] https://courses.cs.duke.edu/spring06/cps108/Assignments/06_vooga/zrf.pdf
[2] ;id=1445
[3] ;id=107

[2] ;id=1445
[3] ;id=107

The URLs to those two specific games, Connect6 and Backgammon, apparently were munged here in copying and pasting them. You can locate those game pages from which the ZRF files can be downloaded, by visiting the ZillionsOfGames.com website and then searching for ‘Connect6’ and ‘Backgammon’, respectively, to locate those pages I tried to link to.