Chess engines' playing styles

1/2/2008 – What's a "chessplaying style"? How do the styles of different chess engines differ? These are the questions tackled in thie latest ChessBase Workshop, in which our columnist defines the playing styles of our various chess engines in basic layman's terms. Workshop...

ChessBase 14 Download ChessBase 14 Download

Everyone uses ChessBase, from the World Champion to the amateur next door. Start your personal success story with ChessBase 14 and enjoy your chess even more!


Along with the ChessBase 14 program you can access the Live Database of 8 million games, and receive three months of free ChesssBase Account Premium membership and all of our online apps! Have a look today!

More...

The topic of this ChessBase Workshop has been discussed before, but I'm still asked about the subject so frequently and so routinely that I've decided it's time for another look. We're talking about the playing "styles" of ChessBase's various chess engines.

First of all, we need to get one thing straight. Chess computers play like, well, chess computers. We flogged this particular horse to death many months ago in our interactive debate about "intelligent mistakes". Computers do tend to play like computers no matter what. When we say that a particular chess engine has "good positional sense", we speaking relative to other chess engines, not relative to, say, Wilhelm Steinitz. Chess engines have come a long way just in the last decade in the area of chess "knowledge" and "positional understanding", and they're definitely many light years ahead of where they were in the late 1980's -- don't even get me started on the '70's, when the "chess-challenged" BORIS and the Atari 2600 Chess cartridge were the toughest (affordable) computer opponents in town. Chess engines still have a long way to go in their understanding of strategic themes, but they are getting better with each passing year.

I'd love to be able to tell you that the "relative scale" we're using in this discussion is based on some firm immutable scientific criteria: "We subjected these engines to 20,000 games and a 10,000 position suite of hand-picked individual positions," blah, blah, blah. I'd love to say that, but it's not true. The whole thing's pretty subjective, based on dozens of games I've played against chess engines, compared with the many hundreds more I've played against human opponents and the several thousand I've replayed from ChessBase databases. Judging from what I read on Internet message boards, however, my experiences and impressions jibe closely with that of dozens of other users who regularly play against chess engines and use them for game analysis. So while one could always argue with the conclusions, said conclusions aren't just being pulled out of thin air.

What constitutes a "playing style"? Here again, the answers are infinite and hardly objective. In general, chessplayers tend to narrow the argument down to a couple of common themes: how aggressive a particular player tends to be, and whether he or she plays in a sharper "tactical" style or a slower, more methodical "positional" one. These aren't the only criteria one could use to characterize someone's chess play, but they tend to be the ones which are most readily understandable to beginning players and advanced players alike; these are the "sliding scales" we'll use in characterizing the four chess engines discussed here.

Chess engines in general these days aren't as blindly aggressive as they once were. Long-time Fritz users still talk about Fritz2, which for my money was the most aggressive chess engine ever. Fritz2 only knew how to attack and did so ferociously (and to the point of foolhardiness), with what might be described as a "blind rage" if we were talking about a carbon-based life form instead of a computer program. Present-day chess engines aren't nearly as wildly aggressive; this characteristic has been replaced in at least one engine with something a lot more interesting.

I'm talking about Junior, easily the most tactics-oriented of the four engines which ChessBase currently offers. The wild aggression of earlier chess programs has been replaced in Junior by speculation; Junior is an engine which has been known to sacrifice material speculatively rather than concretely. Almost any chess engine will sacrifice material for an immediate gain; for example, if a Queen sacrifice results in a forced mate-in-two, you'll see a chess engine sac the Queen with no problem. Junior, though, will sometimes sacrifice minor material to clear a line or to otherwise free its game, which is something almost unheard of among chessplaying programs (I can think of just two other commercial programs which were known to do this). This behavior isn't the result of poor programming or a flaw in the engine's algorithm -- it's a deliberate programming choice which allows Junior to see possibilities which other engines often miss. It can lead to some pretty wild unexpected play; when I think of Junior, the first word that comes to mind is "surprise". Of the four ChessBase engines, Junior's the most fun to play against in my opinion -- the program often surprises me and definitely keeps me on my toes. I don't use Junior for analysis very often, but it's a whale of a lot of fun as a sparring partner.

Next on our "sliding scale" (as we move from tactical toward positional) is Fritz, ChessBase's "flagship" playing program. Fritz has always been a tactical monster, even leading in past versions to some baroque play in which Fritz would offer as game analysis some ten-move material winning variation which no human could ever spot. Fritz has changed greatly over the last decade and a half. It began as a fairly simple little chess engine against which you could easily draw a game if you knew the trick. By its second version, it had mutated into a fearsome attacker (although a skilled player could take advantage of Fritz' blind aggression to turn the tables on the program). By the mid-1990's and its third version, Fritz' aggression had been curbed in favor of a more solid (but still tactical) style of play which took advantage of faster computer processor speeds. As we reached later versions of Fritz after the turn of the century, we saw more chess knowledge being built into the program's set of instructions to give it a better handle on positional play -- the program evaluated fewer positions per second than its predecessors, but "understood" more of what it saw.

The newest two versions of Fritz (10 & 11) have seen another change in its programming philosophy. Instead of being geared toward performing well against other computer chess engines (as has been the case with almost all chess programs since the early 1990's), Fritz' main focus these days is high performance against human opposition. Coupled with the newer training features which have been added to the program's GUI, Fritz' present goal is to be the best available training package for the improving human player. Consequently, while Fritz is still a pretty tactical program, the additional positional chess knowledge programmed into it has placed it close to the middle of the "tactical to positional scale" with a slight tilt toward the tactical.

One of the most talked-about chess programs on Internet message boards is Fritz' "mirror image" on the "tactical to positional scale", namely Hiarcs. The comment you hear most often about Hiarcs is how "human-like" it plays. I've been saying the same thing ever since Hiarcs2 was released back in the early 1990's. But what does "human-like" really mean? I had a hard time defining it, but having played many games against Hiarcs I certainly believed it and repeated that assertion. After considerable thought, I'll try to summarize its style of play. Hiarcs' style of play feels "natural" because it's not an overly-aggressive program. It doesn't play weird-looking moves which then actually lead to material gains (which was a trademark of Fritz' style through several of its past versions -- what I referred to above as "baroque variations"). While having great tactical insight, Hiarcs also knows the value of many basic positional motifs. Hiarcs is like the older master-level player down at the chess club, the guy you always think you're getting a decent game against until he drops the hammer at the end, crushes you, and leaves you shaking your head and wondering what just happened.

Easily the most positionally-oriented of the four engines is Shredder. If Junior is the Tal of ChessBase engines, Shredder is the Petrosian. Shredder never takes unneccesary risks and plays a rock-solid game which is very difficult to pick apart. One hallmark of a great chessplayer is the ability to cause one's opponents to make mistakes; I believe this is why top-level players have such a tough time with Shredder in competitive play -- it's very, very hard to trip up the program. Though I'm a far cry from anything other than "mediocre" as a player, this solid play is why I seldom play games against Shredder -- it's like beating your head against a three foot thick concrete wall. But Shredder is my most trusted analysis engine. If I'm using "Compare analysis" to analyze one of my own games, Shredder is invariably one of the engines I select to perform the analysis, and it's usually my first pick for the single "Blundercheck" engine. It's also my engine of choice when doing "on the fly" analysis within the ChessBase software. Why? As I said before, all chess engines are primarily tactics-based but Shredder seems to have the best positional knowledge of the four main ChessBase engines. If any of the four is going to point out a strategic/positional error I've made, Shredder is the most likely to do it.

Once again, please bear in mind that "tactical" and "positional" are used in this column in a relative sense; no present chess program is going to fool you into thinking it's the second coming of Wilhelm Steinitz. On a relative scale from tactical to positional, the programs stack up in the order presented in this column, to wit:

  • Junior
  • Fritz
  • Hiarcs
  • Shredder

Until next week, have fun!

You can e-mail me with your comments on ChessBase Workshop. All responses will be read, and sending an e-mail to this address grants us permission to use it in a future column. No tech support questions, please.


© 2007, Steven A. Lopez. All rights reserved.

Discussion and Feedback Join the public discussion or submit your feedback to the editors


Discuss

Rules for reader comments

 
 

Not registered yet? Register