Playing the CPU - Honest Viewpoints

It seems like the CPU is often referenced a lot but doesn’t get the recognition it deserves, and I don’t understand why. Maybe that statement alone will rub someone the wrong way - what do I know right..?

Anyway, what I hear most often about it is that it’s not a human opponent and doesn’t think like a human opponent. Either that, or during analysis it’ll recommend a move that doesn’t make any sense and miss seemingly obvious opportunities. Well, humans also make a lot of strange moves that don’t make any sense, especially beginners like myself. Humans also play moves that cost them the game.

But isn’t the whole idea of improvement to recognize tactical advantages and opportunities just as you would in puzzles? So then it's good the CPU does those moves, so you will recognize the patterns and capitalize, right?

I don’t believe the idea was for it to play like a human. Yes, there’s AI now that’s trying to get closer to making ‘human-like’ decisions, but their purpose is to support and provide assistance, not to be human. When GM’s make up their own moves it’s “the genius” at work. If they fail.. well, then there’ll be a breakdown as to what went wrong. Some videos I’ve seen give praise to the CPU, and it’s “non-standard” play - but more often than not I hear playing the computer won't really help you improve.

It’s like - there’s almost the sense that it’s supposed to be perfect, but it’s not...and never will be.  AI is constantly improving. Just throwing some stuff out there out of curiosity, hoping to get a few honest replies on the topic.

This is mainly directed at experienced players. Is there some kind of spite towards the CPU? Why does it seem as though - on one hand it’s accepted, but then on the other hand, even though it could easily beat a great majority of human opponents, it gets pushed aside because of its inaccuracies?

One of the main complaints about bots is their play is inconsistent. These days there's no such thing as a 1000 rated engine, so they take a 3000 rated engine and (one way or another) force it to make mistakes.

When someone your rating blunders a knight you'll probably win, but when a 3000 engine blunders a knight, it might play like a master after that... so the most reliable way to win is don't do anything, just keep defending and making moves, and wait for it to blunder again... obviously this is not a great way to practice because this is not how to beat humans.

---

I do think there's some benefit in playing against engines that are a little too hard... for example maybe it can beat you 8 or 9 times in a row, but it can't beat you 100% of the time. I think engines like this can be useful to help players get out of the habit of playing for tricks (like hoping their opponent doesn't see a threat), and help them form better calculation habits (the engine will reliably punish simple tactical oversights).

Even this can be detrimental in the long run because you don't get much practice winning a won position. When you're ahead opponents will try various things to make it harder for you to win... not only will a tough engine beat you most of the time, but it doesn't know how to create the sorts of practical defensive tries you'll see from humans. For example a human might decide that they have nothing to lose, so they attack your king as much as they can just hoping you miscalculate. An engine will see the attack doesn't work and wont force you to prove it.

Oh, do you mean why don't pros play against full strength engines?

That's sort of like asking why professional runners don't train with horses... horses can run fast right? So why not? The answer is because not all running is the same. Horses run in a fundamentally different way.

This is one thing that new players misunderstand about cheat detection... there's a difference between playing very good moves and playing engines moves. The majority of positions in chess have more than one "best" move in the sense that there are only 3 real evaluations a move can have (winning, draw, or losing). So it's not only that an engine plays well, it's that it plays like an engine. The way it finds moves, and the types of moves it finds, are sometimes not instructive or practical for humans.

In the end, the engine is a very useful tool... and like all tools it depends on how you use it. Playing games vs the engine is not a terrible idea, and with specific goals in mind it can be useful, but that's generally not what the tool is best for.

Great answer and excellent points. I understand the logic behind what you’re saying - there’s no substitute for those desperation attacks/counters that would come from a human opponent. I didn’t consider that at all and was only thinking about blunders and miscalculations.

As far as Pro’s playing the CPU, no I didn’t really think they would, it just seems like higher tiered players will often mention that it’s not ideal to learn against the CPU and I couldn’t understand why. When giving an example of GM’s doing random moves your first explanation really cleared that up. Liked that race horse example too btw. 😏

Rather than sarcasm, which was probably warranted because of my naive experience, you provided real knowledge on the topic and I really appreciate that. You mentioned two things that I’m also curious about  - playing stronger AI and using the CPU with a specific goal in mind. I get the impression that some of the tactics and strengths it has would be too advanced to really take hold - kind of like the more difficult puzzles. What do you think? Also, when playing the CPU what goal(s) should I keep in mind?

I like talking about chess, so when someone asks a legitimate question I'm not rude to them

-

-

My reply to this became monstrously long, so I actually broke off about 2/3rds of it and turned it into a blog post. To read that you can go here:

https://www.chess.com/blog/llama47/basic-reasons-to-distrust-the-engine

---

The part I didn't include in my blog post was your question about how to use an engine:

How to use an engine
If you're going to play against it, then I think it's best used to practice calculation. Maybe something like write down a few candidate moves, then calculate them, and prove to yourself that there are no tactics that punish the move you want to play (but don't play passive defensive moves, play active moves that can't be punished). Inevitably you'll make a mistake, and as you repeat this exercise hopefully there will be some common element tying your mistakes together. For new players it might be something simple like missing knight forks. For more advanced players it may be a certain type of tactic or a certain type of position.

More useful is using an engine to analyze... and since there is more than one good move in most positions, many players set the engine to show its top 2, 3, or more moves. Some wins require elaborate calculation, and some wins are so natural that they require almost none... and sometimes a move that is objectively equal or gives the opponent an edge, is the most practical way to play. A player uses their knowledge of chess when decide which engine lines are worth exploring

Liked and Subscribed - thanks again Llama

Could only get up to point number 2-3 in your blog, even after trying the moves in analysis, I'm just not ready yet. It was an impressive breakdown nonetheless.

Thinking more about what you said in that last post - busting out the pen and pad for chess..that kind of dedication easy to admire. I have one more question though, Do you want some of this? 🤣 I see your ranking.. [smh] those are just numbers man, and I don't even care. j/k j/k 

One of the most helpful exercises I did was like that, writing down my analysis and then checking it (I did it with tactics puzzles). But yeah, at lower ratings the most important thing is to just play a lot of games. That way you get used to everything.

I don't mind playing unrated games vs people... random challenges like "let's play right now!" probably not, but if we set up a time that's fine, and if not me then I know @energetichay is often playing anyone who wants to play (he'll only do unrated daily games i.e. not live games).