Create a New Discussion Button

The 1st forum, where all the questions will be directly answered by Grandmasters!

The 1st forum, where you’ll be rewarded for your answers!

To computer or not to computer - that is the question...

I want to open a discussion on this as I feel there is a little too much of a hardline against engines.

The prevailing wisdom is that players under (2200 or thereabouts) should avoid engine use or at least do some serious analysis first. There are however I believe several assumptions underlying this:
. That the student has a coach who can answer why a move is good or not if the student is unable to come to a conclusion no matter how many of such positions there are
. That the student has a large amount of time to do such analysis
. That 2200 equates to some kind of level without explaining what the underlying level means
. That bad engine discipline leads to overuse (cf. as using a calculator when you should be learning your times-tables)
. That comments and material in courses and books is always accurate and appropriate for the student's level

. That weaker players will not be able to use the engine appropriately or will misread it
. That it's unnecessary for below 2200 play
. That becoming a Grandmaster is the goal

First my own feeling on this:

If you are using an engine appropriately for your level and you are aware of the limitations of an engine, and you do not have a coach on hand, then some engine use is beneficial, probably more so than not using one at all. It can be used to check your own analysis and assumptions as well as speed up your ability to work through material when time is limited.

I think if you ask the top trainers most would say don't use a computer, but they are already coaching their student and some of their students have many hours each day to spend doing things a computer would short-cut (the should cut might cut out some practice or learning of course). However I think this wisdom doesn't always translate well to the club player with a few hours each week to spend on their chess - plus limited concentration span for study when it's a hobby, hence the discussion.

The limitations of the computer need to be understood:
. Computers are bad at endgames, better now, but still need to improve unless it is a tablebase like or tactical position (I consider tablebase position use databases not engines).
. Computers often do not tell you straight out (without some investigation) why one move is better than another (unless the line clearly shows a tactical error etc). Certainly no explanation is given and the few tools that do are still in their infancy.
. Computers perform better in tactical positions than positional (though this is changing) because of the insufficient evaluation function coupled with very deep search ability.
. Centipawn measurements are a little arbitrary
. Psychological or human difficulty factors are not considered in an evaluation (including gambits) which is why the Benko or Sicilian variations get a hard time
. Sometimes computers need to run for a while to get a true evaluation even in positions that aren't so profound to humans.

What the above shows is it's very easy to misuse/read a computer, which is where a fair bit of the adage of don't use a computer comes from.

Second I take a little bit of disdain to mentioning ratings as if they are levels of skill. A 2200 is something that is on the face of it strong, but there is a big rating difference between a player who got 2200 by playing mostly strong players and the few draws or very occasional win got them there, versus someone who plays in clubs and lower level tournaments and the constant barrage of weaker players which the occasional loss or draw keeps the rating lower. The rating pool (sometimes artificially constructed by choosing which games a player plays) is very important. Similarly one might be 2400 in the Sicilian, but 2000 in the Benko, or 2400 in the endgame and opening, but 2000 in the middle game. They might be 1800 after a day at work, but 2000 on a weekend. The weakest point is probably going to determine the rating more than someone who is all round good. In addition rating does not equate with experience: there are plenty of players in the 1800-2000 range some who are pretty good (just not consistent or hobbled by just playing in a small pool of 1600-2000 players) who have been playing chess for 40 years, and there are plenty of 2200s who have been playing just a few years. The question is what does this 2200 really mean in terms of chess skill or maturity.

I feel the time factor for study really needs to be explored further. As I've mentioned elsewhere there are those who can spend 5 hours a day on chess, whereas some only get 5 hours a week. In the latter case it's 'getting the best bang for your buck'. Spending 15 minutes analysing to get an answer why a move was bad (after spending 15 minutes already looking at in the game but without the hindsight of what happen), is often too much of a task, so soon no analysis takes place at all. Similarly trying to understand master moves in a book of 500 positions/fragments/games will take years for just one book if you take this approach with only 5 hours study available per week - most will give up or not finish the course - and while they are studying it nothing else is getting worked on.

I'm somewhere in the middle for study time and here is where I use an engine:
. As a blunder check after online games, and to understand whether a marked inferior move was inferior and to try to tease why another move was superior.
. When I can't understand why a move was/wasn't played after a bit of thinking
. To check my own analysis
. To check for errors in published material before I commit it to memory (some errors in positional based material aren't necessarily a problem if the pattern is intended to be conveyed not the specific example).
. Openings when looking at the database for what was played as there is usually too much complexity for a non-master to properly make a decision on whether a plan or move was a bad one. Plus just because it's a game in a database played by someone strong, doesn't mean there are not errors.
. New ideas or things to consider in a position that I haven't seen before

Examples of where I would say is bad engine use:
. Studying tactical/analysis material before having a proper go at solving it

. To quibble over a couple of centi-pawns whether one opening move was better than another
. To find deep lines to study in openings so 'you know more' beyond what is appropriate in the games you have been facing
. Before actually playing through a published game at least once to get a feel of what went on
. When your eye is more on the engine than the material itself
. Getting definite evaluations of endgames
. As a substitute for thinking (aka the analogy of using a calculator vs mental arithmetic)

Finally what level would I say is appropriate (beyond blunder and material checks) from my own personal feelings (without the view a grandmaster or a coach has):
. You make few blunders, certainly nothing too serious in longer games
. You understand tactics well
. You understand positional concepts well

At least at this point you have the ability to question a computer evaluation as well as less likely to use it as a first point of call. Whether that is 1700, 2000, 2200 of course depends on the factors I've mentioned as well as the individual.

PRO Members