A Brief Guide

to
Engine Analysis

Advice compiled by Tartajubow

 This booklet is a compilation of my Blog posts concerning how to
properly conduct an engine analysis. It consists of advice on analyzing which
engines from such strong players as Jonathan Rowson, Steven Ham and Robin
Smith.

It is advice which I have found helpful when playing correspondence chess on

sites where engine use is both allowed and necessary if you are going to
compete at higher levels of play. Such sites include ICCF, Lechenicher
SchachServer, IECC and probably others.

Besides that, the advice will be helpful to anybody who uses an engine to
analyze their games and wants to get the best possible results.

Subjects covered include:

 Good Engine Analysis is an Art

 Starting with the basic settings:
o Calculation time
o Threshold
o Reference Database
o Side to Analyze
o Annotations
o Training
o Erase old annotations
o Hash Tables
 Engine Comparison
 Some Basic Hints
 Sparring Partner
 Multi-variations
 Multiple Engines
 Engine Tournaments
 Deep Position Analysis
 Full Analysis and Blunderchecking
 Opening Advice for Correspondence Players
 Analyzing Won and Lost Games
 Pre-existing Analysis: Legal or not?
 Making You Own Database and Opening Book
 Critical Points
Capablanca once wrote, “Chess books should be used as we use glasses: to
assist the sight, although some players make use of them as though they
conferred sight.” The same thing could be said of chess engines.

On various forums wherever correspondence chess is discussed

some of the comments are interesting. For example: one player, rated around
2100 OTB, stated he had given it up because he was not good enough to take
advantage of any “weak” moves played by engines, adding that no matter how
much time he spent on the game, even using his own engine, he simply could
not tell a “good” engine move from a “bad” one. He added that an OTB master
and 2400 rated CC player had the same problem.

His opinion was that in order to successfully play CC you have to be strong
enough to know when an engine generated move is not the best and for that
you need to be rated at least 2300.

In response, one FM replied that endgames are still the engines Achilles Heel
but you need to be a master of endgames to succeed against engines. He also
pointed out that in closed middlegames an experienced strong player can come
up with a better plan than engines which often do nothing but shuffle pieces.
Additionally he felt that in order to take advantage of an engine‟s opening book,
you need to be rated about 2300 OTB.

This FM, and many other strong CC players think that to be successful in CC
these days, it‟s the openings where games are won or lost. You need to find a
line with long term strategy or an endgame the engine can't understand or play
an opening innovation where the advantage of the move is over the engine‟s
horizon.

I have to agree with this FM and as a result, most of us won‟t be very

successful at the higher levels of CC play. In fact everything he said is in
agreement with CCGM Robin Smith‟s statements in his book, Modern Chess
Analysis.

Another interesting observation of this player was that the rating difference
between an 1800 rated and a 2000 rated CC player was practically zero and
the difference between an 1800 and 2400 is quite small…probably the only
difference was endgame knowledge. He went on to add that CC still remains a
good way to improve because it requires a LOT of study, but if all you‟re looking
for is wins and a rating then you will find it very frustrating.
It‟s quite clear to me after returning to CC in 2004 after an absence of about 12
years that with the proliferation of engines, legal or not, I am facing the same
problems as the original poster. I‟m simply not good enough to outsmart a
chess engine in the opening, or force the game into a closed middlegame where
I can plan a winning strategy twenty moves in advance while the engine
twiddles its thumbs, aimlessly shifting pieces nor am I an endgame expert. Oh,
I can play K and P and R and P endings pretty good because years ago Basic
Chess Endings and Peter Griffith‟s Modern Chess Endings were my best
friends, but that‟s not good enough because I still lack a 2300‟s understanding
of these things.
The short version is that if you aren‟t 2300 OTB you are reduced to 1) playing
only engine generated moves which is no fun or 2) trying to find another move
that probably won‟t work and likely will lead to losing the game which is no fun
either. Since I am not good enough to play CC at the top level and am not
looking to improve my game, only have fun, why bother with this form of chess
at all? I don‟t know why I do it, but I do.

Good Engine Analysis is an Art

...but for most of it‟s more like a kindergartner‟s crayon drawing.

I use Fritz 12 with the Houdini 1.5a x64 bit version. If you have a single core
processor you will have to use the 1.5a w32 bit version. So far I have not found
any free engine that consistently defeats Houdini and the Fritz 12 engine does
not even come close.

Opening books really do not matter because for correspondence play or

analysis I use a database. There are many good commercial and free DB‟s
available, but I have always used my own. The main thing is to make sure you
keep it updated. For this you can use a site like The Week in Chess to
download new games every week.

I also use the DB that contains all the games ever played at the now defunct
International Email Chess Group because it contains openings used by some
of the world‟s top CC players and their games are good models to follow. I don‟t
search for my own opening innovations like most top level CC players because
I‟m not one…I wisely rely on their research! Openings will need to be solid
mainline openings and not dubious gambits. Successful CC usually means
winning in the ending and to do that you are best advised to play positional
chess.
I prefer to check potential endings at the Shredder Endgame Database. There
are tablebases you can download but access is very slow and if you do choose
to use them download time can run 30 to 80 hours and at least 50 GB of hard
disk space.

Chess Life had an interesting interview with CCGM Stephen Ham. Ham stated
in the interview that his chess tends to be technical with an emphasis on long
range planning and his wins are usually the result of the accumulation of
small advantages resolved in the ending. When analyzing with an engine he
investigates positions that interest him and not the engine. He added that
stron) players know which lines are worth looking at and which are not. He
also added that he relies on his own evaluation of the position and not the
engine‟s which he claims are often unreliable. I have heard other very strong
players make the same claim.

One interesting statement he made was that in closed positions with

hypermodern strategies (e.g. King‟s Indian) engines often play weakly and their
evaluations usually favor White which are unreliable and incorrect. Perhaps
this explains why in top level CC you often see a lot of Queen‟s Gambit‟s and
Nimzo-Indians. He also advised that when playing CC one ought not be
influenced by the engine...often his moves were not even in the top ten engine
choices. This is in line with what I heard another CCIM say: The initial search
for moves should first be broad, not deep. Only when one has selected moves
that look promising should deep analysis begin.

For most of us, we enter a game into our engine, let it analyze at 10 seconds
per move and call the result “analysis.” But, whether one is playing CC on a
site where engine use is allowed or simply analyzing games with the idea of
improving, serious analysis is time consuming and requires considerable skill.

In any case, there follows some generally accepted good advice on how to
conduct a proper engine analysis which applies whether you are playing
serious CC or just analyzing your own games in an effort to improve.

Starting with the basic settings:

Calculation time
The more time you give the engine to analyze, the better the analysis. Best
results will be obtained by overnight analysis but most of us are not interested
in spending that amount of time so at a minimum 60 seconds per move is
reasonable. Still, even at that rate you are looking at 2-4 hours per game.
Threshold
Is a value expressed in 1/100ths of a pawn, and determines which
recommended improvements you receive. For example, if you set this value for
“500”, you won‟t see a recommended improvement unless the game‟s actual
move lost a Rook or more.

For beginners a setting of 100 is good because it means if you made a move
which lost a pawn or more and the engine will show you an improvement.
Stronger players can use a setting of either 50 (half-pawn) or 30 (moves which
lost the equivalent of a tempo.
The higher the setting, the fewer moves you‟ll see analyzed. Using settings
lower than 25 means you will see a lot of analysis, but most of it will be
meaningless.

Reference Database
The engine will search a specified database for similar games and place
opening references into the analysis. Remember though that this will only be
beneficial if you have a database that includes an opening key and the stronger
the players, the better the suggestion is going to be.

Side to Analyze
Always set to Both.

Annotations
Verbose will add short, descriptive phrases at various points during a game
that are inane and meaningless. Skip this setting.

Graphical adds colored arrows and squares to the board for emphasis. Skip
this one, too.

Training
This setting adds timed tactical training questions to a game, especially in
cases where a gross tactical blunder has occurred. Lower rated players may
find this setting quite useful.

Erase old annotations

Wipes out any existing game annotations before analysis begins. Don‟t check
this box if you want to keep any notes to the game.

Hash Tables

This feature is important. It‟s basically a storage system where the computer
stores in a temporary file the positions it‟s already analyzed and their
evaluations. If the engine reaches a position it‟s already analyzed by a
transposition it remembers that position‟s evaluation and not waste time
analyzing it again. This allows the engine to analyze more deeply.
Don‟t set the value to the maximum because there are other processes that
may need to utilize RAM. For best results set it at 75% of the available RAM.
Remember if your hard drive light is flickering and the computer is making a
clicking sound the hash table value is set too high and you‟ll need to lower the
setting.

Engine Comparison

Engines are very strong at tactical analysis, but weaker in quiet positions
where strategy is required.

The endgames of chess programs are often enhanced by the use of endgame
tablebases. But to be really effective these tablebases require a tremendous
amount of disk space. As mentioned above, if one needs ending analysis, use
Shredder‟s online 6-piece tablebase.

The most popular chess engine protocol under use and probably supporting
the strongest engine play is UCI (universal chess interface). These UCI
protocols are stronger than the old XBoard/WinBoard engines.

Some Basic Hints

1) don‟t rely on the fact that just because one side won it means his moves
were the best. Mistakes affecting the outcome may have occurred later in the
game.
2) even with engines, long analysis can be wrong analysis
3) human interaction is necessary when analyzing with an engine.

Most people running an engine analysis of a game simply have Fritz run a full
analysis for a set time limit per move or a set ply-depth, but that does not
always give satisfactory results. Below is a list of some methods that will give
better results, but in the end, if one is going to learn anything or hope for
decent results in high level CC events where engine use is allowed, then it is
going to require interactive analysis where you work with the engine in trying to
ferret out the best move in any position. And that is going to require more than
just plugging a game into Fritz and letting it do its thing. It‟s going to require
patience and, as you will see, making some judgment calls based on your
understanding.
Sparring Partner
You come up with a plan and use the engine as a blundercheck to see if the
move loses material. This is a good way to check out your ideas because
engines will put up tough resistance to your moves. This method is also
effective because sometimes we get so carried away with our ideas that we can
overlook the best moves but the engine will not.

Another excellent way of using this interactive analysis is, once you have a line
you want to analyze, go to the end and keep going back one move and wait to
see if the engine‟s analysis makes a large change. In fact CCGM Robin Smith
advises that this is absolutely vital. If it does then you will want to check out
those side lines to make sure the engine did not miss anything in its original
analysis.

Note: if the engine pauses for a long time and/or it keeps jumping back and
forth between candidate moves, it means it is having a hard time deciding on
the move it thinks best. For you, this means it‟s a good place to stop and
investigate further the move the engine was considering

Multi-variations
In this mode you click on the + or – to increase/decrease the number of moves
variations displayed. When the difference between the first and second choice
is about one Pawn and the position is relatively quiet tactically, then the
conclusion is that the first choice is probably best. However, in a tactically wild
position before concluding that the first choice is clearly best, you need a
difference of 2 or 3 Pawns before concluding that the first choice is definitely
the best one. If the difference is about a half Pawn, you can‟t draw any definite
conclusions.

Multiple Engines
Running more than one engine at the same time is one of the best ways to
analyze a position because different engines will not always come up with the
same first choice.

When you get different suggestions, you will have to investigate them further
and make a decision as to which move offers the best possibilities. Another
point to understand is that when it comes to positional evaluations, engines
can vary greatly in the numbers they display. Usually the difference will only be
a half-P or less which is insignificant, but it can also be quite large. When this
happens we usually either go with the engine showing the largest difference or
the one that‟s rated the strongest, but this can be misleading in positions
where there are material imbalances but one side has attacking possibilities.
So the question then becomes one of which engine are you going to believe?
Unfortunately, there is no easy answer to this question and you will either have
to try running a shootout, engine tournament from the position or rely on your
own evaluation of the position.
Engine Tournaments
Running a tournament with multiple engines from a given position is another
way of testing a position. After the tournament you can look at the games to get
an idea of different liens of play and see where they lead. One word of caution
though…the results can be skewed by later engine errors, so don‟t take the
results at face value without checking them out first. The most effective
positions to evaluate in this manner are those in which the next few moves will
affect the game in some way. Examples: which side to castle on, P-moves that
close the center or result in isolated or doubled P‟s, exchanges leading to
material imbalances, or positions that are complicated and open.

Deep Position Analysis

This method of analysis is one that establishes an analysis tree and you can
adjust the length of the branches. This method is probably the least desirable
method and is best used when you don‟t want to spend time interacting with
the engine and are content to just let the engine do all the work. If one’s
understanding of chess isn’t very good this method is satisfactory.

Full Analysis and Blunderchecking

This is good if you want to check a game for tactical errors, but remember that
this analysis will be pretty meaningless in closed positions and setting
blunderchecking at any value of a half-P or less will result in “improvements”
that are, generally, meaningless

Do not rely on the engine’s evaluation of a position after a long

analysis!! After you have completed the analysis, whatever the time allowed,
go to the end of the analysis and see what the evaluation is then begin stepping
backward through the analysis. It is not unusual to see a significant change in
evaluations and better moves recommended.

You can easily get lead down a false path by following moves that the engine
says lead to an OK position, but eventually you see its evaluation jump from,
say, your having a relatively meaningless half P deficit to it showing your
position is so bad you can resign.

Opening Advice for Correspondence Players

Gary Kasparov proposed what he called Advanced Chess back in1998 and, like
it or not, this is the way CC is played at the master level these days. Even if
you are playing at lower levels where engine use is not happening, you can still
use these hints to get the most out of your engine analysis and increase your
understanding

If one's goal is to have fun or practice for OTB tournaments you may take a
different approach, but if your goal is to win more games and improve results
then it will be necessary to thoroughly research openings. You don‟t necessarily
have to do this before the game starts. You can wait until you know what
options are available, but one thing you should avoid is just blindly wondering
down an opening line only to find out too late that it has resulted in an inferior
position. Know where you are headed before you start down the path

In CC play at the master level it is necessary to know your openings and select
them carefully. Not playing at the master level? Imitation is a good way to
learn. Plus if you are playing at a lower level, opening study should produce
even greater dividends. By opening study, I am NOT talking about memorizing
reams of variations, but rather understanding the ideas behind them and
resulting P-structures and tactics, etc. Of course one needs to know how to
play the middle-game and endings, but all things being equal, an opening
advantage gives you a head start in the effort to win the game. Besides, in this
day of engine use in CC play, starting at even a slight disadvantage can be
disastrous! This is far more true in CC play than in OTB play…unless you are a
GM, of course.

One point…detailed openings study is NOT the best way to make yourself a
better player, but it is necessary in CC play. Deep theoretical study is not
essential to play at a high level, but you DO need a really good database of high
level games. You cannot rely on the opening book that comes with Fritz or
other engines. I recommend putting together a database of correspondence
game played by masters because then you know they are well-played and the
openings have most likely been engine checked.

When you select your opening line, look for wins by the highest rated players
who played that line. That puts the onus of finding an improvement on your
opponent.

If you do not prepare much before the game starts you will have to do a lot of
work during the opening because you will need to review all the games you can
find, survey opening theory, and you may have to try and find new ideas
yourself.

During the game note keeping is also essential. One thing you don‟t want to do
is spend time rediscovering moves. Another thing I strongly advise against if
you are playing on a site that allows using engines is that you DO NOT save
the analysis then blindly follow it to the end! Once you have selected your
move, delete the subsequent moves and if you want, type in the analysis as a
note. Failing to do so can result in some mistakes of a type I won‟t elaborate on
but take my advice…don‟t do it.

Also, when entering your selected move on the server, ALWAYS double check to
make sure the move you are making is actually the position on the board and
not from a position you are analyzing! I have both won and lost games because
I was analyzing with a chess program, decided on my move then went to the
actual position on the server and entered the move only to find out too late that
the move was from a position that I was analyzing and not the actual position.
That happened because I didn‟t look and verify I had the position right. It‟s
stupid, I know, but I‟ve done it, my opponents have done it and you will do it if
you aren‟t careful. I once beat an opponent who outrated me by 200 points
when I played e5 attacking his N on f6 and he didn‟t play …Nd7. Instead he
played the move he would have played on the NEXT move…obviously he didn‟t
look at the position on the server.

Analyzing Won and Lost Games

Sound advice from veteran CC players is that you should always devote more
time to your won positions than your lost ones. A mistake in a lost position
won‟t be a disaster but if you blunder in a won position, it will be very painful!

At the same time you should be aware of the difference between bad positions
and lost ones. Be careful not to write off bad positions prematurely and fail to
give them extra attention. You should devote special attention at critical points
in any position, even in lost positions. In bad positions you should try to set
hard problems for your opponent. When you are really sure that the game is
lost then resign!

Don't be afraid to accept a draw in an equal position! Your thinking time is

probably better spent on winning good positions and saving bad ones than on
trying to find ways to win an equal one. One point worth making here is that
you need to distinguish between truly drawn positions and positions that offer
equal chances.

Endgame technique is very important, so don't underestimate tablebases. If

used properly they can be very helpful. They will help you decided whether or
not to steer for an ending.
Remember, even with engines, understanding their peculiarities is important
because you can‟t rely solely on their output. Human interaction, things like
playing through their analysis, engine vs. engine matches from the given
position, etc. are all necessary before selecting a move are vital, if time
consuming techniques, that will bring better results than simply letting the
engine think for a couple of minutes then playing whatever it suggests. It‟s also
more challenging and fun which is why we play CC anyway!

In many cases even when using an engine it all comes down to who
understands the position better, you or your opponent.

As one veteran CC player so poetically put it, all engines are useful when it
comes to whispering cunning moves in your ear but it is completely up to you,
whether you let your engine automatically play without your "help" or if playing
centaur style, trying to create a plan and checking the moves with your chess
engine. Right or wrong - it was my plan, you can say after your loss.

He offered another hint: Rybka, for example, is sometimes very generous about
sacrificing pawns and often it is correct with its evaluation BUT it also still
suffers from a serious lack of endgame knowledge in some types of endgames,
but not all. You have to understand how engines evaluate

Pre-existing Analysis

These days most all books, including opening books, contain computer
analysis but playing engine generated moves from those sources is legal. The
fly in the ointment for some people is the term "pre-existing analysis" as used
by some online chess sites. What if I‟ve been analyzing an opening line with the
aid of my chess engines and either printed it out or stored the analysis in my
db? Is that analysis illegal if I refer to it during the game?

It is well known that engine analysis is making an impact on opening theory

and some opening database sites will, no doubt, contain much analysis that is
engine generated. How can you be sure what was engine generated and what
wasn‟t? It is legal to use Nunn’s Chess Openings, but right in the preface it is
stated that some analysis engine generated. Must I avoid those lines? If it‟s
legal to consult NCO and play all the lines contained therein, why can‟t I use
my own pre-existing engine analysis? What‟s the difference?

Another problem exists with the phrase "databases consisting of previously

played games between human players." I suppose the addition of such a phrase
eliminates a few opening sites but my own personal analysis is not going to
consist entirely engine generated moves because, as I‟ve previously pointed out,
engines require some input from humans to be used effectively in high level CC
play. Is analysis produced in this way illegal? Serious CC players research
opening lines quite deeply in this fashion and therefore many players insist
such a person is not using their own skills and it is questionable ethically to
use such material.

I think such people are straining gnats on this issue. In my opinion it doesn’t
matter if you use books, databases or previously engine generated material.
Sooner or later a move will be played that has not been previously analyzed
whether it’s at move 3 or move 30. At that point you are on your own and the
outcome will be decided be chess skill alone. So to my mind the whole issue is
moot.

Making Your Own Database and Opening Book

These days with the use of engines in CC DB‟s are very important because
when it comes to opening selection, gone are the days of playing risky
openings. At high level CC you see mostly solid, mainline stuff: Ruy Lopez,
Sicilian, Nimzo-Indian, Q-Indian, etc. In the past I used my own personally
constructed DB consisting of 3.5 million games, NOT the Fritz opening book, to
select opening lines.

A better and more productive idea turned out to be making a database

consisting of top level correspondence games. I can search opening lines that
have been played by strong CC players. This has an advantage of telling me
what lines have been successful plus I know that they have in all likelihood
been engine checked. Here are couple of results from my own CC games:
After the following moves where I am playing Black: 1.d4 d5 2.Nf3 Nf6 3.e3 e6
4.Bd3 c5 5.b3 Nc6 6.Bb2 Bd6 7.0–0 0–0

Fritz‟ opening book shows a result of +4 -5 =2 and my database of CC games

with at least one player rated over 2200 results: +9 -8 =22, so the chances are
about equal and a draw wouldn‟t be unexpected.

On the other hand in the following game where I am also playing Black, after
the moves 1.e4 c5 2.Nf3 d6 3.d4 cxd4 4.Nxd4 Nf6 5.Nc3 Nc6 6.Bg5 e6
7.Qd2 Be7 8.0–0–0 0–0

The Fritz book shows +25 -11 =15 and my new database with both players
rated over 2400 results are not good either: +20 -7 =16, so it looks like I‟m
going to have an uphill battle in this one!
A final example where I am White: 1.e4 e6 2.d4 d5 3.Nc3 Bb4 4.e5 c5 5.a3
Bxc3+ 6.bxc3 Ne7 7.Qg4 cxd4 8.Qxg7 Rg8 9.Qxh7 Qc7

After 9.Qxh7 Fritz‟ opening book shows only two moves for Black 9…Nbc6 and
9…Qa5 with results of +2 -0 =1; not much help. My DB gives a different picture
because after 9…Qc7, which isn‟t given in the Fritz book, a search of the
diagrammed position where both players were rated over 2400 shows +11 -1
=11, so that‟s very encouraging.

Critical Points

According to GM Arthur Yusupov, “To many people it seems that grandmasters

simply calculate variations a little deeper. Or they know their opening theory
slightly better, But in fact the real difference is something else. You can pick
out two essential qualities in which those with higher titles are superior to
others; the ability to sense the critical moment in a game, and a finer
understanding of various positional problems.

The ability to identify critical moments in a game is important because it is

those positions that will influence the further course of the game. In fact GM
Edmar Mednis, writing his great book titled How to Beat Bobby Fischer in
which he analyzed Fischer‟s defeats, always showed the position where he
isolated the losing move; the move where Fischer missed the critical point.
Also, sometimes in GM analysis you will read comments such as, “Not a bad
move, but it embarks upon the wrong plan.” Or some similar comment. These
are critical moments.
Every strong player has the intuition that tells him when he has arrived at a
critical position, but as with any GM‟s intuition, it is often difficult to define
exactly when such positions are reached. I think sometimes GM‟s cannot
explain such things…they just know.

You would think working with an engine would make it easy for us to know
when a critical position has arisen on the board, but that‟s not the case
because engines, critical position or not, treat every position the same and
calculate variations.

GM Jonathan Rowson spent an entire chapter in his book, The Seven Deadly
Chess Sins, discussing how to identify critical positions or key moments.
Rowson wrote, "To miss such moments can be considered „sinful‟ in that it
usually results from a basic misunderstanding of the nature of chess
assessments and of how they can and do change.” He then went on to explain,
“…it is very difficult, if not impossible, to give any clear definition of what a key
moment, or critical position actually is…” Still, Rowson went on to try and
define what the circumstances are and how, when we see them, to know we
have reached such a position. His list:

1-You begin to see pending counterplay for your opponent

2-The prevailing trend seems to have stopped and you can‟t see any way for the
advantage (for either side) to be increased
3-You have lots of reasonable moves, but none seems to be outstanding
4-You opponent‟s last move was unexpected or in some way unusual.