Analysis is a highly non-linear activity, with many tangents. Effective utilization of Chess Assistant requires a willingness to conduct analysis without using the chess engine as a crutch.
Use the computer only for checking your analysis, or as a final step, to search for improvements that you can't find on your own.
In this article, I will provide an example of how I conduct game analysis. It is not necessarily the correct method, since there are a variety of analysis modes that Chess Assistant provides, and everyone will conduct the analysis process a little differently.
It may seem like a very basic fact, but you will get the most out of analyzing your losses, and not your wins. When analyzing wins, it is especially easy to gloss over sloppy play and tell yourself that an inferior move was ok, because you still won the game. Not so with a loss, where it is easier to assign blame.
The following is a game I played against my computer some time ago. The winboard engine I was playing against was called Wildcat. The game ended in an ignominious defeat for yours truly, and as such, provides good fodder for this article. In the course of this article, I will touch on topics such as how to use CA to analyze the opening of the game, and when to use blunder checking, background analysis, and interactive analysis mode.
There are a couple of schools of thought regarding the use of chess engines for game analysis. I am of the opinion that game analysis needs to be done in two phases (time permitting). In the first phase, you need to analyze the game yourself. In the second phase, you use the chess engine for assistance.
Phase Ia, opening analysis
You can download this game in ZIP format. The file can be opened directly from within later versions of CA. The database is in version 5 format.
The first thing we want to do is take a look at the opening phase to find improvements. So, start at the first move in the game, and step through each move. While doing this, take a look at the tree statistics, master evaluations, and CAP data. What you are looking for is moves that either a) score well, or b)have good master or CAP evaluations. Using this approach, we see that after black's move 2...c6, the tree display looks something like the graphic to the right.
What we see is that the move 3.d4 both scores better than the move played in the game (the automatic 3.Bg2), and it's master evaluation (which appears just to the right of the first three moves) is better than that for Bg2 also. So it is a safe assumption that we've found an improvement here. Plus, when you think about it, white can get this thrust in, and recapture on d4 with the queen, without worrying that it will get kicked by a knight developing to c6 (since that square is blocked by a pawn).
Ok, we've found the first improvement, so we can then move through the tree, and look at probable game continuations from that point on. However, I think I will save that technique for discussion in another article.
You can also use one of the automatic opening commenting features that CA has by using the "annotate opening" button on the Comments button bar. You will see that it automatically finds improvements, shows you where the novelty occurred, and will insert model games, without any intervention on your part! The method you chose is up to you, but I use the manual method mostly out of habit (there is nothing wrong with the automatic method). If you don't use this function then you need to put in references to where the novelty occurred (plus model games, if you so desire).
Ok, let's move a little further down the tree. Coincidentally, 4.d4 looks like the main line, since it occurs in most games. When we click in d4 in the tree display, we see that black's usual response is 4...e4. Well, isn't that greedy of him - grabbing space like that, doesn't black know his place? Well, the question after black's e4 move is how white finishes his development. Does he play e3, so as to develop the knight on e2? Well, if we look at the tree again, we see that white usually plays Nh3 at some point. If we continue to move around in the tree, looking at alternative moves, we see that Nh3, f3 and ...f5 by black all seem to be common moves. What we've found is the beginning of a plan for white.
Phase Ib, finding a plan
Let's examine this idea a bit further. Click on the small carat to the right of the search button on the main button bar. Select the menu entry called "Search for current pawn structure". The search dialog box should open up, with the pawns in the proper formation. Let's modify this position slightly to add the knight at h3, the bishop at g2, and a white pawn at f3. You should have something like this:
After setting up the position in the dialog box, click on the button in the lower right hand corner with the little red circle next to it. A dialog box will open, allowing you to select the database you want to search through. Make sure the checkbox next to Hugebase is checked, hit "Ok", then hit the "Search" button on the search dialog. You should get a dataset with a little over 40 games in it.
If you examine the dataset, you should see some common ideas cropping up. One such idea is that the knight on h3 typically goes to f4, to hit the black pawn on d5. And many times, the queen comes to b3 to hit this pawn also. It certainly looks like a plan, or at least a set of ideas to use in future games.
If you want, you can keep a separate database with a collection of these opening ideas in it. If you don't want to do this, I would suggest that you at least assign a class marker to any position that you find to be interesting. Classes and markers will be the topic of a future article.
Ok, we now have a better way of playing this opening, and we have some ideas for how white's play should proceed from there. Now on to some analysis of the middlegame.
Phase Ic, middlegame and endgame analysis
In this part of the analysis process, we step through each move in the game, and attempt to find improvements, without engine assistance. This is not easy, and can be very time consuming to do. Additionally, it is quite difficult to find improvements that you missed during play. However, I believe that this stage is necessary, since it helps one verbalize about what went on in the game. Also, there are many times when you may sense that you missed something, and you can use this intuition with some of the engine analysis tools to find alternative moves (more on that later).
I'll be honest, when I went through this game, I did not find many improvements, except one for the computer. Of course, I felt like I had definitely messed up with 46 e5, but other than that, I'd only seen a the possibility of playing 16. Nb1. As such, when I was annotating the game, I submitted the position after 15...d4 to see what the computer thought of this position, and where it would have put the knight.
You can find the results of my phase I analysis in the ZIP file here, it is the first game in the database. I did not include the background analysis for 16.Ne2, since it essentially confirmed the game continuation.
Phase II, computer engine analysis
One thing I wanted to do was look at the placement of the knight. When I originally played Ne2, I also had thoughts of eventually placing the knight on d5. But in any case, I decided to take a look at the position after 15...d4 with the multivariation mode (multivariation mode can be accessed by starting infinite analysis mode, and then clicking on the "Multivariation mode" button on the engines tab, see screen shot to upper right) . When I did this, it was clear that Tiger felt Nb1 was a lemon, since the difference in the move evaluations was about 0.36. But, I then looked a little further into the resulting variations, and found that there really wasn't much difference in Tiger's evaluation of the position after the knight was maneuvered onto c4. My conclusion after looking at all this, was that 16.Nb1 was probably ok, but Ne2 was more active. Tiger probably did not like the looks of it, since it seemed to "undevelop" a piece.
Ok, after I satisfied my curiosity about the placement of the knight, I decided to run a blunder check on the game. To access this feature, select "Search for blunders" from the "Engines" menu. Once the dialog box pops up for this feature, select the options you want, and click the "Ok" button. You can see the settings that I like to use in the figure below. I usually like to increase the ply depth to 7/9, and use 3 passes of checking with a threshold of two evaluation symbols.
Once the blunder check was completed, I got a game containing a mass of variations. I then went through and looked at all the suggested improvements. Some I chose to ignore, but there were a few suggestions that I thought were extremely important. One of the problems with this mode of analysis is that it can provide too much information. Obviously, if you increase the threshold, you will get less analysis to work through, but you might miss something as well. If you want, you can conduct your own blunder analysis of this game on your computer, with the settings shown in the figure to your right.
While I will not go through every move suggested by the blunder checking step, I did feel as though the following two were important:
You will notice that the computer did not say "Hey Bob, you can get to a bad B vs. good N endgame by playing 29. Bf4!". It only provided a variation, with a favorable evaluation for this move. I had to supply the words myself when I looked at the variation. This is the dirty little secret of computer analysis - in some cases, you need to make sense of the variations that the computer spits out. Clearly, this is not an ability that everyone has, but you need to attempt an explanation for important moves like this, when you feel comfortable doing so.
Incidentally, please take my analysis with a grain of salt. The last time I checked, my answering machine did not have an invitation for a skittles game with Gary K. on it :-).
Ok, one problem still remained. I know that I screwed up in the endgame, but I was not entirely happy with the computer's suggestion of 46. Rc2. It seemed like a passive move to me, and I thought there had to be something better. So, I decided to go back and employ one of CA's analysis tools that was introduced with version 6.
I am speaking of the "Interactive analysis function". It is feature that I am sure many CA owners are totally unaware of.. It is an extremely sophisticated analysis mode that can be used to smell out improvements, and take a really deep look at a position. You start this mode via a button on the "Engines" button bar. Interactive analysis mode essentially plays out variations in the game, and then retraces the analysis to check its results. Typically, this results in a new variation which needs to be examined. This mode is very effective, since it effectively employs hash table that it has created from processing moves further down in the move tree. But first, let's refresh our memories about the position after move 44 (you'll see why in a minute).
I decided to start the analysis a couple of moves back from move 46 (I started it after black's move 43). The result was that the computer suggested a line starting with 44. Rb3 Ra8 45. e5!. Magnifico. When I looked at this variation, I saw certain elements in common with my calculations in the game when I played ...e5, but I also saw that by playing 45.Rb6, I wasted a valuable tempo. Thus, I added a third mistake to my list - I should have played ...e5 one move earlier. Clearly, this is a move I should have been looking at, with the idea of creating a passer on f5.
Note that this move would probably have been hard to find if my intuition hadn't told me that it was there.
I realize it has probably taken a while to go through this particular article. I can assure you that it took even longer to write :-). Game analysis is not necessarily a linear process, since there are many tangents, and what I like to call "Voyages of exploration" to go on. Although, these journeys can be one of the more rewarding aspects of the analysis process.
There are several things that you need to keep in mind when performing this process. First - you need to use your brain whenever possible. While time does not always permit a thorough human analysis of a game, you should try to do as much as you can. That means putting in your own analysis, without the engine running. If you simply look at the engine churning out variations, then you aren't learning very much. The utility of the engine is to check your work, and to point out things that you missed, that's all. The final thing is that you need to conduct a dialog while you analyze, and attempt to put words to the variations that the computer is supplying to you. This is not always easy, and it presupposes a certain amount of knowledge on the part of the user.