“Blundercheck” analysis mode in Fritz/Rybka

Last time around we learned how to analyze a chess game using the “Full analysis” mode in Fritz and its related playing programs (Rybka, Hiarcs, Junior, & Shredder). “Full analysis” is a decent tool for chessplayers (especially beginners) who don’t want to be overwhelmed by long, and sometimes complex, variations and who would rather have verbal cues and symbolic notation instead of numeric evaluations. But for players who are a bit more advanced and aren’t afraid of numbers, there’s another analysis mode available. “Blundercheck” provides precise numerical evaluations – you’ll not only see that the chess engine’s suggested variation is better than what was actually played, you’ll see exactly how much better.

You’ll recall from that previous blog post how to open a database in Fritz and highlight a game you want the software to analyze. This time around, though, you should click the “Blundercheck” button in the “Database” menu’s ribbon:

Fritz12 and Rybka4 chess playing computer software from USCFSales.com

After you click it, you’ll see the following dialogue:

Fritz12 and Rybka4 chess playing computer software from USCFSales.com

It would be pointless for me to repeat the descriptions of the “Side to analyze”, “Threshold”, and “Storage” settings – these are all described in the immediately previous post to this blog. However, this dialogue does contain some new settings we’ve not seen before.

First of all, I should point out that the “Time” and “Depth” settings are mutually exclusive: you can set one or the other, but not both. As always, I recommend the use of “Time” instead of “Depth”; this allows Fritz to budget its own time and, in lively positions with forced tactics, you will likely get a deeper search by using “Time” (the tradeoff being that you might get a shallower search in closed positions).

However, if you’re going to use “Depth” please set it for an odd number, as some chess engines become tactically “blind” at even numbered search depths, failing to consider proper replies to a tactical shot which happens at the end of its search.

Output” determines whether the engine’s output will appear as a text annotation (and by text, I still mean algebraic notation move sequences, not text explanations of moves and variations) or as replayable variations (in which you can replay the suggested variation directly on the screen). I recommend “Annotate as variations” so that you can play them out on the screen as you’re reviewing the game.

As in “Full analysis” mode, the “Erase old variations” command wipes out any existing commentary in a game (such as a GM-commented game from Mega Database), so please use this toggle with caution.

Selecting “Write full variations” means that the engine will include the entire recommended variation it finds, while unchecking it will cause the software to truncate the variations, shortening them to just a few moves (as we’ve previously seen in the output of “Full analysis”).

A check in the box beside “Training” means that the software will generate an occasional timed training question if a major tactical blunder occurred in the game being analyzed (this, too, was discussed in the immediately preceding blog post).

Store evaluation” means that the chess engine will not only provide a positional evaluation (in numeric form) at the end of each recommended variation, but will also provide one for each move of the actual game. This is useful for a large number of reasons, not the least of which is the generation of an Evaluation Profile graph for the game (as we’ll see shortly):

Fritz12 and Rybka4 chess playing computer software from USCFSales.com

Check main line” means that the chess engine will evaluate the moves of the actual game. “Check variations” means that the chess engine will perform its analysis only on existing variations in the gamescore. This latter function is, in fact, the original intended purpose of the “Blundercheck” feature. Say, for example, you’re annotating a game for publication and have included several variations in your notes. If you select only “Check variations”, the chess engine would ignore the move actually played in the game and analyze just the variations you added. This is a really handy way for chess writers to double-check their own evaluations prior to publication.

The combination of these two features provides you with a lot of flexibility for having a chessplaying program analyze a game. You can start with an annotated game (one you’ve annotated yourself, for example, as a means of practicing game analysis) and have your chess engine analyze only the main moves of the game, only your added notes, or analyze all of it (but do remember to turn off “Erase old variations” first!).

After you’ve made all of your selections in this dialogue, click “OK” to start the process. Just as with “Full analysis”, the chess engine will start at the last move and work its way backwards through the game. When the analysis is finished, you’ll see something like this:

Fritz12 and Rybka4 chess playing computer software from USCFSales.com

Click on the picture for a larger view

Each move of the game has a numerical evaluation in blue, expressed as two numbers. The first tells you who is ahead in that position and by how much (expressed in pawn units). For example, after 6.Qxd4 White is ahead by 1.25 pawns; even though the material is even, White has developed two pieces to Black’s none, and Fritz thinks this is the positional equivalent of being ahead by a pawn and a quarter. Please note here that a value expressed as a negative number means that Black is ahead in the position.

The second number (after the slash) is the search depth: how far ahead the software looked to make that evaluation. For example, after 6.Qxd4 Fritz thinks White is ahead by 1.25 pawns and searched 17 plies (8.5 full moves) ahead.

We can also see how the “Threshold” value works. When I had Fritz analyze this game, I specified a “Threshold” value of 30. Fritz has provided a better move (in reality, a complete variation) at every point in the game where it discovered a line of play which is at least 0.30 pawns better than what was actually played. For example:

Fritz12 and Rybka4 chess playing computer software from USCFSales.com

Click on the picture for a larger view

Here Fritz has found an improvement for White’s seventh move. In the actual game, White played 7.Qe5+, which results in a position which Fritz evaluates as being 1.25 pawns better for White. But had White played 7.Qd3 instead, and had the remaining moves of the variation through 16…Rd8 been played (with each move of this variation being the best move for either player in Fritz’s estimation), the result would have been a position in which White would have been 2.05 pawns ahead – an improvement of 8/10ths of a pawn (which, by the way, rates as a significant improvement).

All of those numbers in blue in the game’s main line, by the way, were generated because I had checked “Store evaluation”. If you go to the “View” menu and select “Evaluation profile”:

Fritz12 and Rybka4 chess playing computer software from USCFSales.com

Click on the picture for a larger view

…you’ll see a new pane appear in Fritz’s main chessboard screen, a pane which contains the game’s evaluation profile:

Fritz12 and Rybka4 chess playing computer software from USCFSales.com

Click on the picture for a larger view

Guess what we’ll be discussing on Friday? Until then…

Have fun! – Steve Lopez

Chessplayers who have purchased their ChessBase brand chess computer software from USCFSales can receive free technical support and advice on their purchases straight from me; just shoot me an e-mail (steve@uscfsales.com), but please remember to include the USCFSales order number from your ChessBase software purchase. – Steve

Advertisements

3 Comments

Filed under chess, Chess DVD, Chess playing software, Chess software, ChessBase, Fritz, Fritz Powerbook, Hiarcs, Junior, Rybka, Shredder

3 responses to ““Blundercheck” analysis mode in Fritz/Rybka

  1. Pingback: Post-analysis evaluation profiles in Fritz/Rybka | USCFSales

  2. Radovan

    In what way can it be looking for a great move? This option is for the wrong moves, but nowhere does it show that it played a great move.

    • I have no idea what you’re asking here. You set a value for “Threshold”; if the engine finds a line of play which results in an improvement higher than the Threshold value, it displays that line of play. It also provides a numerical evaluation of that line of play, along with one for move actually played so that you can compare the values.

      If you set a low Threshold of, say, “25”, any move which doesn’t get a computer-generated variation was evaluated as being within 0.25 pawns of the best line of play — in other words, close enough to the “threshold” that any improvement isn’t much of one. So any move in which Fritz doesn’t show an improvement can be considered a good move.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s