Translate this page
Filters and Masks
Quote of the Month: I'm only interested in what's important.
We all have filters attached to our senses to restrict the type of input we want to process. For example, when driving a car, you do not try to take in all the scenery. Instead, you focus on items that might affect the safety (or direction) of your car. If something in your peripheral vision, like a ball rolling onto the street, becomes a safety issue, then it becomes important; otherwise most "side" issues are easily missed. The idea that the brain "tunes out" less important input is necessary for our sanity, and we can think of this as a filter – the brain allows through to our consciousness only the most high priority items, so we do not have to process everything.
To make a software analogy, a filter is like an "IF" statement that calls a subroutine. It states something like "If X is true, then call subroutine Y" or "If X is less than amount Z, then call subroutine Y." The parameter X is acting as a filter as to whether you would process Y; while Z is acting as a threshold for the filter.
In chess we develop filters too. For example, when I ask players of all levels to list all the checks in a position, often lower-ranked players do not quickly report all the checks that are not safe. They "filter out" those moves as not really possible, even though I am simply asking about the legality of a check, not the "goodness" of it. Interestingly, stronger players usually can quickly name all the checks, no matter how terrible the move might be – they are better able to "turn off" their filter and just see a move as a check or not.
This ability to control a filter turns out to be important in many board vision and tactical vision issues. Many inexperienced players are taught to "filter out" certain moves to help quickly determine safety, but in doing so they sometimes go too far and end up missing ideas that could be important. I first discussed this issue at length in Sneaky Pins and Invisible Moves. It is a filter that often causes these potentially good moves to become invisible. These types of difficult-to-spot moves are the subject of the fairly advanced book Invisible Chess Moves by Neiman and Afek.
Many players complain that when doing a tactics puzzles they can easily find a pseudo (sham) sacrifice that eventually leads to win of material, but in a game they miss these sacrifices. Part of this failure is due to not recognizing The Seeds of Tactical Destruction, but not all misses can be attributed to this lack of pattern recognition. I believe what happens during the game is that a weaker player's filter threshold is set "higher" than it is when they are doing the puzzles, since the solver knows there is a solution and finding that solution may involve a possible sacrifice. During a game these less experienced players do not want to waste time looking at frivolous sacrifices, so the filter threshold is set back "on" and they miss tactics that stronger players find, because stronger players both recognize the possibilities (pattern recognition) and also have better control over their chess filters.
Similarly, many less experienced players filter out "bad" moves when asked to find threats. A threat is a move, which if unstopped by the opponent can do something clearly positive next move; e.g., win material, checkmate, ruin a pawn structure, infiltrate, etc. One way to help find threats is, after visualizing a potential threat, to then apply the null move, which assumes the opponent passes and does not move at all, and then see if something positive can be accomplished. As an example of threats, I sometimes use the following position from Bain's Chess Tactics for Students (colors reversed); I ask my students to find all the moves that are threats, but not also captures or checks:
White to play and win
Some find 1.Be5 or 1.Ba5, with the threat of 2.Bxc7+ (removal of the guard/deflection) 2...Kxc7 3.Rxe7+ winning a pawn and more. But it is more difficult for weaker players to spot the "unsafe" threats 1.Ba6 and 1.Bc6 – both threatening 2.Bxb7, and 1.Bb4 threatening 2.Rxe7 and 2.Bxe7+. Although these are not "good" moves they are perfectly good threats even if easily met. Clearly this is a case where a player's "safe" filter makes these threats more difficult to find.
By the way, the answer to the "Play and win" aspect of this problem is 1.Rxe7 Kxe7 2.Bb4+ and White nets a piece. This pseudo-sacrifice is just the kind of move that inexperienced players find easily when told it is there ("This book is too easy for me!"), but often miss in games.
Strong players can more easily recognize when a possible sacrifice is worth investigating and thus can play with their threshold set lower. In other words, strong players will often consider superficially unsafe moves if they think it has a decent chance of reaping rewards, while lesser players reject these moves out-of-hand even if the potential reward should act to advise them to "lower the filter!" For example, if a move seems superficially unsafe, but it could potentially win the game, one should probably lower the threshold for the filter and take some time to consider "If I could just make this work, I can win the game" and then analyze possible ways to make this "high risk, high reward" idea work. In the previous position this ability allows the stronger players to avoid quiescence errors and find eventually-safe ideas like 1.Rxe7 both more quickly and more often (in game situations).
I have an anecdotal example of a less experienced player filtering out a relatively easy sacrifice. About ten years ago I played Dr. Steven Rolfe (who recommended I write the book Looking for Trouble). Dr. Rolfe, a 1500 player, placed a knight on e4 (1.Ne4) where I could capture it with my rook, but the knight was defended by one of his rooks. Therefore, the immediate effect of capturing 1...Rxe4 was that I would lose the exchange.
However, it was obvious to me that after that sequence 1...Rxe4 2.Rxe4 that I could then use a bishop to pin Steven's rook to his king. Furthermore, getting out of the pin would be problematic for my opponent. I then calculated that I could, indeed, later win an entire rook, thus netting a piece. So, after a thorough recheck, I played the pseudo-sacrifice 1...Rxe4.
Dr. Rolfe was surprised that I made the initial capture – his filter had caused him to erroneously conclude not to be concerned about 1...Rxe4, since he could win the exchange with 2.Rxe4. A few moves later, after I had netted the piece, he resigned. Afterward, we were reviewing the game and I asked him why he allowed this relatively easy combination. His insightful reply was "I didn't think you would take it – it was guarded by my rook." Yet if you had given my position to Steven as a problem in a tactics book "Black to play and win" he almost certainly would have found the combination – it was the only chance in the position to win material.
By Dr. Rolfe's remarks, we can clearly see that he had set the threshold of his filter for consideration of my sacrifice too high. He simply assumed the knight was safe just because it was attacked by a rook and guarded by a rook. My threshold for continuing my analysis of the sequence was lower, as I asked "What would happen after 1...Rxe4 2.Rxe4?" I did not filter out calculating the rest of the possibilities just because the combination began by losing (i.e., sacrificing) the exchange. This type of error is quite common among lower rated players, as discussed in Quiescence Errors. So we can rightly attribute many quiescence errors to setting a filter threshold too high, but this is not the only type of error that is caused by a poorly set filter threshold.
One of my standard lesson problems is to address the safety of the candidate move 1.Nc3 in the following position:
When discussing how to figure out if 1.Nc3 is safe, I ask the following:
"After 1.Nc3, list all of the white pieces which are not guarded."
This is in accordance with Nunn's Dictum: "Loose Pieces Drop Off" (LPDO), which signifies the most common Seed of Tactical Destruction. Almost all students quickly find that the pawns on c2 and d4 are not guarded, but some players have difficulty seeing that the rook on a1 is also not guarded. It is so "sheltered" that the fact that it is not guarded seems to elude some players.
However, a bigger "filter" issue is that White has a fourth piece that is not guarded – the king! Some stronger players reply something like "Well, the king is not guarded but that's not what you mean, right?" And they are correct that whether a king is guarded or not is never a safety issue, so it seems that it might be excluded. But if you completely turn off the filter "how important is it for a king to be guarded?" and just ask "What are all the unguarded white pieces after 1.Nc3?," it becomes clear that the king should be included. Yet, whereas stronger players usually see that turning off the filter is possibly the proper course of action and thus question the inclusion of the king, the great majority of weaker players have trouble turning off this filter – even after I point it out! I have no doubt that the fact that these same players who have difficulty accepting the king as unguarded also make more quiescence errors is tied to the similar issue of those players having trouble adjusting their filters.
Once I agree with my students that an unguarded king – as opposed to an exposed king – is not a safety issue, I ask the follow-up question:
"After 1.Nc3, how many black replies attack at least two of the three other unguarded pieces: Ra1, Pc2, and/or Pd4?"
Inexperienced players have less trouble finding 1...Nb4 with a discovered attack by the rook against the pawn on d4 and a direct attack on c2, than they do identifying the "unsafe" 1...Ne3+ with the same discovered attack by the rook against d4 and direct attack on c2. At least this time there is less controversy – once they see it, almost everyone agrees these are the two moves that create such multiple threats. Once we have these two moves on our list, then we can reject 1...Ne3+ as easily met by 2.fxe3 or 2.Bxe3 and concentrate on the "safe" threat 1...Nb4. Doing so in turn answers the original question, as after 1.Nc3 Nb4 White cannot save c2, much less save both c2 and d4, and thus 1.Nc3 is not safe.
The point of presenting this problem in a lesson is not to point out filters (that is just a bonus for this column), but rather to show that often only careful analysis can determine if a move is safe. Relying on just pattern recognition (knowledge) would likely not work in this case, since this position is not a "common" safety pattern. You are very unlikely to find many similar patterns in tactical books or software problem sets. One reason similar problems are not in tactics books is that positions where one side can just attack a pawn and it cannot be guarded (such as c2 after 1.Nc3 Nb4) are not very "spectacular" and will not sell many books! In addition, this tactic cannot be clearly categorized into, say, a chapter on discovered attacks, since the discovered attack on d4 can be met and turns out to be gravy, and not the main theme.
There is an excellent example in Looking for Trouble where my own filters failed me. I was playing a practice game against a 1500 player and we reached the following position with me as white:
White to play after 12...c7-c5
Inexperienced players usually have no difficulties with their filter on whether castling is legal (although their threshold for wanting to castle is often set too high!). However, their ability to realize that en passant is possible is often filtered out, or possibly forgotten altogether. This is definitely not true for stronger players, so the first thing I realized was that 13.dxc6 e.p. was legal. Nevertheless, I do have a clear filter against anti-positional moves that help my opponent, and 13.dxc6 seemed to allow only things I did not want to happen after the obvious reply 13...dxc6:
So I promptly "filtered out" 13.dxc6 as positionally terrible, and, after some consideration, went on to play another move. After I returned home and started analyzing the game with a strong engine, imagine my surprise when it promptly spat out 13.dxc6 as the clearly best move! Stunned, I forced it to show me why and it displayed 13.dxc6! dxc6 14.h6! threatening mate. Black has no good defense: 14...gxh6 15.Bxh6 f6 16.Bxf8+- (among others), or 14...f6 15.hxg7 Rf7 (15...Kxg7 16.Bh6++-) 16.Qc4+- So this is a position where the tempo gained from making Black recapture on c6 – instead of retreating the queen – allows White the time to play 14.h6! In other words, all the positional negatives of the move are completely wiped out by the positive tempo. This is also a good example of The Principle of Tactical Dominance. Too bad my "strong" positional filter rejected the move right away!
Another area where "internal filters" cause problems is making errors of acquiescing. Acquiescing is different than quiescence; it is giving up too easily when faced with a critical threat. For example, your opponent's previous move seems to have your queen trapped, but you can throw away a pawn to save a queen, good players will spot this since they "lower" their filter to make a pawn loss acceptable, whereas normally it is not. But inexperienced players often reject such sacrificial moves as "unsafe" even though the consequences would be to allow a much more unsafe move, such as the loss of the queen. Even worse, they may also acquiesce and just quickly and unnecessarily give up the queen for a piece, leaving them with lots of time to play out a lost game. For more on giving up too easily, see Acquiescing.
As a combination of the previous two subjects (a master filtering out a good idea and someone acquiescing and giving up too soon), we have the extreme examples of strong masters resigning in positions where they are not losing. Of course, this is the result of not seeing a saving tactic, which sometimes results from a filter that is so strong that even imminent resignation did not lift it. The most famous example is Von Popiel – Marco, Monte Carlo 1902:
Black to play
Black is faced with the threat of losing his pinned dark-squared bishop on d4. Of course, he would love to move the bishop if he could check and win the white queen, but the white king is on a light square, so check is out. And Black would be willing to retreat his bishop to guard the rook, but the rook is also on a light square, so retreating the bishop just allows 2.Qxd7. Therefore, faced with these and other constraints, Black resigned. It was not until later that others pointed out there was a third way to safely move the bishop – threaten mate. This is possible with the filtered-out 1...Bg1!, threatening both 2...Rxd3 and 2...Qxh2. There is no good defense, so it is Black who should be winning, not White! This is a good example of an "invisible" move.
Masks are different than filters, although you could consider some masks as filters always locked completely "on," not letting anything through. A mask is something that prevents you from seeing something else. The most common type of mask I encounter while teaching is a student playing too fast. If you make bad moves and lose games because you play too fast, that masks the types of errors that would cause you to lose games if you tried your best and took your time.
One of my only requirements for taking on a student for lessons is that they be willing to take their time in long time control games. Why? Because my job is to help them play better, and to play better they have to find better moves (see Chess is Decisions). If they are not going to look for better moves, then it is very difficult to help them find those better moves! But there is a secondary reason: if someone does not try their best and/or plays too quickly, then it is not clear how much they know and how well they can play if they did. And if they lose because they made bad moves without thought, what would be their cause of losing games if they didn't? So playing fast masks those problems and the first thing I need to do when I get a student who plays too quickly is to convince them that if they want to play better, they have to try to take their time in slow games (see The Goldilocks Principle). Once they do that, they not only will play better, but playing too fast will not be their universal cause of loss, and we will be able to spot their underlying weaknesses and start working on them, too.
Interestingly, playing too much to your opponent's rating is a type of mask as well. For example, some players are so afraid of higher rated players that they win a much lower percentage of games against them than the rating system would predict. And others take lower rated players so lightly that they do not play anything near their best and lose a higher percentage of games than they should. In these cases the preconceived, perceived difficulty or ease of the game, based on the opponent's rating, masks the player's true capabilities.
Not all masks are bad. If you can "mask out" outside influences and concentrate fully on each move to find the best move in the given situation (time and position), that is a great trait. Worrying about an earlier mistake, how the outcome will affect your rating, or even where you are going to eat after the game, is only going to be counterproductive. For more positive abilities, check out Traits of a Good Chessplayer.
Question I have identified that my biggest need for improvement is time management. (I work my way up to 1300+ USCF and then I dip down to about 1200 and then repeat.) I do not get into time trouble in the opening, but once I am out of my "book," I take too long to decide if my candidate moves are safe. Because of this, I am usually in time trouble in the endgame and make poor moves and lose the game. I am overly cautious until it is too late. Are there some exercises that I can do that will help speed up my thinking process to determine safety? I spoke with one of the Experts at our local club and he thinks the safety check should only take a few seconds. (Maybe for him, but for me it is not that simple). Note: I have Looking For Trouble, but I find myself spending too much time on the puzzles. So I am guessing there is a better method for me.
Answer Repetitive study of easy tactics, such as Bain's book Chess Tactics for Students will help safety recognition (see A Different Approach to Studying Tactics and Tactical Sets and Goals). And a good thought process asking "Is it safe?" about each candidate move and having a sanity check, if necessary – will help. But many players, even after doing this, are "too cautious" and cannot pull the trigger anyway (see Speeding Up). Moreover, study of instructive game collections (many!) will help you recognize what to do and allow you to play somewhat faster. Lastly, time awareness is important. After each move, when your opponent is thinking, ask yourself "Am I playing too fast or too slow?" and adjust accordingly.
If you are 1300, then Looking for Trouble may not be the best book for what you need, though the one- and two-star problems should be very helpful. Check out The Improving Chess Thinker, The World's Most Instructive Amateur Game Book, and A Guide to Chess Improvement, which have many suggestions such as the above. Also see my multiply-indexed Novice Nook columns.
Question Recently I have been working on my endgames and I have a decent basic knowledge of K&P, K&R, opposition, and so on. Where I struggle is with the larger endgame, when there are a few minor pieces and several pawns, the king starts to join the action and plans and mistakes are difficult to assess. Any ideas how I might tackle improving this part of my game?
Answer Your bigger question includes most endgames and an almost infinite number of positions. As I always tell everyone, I got better at these endgames by:
1) Playing long time control games (such as 48/2, followed by 24/1 or SD/1) where I had lots of time to play the endgame and could think about each move carefully.
2) Reviewing the endgame with my opponent and/or a strong player to see what was possible. These days we can include strong computers and tablebases, too.
3) Learning how to analyze well in all positions and applying that same thinking to the endgame.
4) I did read some basic endgame books, such as Averbakh's Essential Chess Endgames or Chernev's Practical Chess Endgames (puzzles), but mostly from reviewing more than 2,000 annotated master games, many of which got to endgames.
You can see much of this advice (and more) in the Endgame chapter of my most recent book The World's Most Instructive Amateur Game Book.
Question Could you do a column on the concept of "equality" in chess? I often come across game annotations that state the "position is now equal" as if this is obvious to the reader; when it is far from obvious to me.
Answer Early in the game an evaluation of "equal" usually means both sides have equal chances to win, and if both sides played perfectly, the game should end in a draw. Computers most often evaluate these positions between +0.2 to -0.2 pawns, roughly.
In the late endgame "equal" usually means more than this; it means that if both sides played perfectly, the game would end in a draw. Often this is differentiated by using the term "drawn" rather than "equal."
Novice Nook #156 (Ebook)
In PDF format. Viewable in any PDF viewer.
© 2014 Dan Heisman & BrainGamz, Inc. All Rights Reserved.
Dan welcomes readers' questions; he is a full-time instructor on the ICC as Phillytutor.
Comment on this month's column via our official Chess Blog!
Purchases from our
Home Page] [ChessCafe
© 2014 BrainGamz, Inc. All Rights Reserved.