Winning starts with what you know
The new version 18 offers completely new possibilities for chess training and analysis: playing style analysis, search for strategic themes, access to 6 billion Lichess games, player preparation by matching Lichess games, download Chess.com games with built-in API, built-in cloud engine and much more.
Some people say the high draw rate in elite professional chess is a problem, while others say it is not. Neither group can speak for the other. Personally I feel it is a big problem. I believe the example of shogi confirms that the related problem of unfought draws is an artifact of the high draw rate among hard-fought games.
In this essay I explain my rather different perspective on how the high draw rate problem could be addressed.
In the ChessBase.com articles, all the suggestions for fixing the high draw rate problem in elite professional chess have two things in common: (a) they are off-the-board rules changes, not on-the-board; and (b) they fail to subdue the high draw rate (unfortunately).
For instance, the popular Sofia rule of banning draw offers has resulted in a draw rate of 56% (see M-Tel tournaments 2005-2007). That is barely any lower than the whopping 60% rate from the five latest world or FIDE championship events (Kasparov-Kramnik 2000, through FIDE Mexico 2007).
Since most chess players would automatically reject all on-the-board rule changes even before reading the proposal, I agree that on-the-board rule changes are an oxymoron for chess. Thus, the high draw rate problem in chess can never ever be fixed.
None of the off-the-board rule change proposals would succeed at cutting the chess draw rate in half. The off-the-board approach is probably a dead end. Therefore it is time to add a new angle of investigation toward this whole high draw rate problem. As an academic investigation, let us design new chess-like games that have a low draw rate; even though we do not intend to play these games.
Perhaps we can advance our academic understanding of chess by engineering new hypothetical games in the chess family, new implementations of the abstract chess concepts. Exactly what about the rules of chess cause its high draw rate? How close did the chess rules come to avoiding this problem? We will never change any on-the-board rule in chess, but we seek better knowledge about our game for its own sake.
Many chess variants have been invented, but their goal is to do "cool" new things. That is not the goal of this new engineering challenge. Instead...
I propose that ChessBase.com challenge its readers to engineer the best answers to the following question:
"What chess-like game could be invented that would both (a) keep its draw rate low, yet (b) feel as much like chess as possible?"
The best answers to this challenge could deepen our knowledge of chess. I interpret Mark Dvoretsky's writings as also calling for as small a change as is necessary:
"It's not an easy thing, psychologically, to abandon centuries of tradition. But if it has to be done, then let us try to reach the goal we have set (in this case, getting rid of opening theory) by the most economic means, with the smallest possible changes."
My answer to the challenge is chess-3, which I discuss further below. Before discussing chess-3, I present several philosophical ideas about the challenge.
Shogi
Shogi (loosely called Japanese chess) is very similar to chess, so shogi could be another answer to this challenge. The draw rate in shogi is under 2%. Due to its low draw rate, shogi suffers no unfought draw problem artifact. Unfortunately, the differences between shogi and chess are large enough that shogi has a very different feeling than chess has.
Gothic Chess
Gothic Chess could be seen as a response to this challenge. Its two extra pieces, Capablanca's N+R and N+B, add so much power that tactics are supercharged and complexity is increased. The opportunities for decisive gains are clearly larger than in chess. I would guess that in elite Gothic Chess that the draw rate would be about half of what the rate is in chess. That amount of reduction would be a big success.
Capablanca's two powerful pieces merely recombine atomic powers that are already present in chess. But they add so much more power that defensive play becomes a secondary art. Knights become almost superfluous. Often the role of a knight is to sacrifice themselves for a pawn. These differences wander far away from the chess feeling. Further, the larger 10x8 board also gives Gothic Chess its own very different feeling.
Gothic Chess is not my preferred taste. But if you like long sequences of clever tactical maneuvers with chess-like pieces on a checkered board, you might want to at least once replay a Gothic Chess game like this one (from move-pairs 13-44).
Fischer Random Chess
FRC (chess960) is Bobby Fischer's answer to this challenge. FRC was intentionally designed to retain the feeling of chess, which it does extremely well after the first 6-7 move-pairs. However, I estimate that FRC would have only a modestly lower draw rate, perhaps 50% instead of 60%. By itself FRC is an insufficient solution.
Dvoretsky Random Chess
Mark Dvoretsky urges that some major chess tournaments switch to a toned down implementation of the FRC concept of varied start positions. Dvoretsky would use dice to determine the initial placement of one white and one black pawn on their third ranks (thus 64 setups, most asymmetrical). White could start with his 'd' pawn as Pd3, while Black could have Pf6. Then the game would start. Like FRC, Dvoretsky's DRC does not add piece power, so I doubt it would make more than a slight dent in the draw rate.
I have quickly dreamed up a few other answers to the engineering challenge. These are only for illustration (I do not like these ideas):
Games could be engineered to target the endgame phase for decisiveness. This engineering strategy would be intended to avoid the butterfly effect: the notion that a small effect now sets off a chain of effects that accumulate to a large effect later. In this challenge we might try to delay the effect of a rule difference in order to duplicate the opening and middle-game feelings of chess without disturbance until the endgame phase.
Beforehand, we must objectively define the start of the endgame phase. I can think of three ways to objectively define the start of the endgame (whichever of the following comes first):
Let us consider games that are entirely like chess until the endgame (though I am not recommending these):
Game 1. Exactly like chess; except pawns can move (or capture) diagonally after the endgame phase starts. This example game is mentioned mostly to illustrate how an on-the-board rule difference might be engineered to minimize its effect on the whole game. We must ask whether knowing about this diagonal pawn potential would affect play in earlier phases, thus disturbing the chess feeling of those earlier phases by enough to matter. My guess is this rule difference would have only a small effect on opening and middle-game play. I am not confident that this game would have a far lower draw rate than chess has.
Game 2. Exactly like chess; except that a king cannot capture enemy pawns that are in the king's home half of the board. Illegal would be 88...Ke8:Pe7. Again I can only guess whether this would have a much lower draw rate.
From these examples, does it seem that a decisive rule change delayed until the endgame would preserve the feeling of chess in the earlier phases?
The butterfly effect theory I used to engineer games that are exactly like chess until the endgame phase might be a weak theory with many exceptions. Fischer Random Chess (FRC, chess960) stands as an exception to the application of the butterfly effect to chess game engineering.
FRC has an enormous effect on the opening phase. But instead of the effects growing during the game, the opposite happens. FRC does produce middle game positions are chess never does. But those positions appear quite chess-like until one takes a studious look. Though distinguishable from chess middle games, these FRC positions are logical, and they feel completely like chess.
By the time the FRC endgame phase arrives, all effects from the FRC difference are merely statistical and cannot be detected in any individual game.
Chess-like games can avoid high draw rates by having more piece power than chess has. The chess-3 answer to the challenge is to add power to a few pieces directly. This produces a nice corollary dividend in that the casual exchange rate is reduced, thus indirectly adding power by retaining it on the board. Yet another indirect anti-draw dividend is more imbalances, which Jeremy Silman emphasized are needed for decisive outcomes.
For chess-3, here are the two major rules that differ from chess:
1. Enhanced Movement: Give enhanced powers of movement (not of capture) to one rook, one knight, and one bishop.
[T] One rook becomes a tower. A tower is a rook that can also move (again not capture) like a bishop. A king can castle with a tower.
[J] One knight becomes a jous (from the word jouster). A jous is a knight that can also move one square along a row or column (not diagonally).
[V] One bishop becomes a vicar (dictionary defines as a substitute bishop). A vicar is a bishop that can also move one or two squares (not three or more) along a row or column. (Allowing the vicar to move only in increments of two, four or six squares barely increases its value over a bishop.)
Note: No new chess equipment is immediately necessary. Instead use small colored hair ponytail bands as belts or collars around the enhanced pieces.
2. Varied Setups: Rather than massive start position variation like FRC uses, chess-3 has 48 unique setups. These 48 setups are nuanced variations of only six basic or skeleton setups. A single six-sided dice is used to randomly determine the exact setup shortly before the start of each game.
(a) The queen and king always start on columns 'd' and 'e' respectively, as in chess.
(b) The two rook-ish pieces, the rook and tower, are restricted to columns 'a' and 'h'. Randomly assign them to these columns.
(c) The vicar, bishop, knight, and jous are restricted to the four columns b,c,f,g. The vicar and bishop are allowed to start on the same shade of square. Randomly assign each of these four pieces to these four columns.
White and Black pieces of the same type always start on the same column.
There are 48 unique setups. Informally, each setup is an implementation of
one of the 6 different skeleton setups.
For example, the eight setups below are all based on the skeleton that is
extremely similar to chess (R NB QK BN R). The following eight setups all
have knight-ish pieces in columns 'b' and 'g', and bishop-ish pieces in columns
'c' and 'f'.
R NB QK VJ T
R NV QK BJ T
R JB QK VN T
R JV QK BN T
T NB QK VJ R
T NV QK BJ R
T JB QK VN R
T JV QK BN R
3. Fair First Move: After the first player completes the first ply of the game (with a white piece), the second player chooses which color pieces he wants to play for the whole game. Then whichever player is Black makes the next ply, and the game continues normally.
Thus White and Black will average the same percentage of wins; meaning Black no longer begins the game trying for a draw.
4. Opposite Wing Castling: The second player to castle has the option of leaving his king on any column a-c, if his opponent castled king-side (to the 'h' wing). Or the second player can castle his king to any column f-h if his opponent has already castled queen-side. For example, the castling ply "Ke8a8/Ra8b8" could be legal.
Games with opposite wing castling have a lower draw rate in chess. This rule encourages more opposite wing castling.
5. (A few other minor rules.)
The next several sections discuss the engineering logic behind chess-3.
Exchange sacrifices of R:N or R:B are rare among all exchanges in chess. In contrast, chess offers numerous opportunities for common and relatively casual exchanges. The casual exchanges include R:R and N:N and B:B, which are equal exchanges. In chess, opportunities for equal exchanges are doubled by the fact that each color has two of each of those officer types (R N B).
The exchange problem gets worse. It is quite something that a bishop and knight are very nearly equivalent in value. This means exchanges of N:B and B:N are also considered casual.
And it gets even worse.
The traditional start position from chess is one of the most exchange prone in all of FRC. By comparing chess and FRC we can see that the traditional setup is more exchange prone than roughly 82% of all FRC start positions. Only FRC setups that put all four bishops on corner squares can be worse.
In the start position of traditional chess:
All these factors interweave to create a perfect storm that too often tosses away piece power too early. The high draw rate is the unpleasant effect.
No chess-3 start position is the same as in chess, because the pieces available are different. Yet some chess-3 setups are highly similar to chess (such as RNBQKVJT or RJBQKVNT). Those chess-3 setups could be highly prone to early exchanges. Since these exchange prone setups come from only one of the four chess-3 setup skeletons, 75% of all chess-3 games would be less, or far less prone to early exchanges than the chess setup is.
In deference to the opinion of Mark Dvoretsky and others, who feels the FRC setup variation is too extreme, chess-3 gives players some consistency in the look and feel of all its 48 setups. Chess-3 does this without introducing the potentially unfair pawn asymmetry that Dvoretsky described.
The enhanced pieces add about 4.0 pawns worth of power to the board (per color). This is significant given that the board remains 8x8.
In chess-3, zero of the white king's seven officer pieces have a teammate of equivalent value; whereas in chess six of seven do. This may greatly reduce the rate of early exchanges.
In chess the 12 R+N+B pieces provide 18 equal or equivalent exchange combinations (such as R:R or B:N). But in chess-3 there are only seven such exchange combinations! (7/18 = 39%)
Most possible exchanges of officers in chess-3 create an imbalance that favors one player over the other. In other words, exciting exchange sacrifices would occur much more frequently in chess-3 than in chess. The resulting imbalances should be more decisive than the B versus N imbalances that Jeremy Silman's books emphasize.
An endgame of K+B+2P versus K+B+2P can be very drawish. But a percentage of these endings in chess-3 would instead be K+B+2P versus K+V+2P, where the vicar's extra mobility may make the game decisive. The advantageous imbalance of a vicar over a bishop during the middle game could prove decisive before the endgame is reached.
Is an exchange of Q for R+N wise? Is an exchange of a Q for a T+J wise?
Overall I suspect that the draw rate for chess-3 (among elite professionals) might be cut by almost in half, compared to traditional chess. The majority of the cut would be due to the increase in piece power. The smaller portion of the cut would come from the elimination of memorized opening variations, and by the use of start positions that are less prone to early equivalent exchanges.
So chess-3 may meet the lower draw rate demand of the challenge. But what about preserving the chess feeling?
If the towers and jouses and vicars rarely move with their enhanced mobilities, then the traditional chess feeling would be well duplicated by chess-3. Otherwise, the only other hope would be that with exposure the enhanced moves would eventually seem normal enough.
Imagine that you set up a traditional chess board. You place ponytail bands on one white R, N, and B, and repeat for three black pieces. Then you replay a grandmaster game score. As you replay each ply, consider whether an enhanced move (of a T or J or V) would have instead been preferred. When I do this, I find surprisingly few occasions for enhanced moves. Their frequency of use probably varies with the phases of the game. This surprising infrequency suggests to me that the disturbance of the chess feeling is of modest size, or less than I at first predicted.
All puzzles are White to move.
Problem 1
White: Ra1, Ke1, Th1, Ba3, Jc3, Qg3, Pa2-b3-e4-f3-g2-h2.
Black: Ra8, Ke8, Th8, Bb7, Jc6, Qf6, Pa7-b6-e5-f7-g6-h7.
Solution: 1. Jc3d5 Qf6d8 2. Jd5d6+ Ke8f8 3. Jd6:Bb7
Problem 2
White: Rf1, Kg1, Qg2, Th4, Bh6, Pa2-b2-d3-f4-g4-h3.
Black: Ta8, Kg8, Qf7, Ng7, Rd5, Pa6-b5-c5-d4-g6-h7.
Solution: 1. Th4e7 Qf7:Te7 2. Qg2:Rd5+ Kg8f8 3. Qd5:Ta8
Problem 3
White: Vd6, Jg4, Kh2, Pa3-g2-h3.
Black: Bd5, Re2, Kg8, Pa4-b5-g7-h6.
Solution: 1. Jg4f4 Re2d2 2. Jf4:Bd5 Rd2:Jd5 3. Vd6e6+
Problem 4
White: Ra1, Nc3, Vf2, Bg2, Kg1, Pa2-b2-f3-g4-h2.
Black: Ra8, Nc6, Je7, Vg5, Kg8, Pa7-b6-c7-f7-g7-h7.
Solution: 1. f3f4 Vg5:f4 2. Vf2f3
Chess-3 endgame position.
Loosely from V.Smyslov - G.Sax 1979, 0-1.
See John Nunn's "Understanding Chess Move by Move", page 221.
After 32. h2h4 Kd6c5, White to move.
White: Je3, Kg2, Pa4-b3-c2-f2-g3-h4.
Black: Nd4, Kc5, Pa7-b6-c7-f7-g6-h7.
Insufficient ply: 33. Kg2f1.
Alternative ply: 33. Je3d3+.
Here is a handy condensed list of the off-the-board rule change proposals discussed on ChessBase.com and elsewhere that I remember seeing.
Draw Offer Duration: Increase the lifetime of each draw offer (thus PGN improvements essential).
Third Occurrence of Position: Make it illegal to repeat a position for its third time, at least before the first time control.
Invitations: Do not invite drawish players, or those that open with the Petroff or Berlin defenses. Also, deny draw-prone players publicity.
Computers Decide: When a draw is offered, Fritz decides which players gets the win.
Forbid Short Draws:
(a) Sofia: Ban draw offers except those authorized by an arbiter who forbids
premature draws. Fine.
(b) Minimum Move-Pair Count: Ban draw offers until a minimum number of move-pairs
have been completed (and maybe a minimum clock time expenditure too).
Various Point-Fiddling Schemes: These seek "aggressive" or "risky",
which sounds like euphemisms for "unsound" moves; but the effect
would be so small that I do not care.
(a) Points Based on Elo: Give the lower Elo rating player more points for
a win or draw than the higher rated player gets.
(b) Draw Point-Fiddling: Give Black more points than White for a draw (but
less extreme than full Armageddon Chess).
(c) Win/Draw Ratio Point-Fiddling: Such as Bilbao, Win/Draw/Loss = 3/1/0.
(d) Points for each Loss: Idea is to reward you for the risks you took trying
to achieve a decisive victory.
(e) Money: Give players normal points for wins and draws to avoid confusing
the rating formula, but award prize money only for wins (thus making the
points mislead the rating formula).
(f) Non-Deterministic: Complex scheme to feed later results into a formula
that tells you how many points you get for a draw. Idea is that a player
could not guarantee himself a certain ranking in the final standings by
aiming for a draw.
Time Crunch: These ideas do not directly target the long time control game
(the first game) that we care about the most.
(a) Blitz Tie-Break: When a long time control game ends in a draw, play
Blitz games until one is decisive. This does nothing to improve the long
game, the only game we care about.
(b) Resume Clocks: When a long time control game ends in a draw, pause the
clocks, start a new game with colors reversed. Eventually someone runs out
of time (or mates). This does nothing to improve the long game, the only
game we care about.
Endgame King Situations: Checkmate no longer the only way to win.
(a) Denuded King Loses: A player loses when he has only a lone king left,
and his opponent has at least one officer (non-pawn piece), such as K vs.
K+N.
(b) First King Across Frontier Wins: The first player to get his king into
his opponent's home half of the board wins.
(c) Stalemate treated as checkmate.
I dislike the Bilbao point-fiddling scheme. I think its looks ever more flawed the more closely it is scrutinized (show us the specific alternative moves the players would play). Nor do I like replacing draws with an immediate subsequent Blitz tie-breaker game. The goal is produce a better long time control game, yet this Blitz proposal encourages players to quickly draw the long game and go directly to the brief energy-saving Blitz game.
But could these two weak ideas be combined into one better idea? The proponents of Bilbao and of Blitz tie-breakers should get together and suggest the following "Bilbao-Arm" proposal (I am not suggesting it):
Bilbao-Arm:
The points for the long time control game are W/D/L = 3/-/0. If the long game
ends in a draw, then the draw is immediately replaced by one Armageddon game
(where players bid for clock time with Black). The points in this Armageddon
are for White W/D/L = 2/0/0, and for Black 2/2/0.
Bilbao-Arm would eliminate the players' desire to rush through an easy draw in the long game. They would know a victory in the long game earns more points than in the tie-break game. But the long games would still suffer a high draw rate, despite the counter-incentive of the stingy point payout from the Armageddon tie-breaker. At least the Armageddon would provide the decisive outcomes that tournament narratives need to entertain their readers (albeit somewhat artificially).
About
the authorGene Milener was born in Brooklyn and raised in Oneonta, New York, USA. There was no direct value gained, he says, by his Bachelor's degree in a subject like Philosophy, but some concepts he learned from Philosophy continue to influence the way he thinks. His graduate degree was in Human Experimental Psychology, which did two great things for his life: he had to teach himself computer programming to run his experiments, which led to a new profession: since the early 1990's Gene has been employed at Microsoft in Redmond, WA, USA. And secondly, grad school was where he met Amy. They have two heartbreakingly wonderful daughters, Patia and Sara. In 2005 Gene "accidentally" wrote a book entitled Play Stronger Chess by Examining Chess960: Usable Strategies of Fischer Random Chess Discovered. "Think what you will about FRC-chess960," he says, "but when I researched it I found there was a whole lot more to say about FRC than the community had yet realized. To a significant degree, FRC is misunderstood." He expects to finish his next chess book in 2009.
|
|
Reader feedback: the great draw debate continues
|
|
|
Unfought draws – reader feedback |
|
|
Unfought draws – mathematical, logical and practical
considerations |
|
|
The problem of draws – feedback from our readers 04.01.2008 – The perceived problem of too many unfought draws in chess has led to a number of imaginative cures being proposed, involving the modification of the rules of the games, the scoring system and the prize distribution. Over Christmas we presented a particularly clever one: let the draw offer stand for the rest of the game. Here are reactions to this proposal and new ideas. Long interesting read. |
|
|
The problem of draws – a Christmas solution 29.12.2007 – It is perhaps not appropriate to take up the subject while a tournament in Moscow is registering the lowest drawing rate in recent memory. But the question of quick, unfought draws still occupies the attention of our readers, and many have sent in new and imaginative proposals. One is so clever that we advocate trying it out immediately. The first organiser to do so gets to name it after his city. |
|
|
Chess, football and the Bilbao Rule – Part II 15.11.2007 – The debate on the perceived problem of too many – unfought – draws in chess, and what to do about it, continues. The letters pour in and we keep receiving extensive, well thought through proposals that attempt to create incentives for playing to win. Josu Fernández presents closing arguments for the Bilbao System, while Serbian GM Dragan Solak tells us why he thinks it cannot work. |
|
|
Chess, football and the Bilbao Rule 08.11.2007 – The discussion and the search for remedies for the perceived problem of short draws in chess continues. Josu Fernández, a Spanish organiser, sent us a report on the effects of the Bilbao 3-1-0 system on the football league in his country, and on what this means for chess. Other readers too have submitted thoughful papers on the subject. Again, it is a long interesting read. |
|
|
The 'Bilbao Draw' – how it doesn't solve the problem 28.10.2007 – Chess fans and organisers all over the world are worried about the problem of too many draws in chess. Actually: about pre-arranged or unfought draws. Many remedies have been tried, including threats, prohibition and, most recently, the Bilbao system of awarding three points for a win and one for a draw. Is that the solution? No, says one astute reader and points to a possibly fatal flaw. |
|
|
ACP Survey: what the players think about draw offers 22.03.2007 – In February the Association of Chess Professionals asked its members what should be done to combat short, unfought draws, which are often perceived to be the bane of chess. The ACP published a questionnaire, 171 members replied. |
|
|
ACP Survey: What do you think about draws? 11.02.2007 – Short, unfought draws are the bane of chess fans. That at least is the public perception. The Association of Chess Professionals (ACP), which has 227 members, has launched a questionnaire to find a remedy. |
|
|
Embracing Risk in Tournaments 14.12.2006 – The issue of playing style is not normally given much consideration in chess. In an interesting article computer scientist Darse Billings maintains that it is an important factor in the probability of winning. A player who tends to win or lose games has a significantly better chance of success in a tournament than a player who draws a lot of games. Read and consider. |
|
|
The draw problem – a simple solution 10.11.2005 – Recently Ignatius Leong and Leung Weiwen made a very radical solution to the problem of too many draws in chess. This led to a vigorous debate amongst our readers – we bring you a selection of their often very interesting letters. But we start off with the voice of reason: John Nunn analyses the problem and proposes a much simpler solution. |
|
|
A Cure for SAD (Severe Acute Drawitis)? 03.11.2005 – Draws, draws, draws – the problem has always faced chess, and it seems that there is no clear way to solve it. However, Ignatius Leong and Leung Weiwen, both of Singapore, offer a radical new proposal that would decide every game of chess in a sporting fashion. Will it catch on? |
|
|
Draws forbidden in Super-GM tournaments 01.04.2005 – When a bunch of world class players get together for a tournament the danger is that there will be a lot of draws. A new organiser who is staging a Super GM event in Sofia, Bulgaria, has come up with a new idea: ban draw offers. The participants have to play on until the arbiter says they can stop. Will this become a fixed feature in chess events? |
|
|
Short on draws 18.03.2004 – "I know that with perfect play, God versus God, Fritz versus Fritz, chess is a draw," writes Nigel Short, who describes a deadly disease called Severe Acute Drawitis. "Those afflicted with SAD display an uncontrollable urge to offer or accept premature peace proposals." Read about it in Nigel's highly entertaining Sunday Telegraph column. |
