Handicap (chess)

Handicaps in chess are handicapping variants which enable a weaker player to have a chance of winning against a stronger one. There are a variety of such handicaps, such as odds, extra moves, extra time on the chess clock, and special conditions. Various permutations of these, such as pawn and two moves, are also possible.
Handicaps were quite popular in the 18th and 19th centuries, when chess was often played for money stakes, in order to induce weaker players to play for wagers. Today handicaps are rarely seen in serious competition outside of human–computer chess matches. As chess engines have been routinely superior to even chess masters since the late 20th century, human players need considerable odds to have practical chances in such matches – as of 2024, approximately knight odds for grandmasters.

History

According to Harry Golombek, "Odds-giving reached its heyday in the eighteenth century and the early nineteenth century." Indeed, it was so prevalent in the 18th century that Philidor played the vast majority of his games at odds. About fifteen percent of the known games of Paul Morphy are games in which he gave odds. Other strong odds-givers of this time were George Henry Mackenzie and Wilhelm Steinitz.
Howard Staunton in The Chess-Player's Handbook advised inexperienced players to accept odds offered by superior players and, upon improving to the point that they can themselves give odds to some players, to avoid playing such players on even terms, warning that doing so is apt to induce "an indolent, neglectful habit of play". In 1849, Staunton published The Chess-Player's Companion, a 510-page work "chiefly directed to the exposition of openings where one party gives odds". Just over 300 pages were devoted to odds games: Book I contained games played at various odds, and most of Book V discussed various types of odds, including exotic and unusual ones. The late-19th century chess opening treatise Chess Openings Ancient and Modern, by Edward Freeborough and Charles Ranken, included fourteen pages of analysis of best play in games played at odds of pawn and move, pawn and two moves, and either knight.
Macon Shibut writes that in the mid-19th century "chess was a gambling game.... Individual matches for stakes were the focus of organized play. Matches between leading players attracted a wide following so masters often succeeded in finding sponsors to back their personal wagers." The available sums were generally relatively meager, however, and travel was arduous, so the amount of money obtained in this way was not sufficient to enable professional chess players to support themselves financially. Moreover, the first major chess tournament was not organized until 1851, and chess tournaments remained a rarity for several decades following. With tournaments an unreliable means of making a living, odds-giving became a way for masters to entice amateurs into playing for wagers, since the odds gave the amateur a fighting chance. The odds system even became the earliest rating system: amateurs were graded according to what handicap they needed to compete against a master, and were referred to as a "Rook player" or "Pawn and move player", for example, as many people would today speak of players by their Elo ratings.
The playing of games at odds gradually grew rarer as the nineteenth century proceeded. Today, except for time odds, they have all but disappeared. Shibut posits that games played at material odds became unpopular for technological, political, and philosophical reasons. Taking these in turn, first, the introduction of chess clocks gave rise to a new way to give odds, one that has today supplanted material odds as the preferred mode of odds-giving. Second, the Soviet Union supported chess masters and sponsored chess education, but expected chess masters "to be cultural icons, not hustlers". Third, chess began to be treated in a scientific, logical way, "with an assumption of idealized 'best play' to underpin all analysis". From this perspective, a game beginning from a "lost" position becomes less interesting, even distasteful. Writings by Wilhelm Steinitz, the first World Champion, and James Mason are consistent with the last point. GM Larry Kaufman argued in 2024 that another factor was that chess became more popular and the standard of play rose, so that it was no longer reasonable to give piece odds to strong players.
World Chess Champion Bobby Fischer often gave odds, as did IM Israel Albert Horowitz before him. Against FIDE Master Asa Hoffmann, Fischer first gave pawn and move, then pawn and two moves, and then pawn and three moves. In an interview with Ralph Ginzburg published in the January 1962 issue of Harper's Magazine, Bobby Fischer was quoted as saying that women were weak chessplayers and that he could successfully give knight odds to any woman in the world. Fischer later claimed that Ginzburg had distorted what he had said.
In 2001, London businessman Terence Chapman, a master-level player, played a match against former world champion Garry Kasparov, with Kasparov giving odds of two pawns in each game ; Kasparov won the match by two games to one, with one draw.
Rybka, a top-rated computer chess engine designed by International Master Vasik Rajlich, played a series of handicap matches against strong human players. In March 2007, Rybka defeated Grandmaster Jaan Ehlvest after giving pawn odds. In January 2008, Rybka defeated GM Joel Benjamin after giving draw odds. In March 2008, Rybka gave pawn and move to GM Roman Dzindzichashvili, drawing the match 4–4. In June 2008, Rybka gave knight odds to FIDE Master John Meyer, losing 4–0. On July 6, 2008, Rybka gave Meyer odds of pawn and three moves, winning 3–1. In 2015, Komodo defeated strong grandmasters at the f7-pawn handicap and rook for knight, without losing a single game.
Top human players still occasionally play odds matches against engines. In 2016, Komodo played Hikaru Nakamura in four odds games, giving pawn & move odds, pawn odds, exchange odds, and 4-move odds. The first three ended drawn, while Komodo won the last game to win the match 2½–1½. In 2018, Komodo played another handicap series against Maxime Vachier-Lagrave. Komodo won four games at pawn and two moves odds, two-pawn odds, queen for rook + queen's knight odds, and knight for f7-pawn odds. Maxime Vachier-Lagrave won at odds of exchange and pawn for knight, while the last "Knightmare" game was drawn. Finally, in 2020, Komodo played a 6-game match against GM David Smerdon at knight odds. GM Smerdon blundered away the first game, but rallied to win the remaining five.
From 9 November 2023, Leela Chess Zero supports playing at knight odds on Lichess. Against "par" grandmasters, the bot scored +24−5=2. Kaufman suggests that "only an 'elite' grandmaster could win a rapid match from LeelaKnightOdds". Later, queen for knight odds and queen odds were also added, along with the option to play as either colour Rook odds were added on 29 February 2024, with the choice of rook and colour.
The first version of LeelaKnightOdds played a ten-game knight odds blitz match against GM David Navara on 30 March 2024, alternating between the b1 and g1 knights: the first two games were at 5+3, the next four at 3+2, the next two at 3+1, and the last two at 3+2 again. Navara won 7–3. After upgrades to use a specialised network for chess handicaps, LeelaKnightOdds played a match against GM Awonder Liang on 12 December 2024, alternating between the b1 and g1 knights. Liang scored +1−6=7, going +0−6=4 among the 3+2 games.
Kaufman argues that odds chess is excellent for training. He argues that if a master plays a weaker student for training, then the following options are available, and he finds the last one preferable:

The master can point out and take back the student's errors, but then that amounts to playing a game against oneself instead of against the student;
the master can deliberately make blunders, but then that defeats the purpose of allowing the student to play against a master;
or the master can offer a fair handicap, allowing the student to tangibly feel progress while advancing to lesser and lesser handicaps. In Kaufman's experience, there is always some handicap available that will any given difference in strength.
Handicaps

The purpose of a handicap, or odds, is to compensate for the difference in skill between two chess players. There are a variety of handicaps: odds; extra moves; time odds; special restrictions ; weighting of results ; differential stakes; and physical restrictions, such as blindfold chess. Many different permutations of handicaps are also possible, as are countervailing handicaps.

Main

The following list is based on that of Larry Kaufman, mostly describing the 19th-century situation.

Draw odds ;
Pawn – quite rare, since time odds could substitute;
*When a single pawn is removed, it is always the f-pawn. This is because removing any other pawn gives some compensation in the form of opening a file or diagonal for a rook, bishop, or queen.
Pawn and move;
*This is a much larger handicap than f2-pawn odds, because after 1.e4, the threat of 2.Qh5+ means that Black has few options: 1...e5 loses immediately, 1...c5 and 1...d5 are gambits, 1...c6 fails because 2.Nc3 prevents 2...d5, and 1...g6 and 1...d6 are very risky because of the plan h2–h4–h5. The best options are probably 1...e6 and 1...Nc6; 1...Nh6 2.d4 Nf7 is also possible but loses time. Therefore, pawn and move is more like a 1.5-pawn handicap, though it demands both players to have specialised opening knowledge.
*Staunton sometimes substituted rook for knight , which is similar in degree but demands less specialised knowledge than pawn and move. Exchange and move was also occasionally given historically. Exchange odds were never historically popular, however, because of differing values of pieces across the game: the knight is more useful than the rook in the opening, but the handicap becomes felt when pieces are traded to give an endgame, where being up rook for knight is almost always decisive. Kaufman writes that rook for knight is a larger handicap than f-pawn if the stronger player has the same colours in both, but that pawn and move is more severe than a1 for b8. Play is different in both: f-pawn handicaps necessitate defensive play and attempting to reach an endgame, whereas exchange handicaps necessitate attacking play.
Pawn and two moves ;
*Almost always, 1.e4 2.d4 was played. However, Kaufman notes that 1.e4 2.Nc3! is stronger, as the otherwise desirable move 2...e6 loses a second pawn. This can be avoided by 2...d6 or 2...Nh6, but in either case Black lacks counterplay against White's advantages in material, king safety, space, and development.
*Two pawns is an alternative, comparable to but slightly less than pawn and two moves. It was used by Komodo Dragon against Nakamura.
Pawn and three moves ;
*It is forbidden to cross the fourth rank, because 1.e3 2.Bd3 3.Qh5+ g6 4.Qxg6+ hxg6 5.Bxg6 would win immediately. Asa Hoffmann played 1.e3 2.Bd3 3.Qg4 against Fischer at these odds, forcing the win of a second pawn, but Kaufman suggests that 1.e4 2.d4 3.Bd3 or 1.e4 2.d4 3.Nc3 are better.
*Historically, opinions differed on how many moves at pawn handicap were valid thresholds before reaching the knight handicap. Pawn and three moves was often considered to be too close to the knight handicap historically, but Staunton, Morphy, and Fischer all used it as an intermediate step between pawn and two moves and knight odds. Fischer sometimes even gave pawn and four moves. Test games from modern engines show that pawn and three moves is useful, being about midway between pawn and two moves and knight odds; on the other hand, pawn and four moves is approximately as big a handicap as knight odds, making it not very useful as knight odds gives a more normal game.
*A comparable handicap is two black pawns, which Kaufman recommends to substitute pawn and three moves as producing more normal play.
*Philidor used instead knight for second move: White plays 1.e4 and gets to move again, but gives up a knight. Staunton also thought favourably of this handicap, as does Kaufman.
Knight ;
*Before 1870, it was more common to give the b1-knight; after 1870, the g1-knight handicap becomes more common. In particular, Horowitz always gave the g1-knight. In both human and engine play, it is somewhat easier to give the g1-knight. This is because without the b1-knight, 1.e4 is a bad move that is best answered by 1...d5!, because after 2.exd5 Qxd5 there is no knight to chase the queen away. The Queen's Gambit also does not work well when no knight can come to c3, so that 1.f4, 1.b3, or the London System become White's best options. On the other hand, 1.e4 and the Queen's Gambit both work well without the g1-knight, though Kaufman suggests that the Jobava London is even better since the plan with 0-0-0, f3, and g4 with a attack does not need the g1-knight.
*Morphy used an extra step on the handicap ladder, ranking b1-knight odds with the restriction that Black must meet 1.e4 with 1...e5 as one step lower than unrestricted b1-knight odds. But even with this restriction, best play for Black is 1.e4 e5 2.Nf3 d5!
*The bishop handicap was hardly ever used historically, because the bishops are not interchangeable: the player with a missing bishop is incentivised to place friendly pawns on the opposite colour of the remaining bishop. Thus it gives a less normal game than a knight handicap.
*The knight handicap has recently been used in many matches between engines and GMs, with the knight given up alternating between games. In April 2020, GM David Smerdon won 5–1 against Komodo ; in July 2021, GM Anthony Wirig won 6–3 against Komodo Dragon ; in January 2022, GM Eugene Perelshteyn and GM Ben Finegold both won 4–2 against Komodo Dragon ; and in March 2023, GM David Navara won 7–3 against LeelaKnightOdds.
Knight and move ;
*This was rare historically. Black cannot create counterplay or force as easily as White, so that the move is a significant advantage.
Rook ;
*Historically, the a1-rook is always the one given up, so that the more common kingside castling is still possible.
*Sometimes castling was allowed without the a1-rook, but Staunton argued that it should not. Kaufman suggests not allowing this option in the modern age, because databases and servers do not allow it.
*Some players added an extra step before rook odds, with the a-pawn already moved to a3 so that it is protected. However, Morphy and Staunton did not make use of this variant.
Rook and move ;
and then the whole cycle could be repeated with an extra rook, with queen odds substituting two rooks.

Even with the "no moves beyond the fourth rank" proviso, Black cannot give White an unlimited number of moves. Doing so would allow White to set up the position at right, when White's dual threats of 1.Qxf7# and 1.Ned6+ cxd6 2.Nxd6# are immediately decisive. Kaufman also points out that pawn and four moves is already problematic due to 1.e3 2.Bd3 3.Qg4 4.Nc3, and now Black is forced to give up the e-pawn and trade queens.
For the modern era, Kaufman advocates the following list:

Handicap	Eval	Notes
Odds of the move	0.17
Two moves	0.48	White is better, but objectively this is not a winning advantage. However, three moves would be.
Pawn	0.71
Pawn and move	1.20
Pawn and two moves	1.59
Two pawns and move	2.03	To replace the historically unpopular pawn and three moves; odds of two pawns gives a more normal game.
Knight for second move	2.44
Knight	2.81	For a match, alternate the knight removed. Kaufman calls this "the critical handicap between engines and grandmasters", and notes that it is also useful and important between good players. It is a smaller handicap to give the g1-knight.
Knight and move	3.28	Historically rare, but fills the large gap between knight and rook odds.
Rook	3.70	0-0-0 is not allowed; the a-pawn starts on a2.
Rook and move	4.07
Rook and pawn	4.48
Queen for knight	5.21	Replaces Morphy's preferred and comparable rook, pawn, and move, as this produces a more normal game.
Queen for knight and move	5.48	Replaces Morphy's preferred and comparable rook, pawn, and two moves, as this produces a more normal game.
Two knights	5.64
Two knights and move	5.84
Rook and knight	6.59
Rook, knight, and move	6.76
Rook, knight, and pawn	7.33
Rook, knight, pawn, and move	7.56
Queen	7.95
Queen and move	8.20
Queen and pawn	8.80
Queen, pawn, and move	9.08
Queen, pawn, and two moves	9.44

Kaufman mentions even greater handicaps, such as queen and knight; queen and two knights; and queen and two rooks.
Leela Chess Zero supports the following odds, with both colours:

Knight;
Rook;
Queen for knight;
Two knights;
Two bishops;
Rook and knight;
Two rooks;
Queen;
Two bishops and knight;
Rook and two knights;
Rook and two bishops;
Queen and knight;
Two bishops and two knights;
Queen and rook;
Queen and two knights;
Queen and two bishops;
Queen, rook, and knight;
Queen and two rooks.