USCF Home Chess Life Online 2008 July Playoff Theory
|By Tom Braunlich|
|July 31, 2008|
An Analysis of
Tiebreak Options - Old and New - and Proposal for Improved Tournament Playoff
"Everybody talks about the weather, but nobody does anything about it."-Mark Twain
Much the same can be said about chess issues. The recent controversy over the final playoff game of the 2008 U.S. Women's Championship showed that, although there is no shortage of opinions, there are very few actual guidelines for organizers on how to break ties in a chess tournament.
As Chief Organizer of those championships, I not only read with interest the voluminous online commentary on the issue, but also received a lot of emails from top players about it.
For example, GM Joel Benjamin wrote, "The criticism directed at the U.S. Championship organization for the unfortunate events of the Armageddon playoff is ridiculous and overblown. People may not like what happened, but the playoff procedure was fairly standard and spelled out for the players, despite the contentions of many Internet ‘commentators'." This contrasted sharply with many of the online statements, which flatly declared the playoff method was "boneheaded," "degrading," or "demeaning to chess."
I began to wonder, "What actually is the theory of playoffs? What system of breaking ties is the most suitable for chess?" Surely playoffs are a relatively straightforward thing, and should be no great mystery, even if they are fairly awkward due to the peculiarities of chess. So I contacted several top players and tournament directors to investigate these issues, and researched the rules and regulations for this article. Responders included a mix of players, organizers and TDs: GMs Joel Benjamin, Yasser Seirawan, Sergey Kudrin, IMs John Donaldson, Larry Kaufman and Greg Shahade as well as TDs Mike Atkins, Frank Berry, Bill Goichberg, Stewart Reuben, and others.
This article aims to analyze the current chaotic situation - keeping in mind the needs of players, organizers, and sponsors - and come to some conclusions. The goal is to do more than just generate more rounds of endless Internet debate - but instead to actually move the debate forward toward some positive action.
First, to make the underlying "theory" easy to understand, I start with some general comments and then review in detail the various options for how to handle playoffs that have been proposed, discussing each in turn and listing their pros and cons:
After that, we have some conclusions, and then a proposal for chess playoff guidelines featuring the best elements of the above options.
The Need to Break Ties
All sports have a need to break ties, at least at certain times. Some sports do it more elegantly than others. The "tiebreaker" in tennis is well regarded, for example (even though it was derided when first proposed). In soccer (football), the use of penalty kick "shootouts" is the approved method, though not liked by purists.
Sometimes tiebreaks can be especially exciting and popular, such as the playoff between Tiger Woods and Rocco Mediate at the 2008 U.S. Open in golf.
Chess has a special disadvantage when it comes to breaking ties, due not only to the possibility of drawn games but also to the asymmetrical nature of the game caused by the advantage of playing white, which leads to the awkward need to play sets of two playoff games to balance it.
Perhaps surprisingly, neither the USCF nor the FIDE tournament rules have much to say about playoffs. There is a lot of information about mathematical tiebreak methods, but both rulebooks imply that a playoff of some sort is to be preferred if time allows. Yet actual recommendations for how such playoffs would work are scant.
The USCF rules , when discussing the breaking of ties (34.A), state, "In other events where time is not pressing, playoffs provide a better alternative to traditional tie-break systems. Playoffs are often conducted at a faster time control than the tournament, even five-minute games have been used." That is about it. There is also rule 34.E12, which is similarly vague: "The speed play-off, an exciting way to wind up a tournament, has been used as the first tie-break to determine the title at several major events."
The FIDE tournament rules have an annex that discusses the breaking of ties. The section on playoffs specifies a few obvious things such as the system must be announced in advance and should be controlled an arbiter, etc., but the guidelines listed for how a playoff should actually work are only this:
"6(k) The following is an example where time for playoff is somewhat limited.
1. If two people tie for first:
(a) They play a two game mini-match at the rate of all the moves in 3 minutes add-on 5 seconds for each move from the first. If this match is tied:
(b) They play a one game Armageddon (sudden death), White has 5 minutes and Black four, in the case of a draw Black wins first prize."
After discussing variations on this scheme when dealing with multi-player ties the FIDE rules add:
12. Where only two players are involved in the playoff, if time permits, they may play at a slower rate of play by agreement with the CA and CO. If the original games run very late, the playoff may go straight to Armageddon.
So, contrary to the many online commentators who were certain that the use of an Armageddon blitz game as the fifth playoff game in the 2008 U.S. Women's Championship was an unprecedented "abomination," it can be seen that the FIDE guidelines are similar and require much less. (More on the appropriateness of blitz playoffs below.)
In any case, for both USCF and FIDE, these rather sparse guidelines seem inadequate for such an important aspect of tournament chess.
In my 20-year career as a professional game designer I wrote rules and regulations for dozens of commercial games. It can be deceptively tricky. "Rules lawyers" and critics pick every word or phrase apart. Clearly, the writers of the playoff regulations quoted above wanted to avoid being too specific, as too much detail might prove to be inflexible and impractical for organizers in some circumstances.
However, it should be possible to elucidate some useful guidelines without being overly detailed, because other sports do it successfully. The trick is to outline recommendations relative to the variable factors (time, goals, etc.); including when to settle for co-champions, when to use math tiebreaks, when to use playoffs, etc.
The example of golf provides an excellent comparison in this regard, and in view of the analysis of the options (below) it is clear that something similar could be done for chess. I quote how they do it here. Note how clear it is, without being inflexible, even though there are several different types of golf events (stroke or match play, handicaps, etc.).
From the Rules of Golf (PGA):
11. How to Decide Ties
In both match play and stroke play, a tie can be an acceptable result. However, when it is desired to have a sole winner, the Committee [what we call the Organizers-TB] has the authority, under Rule 33-6, to determine how and when a tie is decided. The decision should be published in advance.
The USGA recommends:
A match that ends all square should be played off hole by hole until one Side wins a hole. The play-off should start on the hole where the match began. In a handicap match, handicap strokes should be allowed as in the stipulated round.
(a) In the event of a tie in a scratch stroke-play competition, a play-off is recommended. The play-off may be over 18 holes or a smaller number of holes as specified by the Committee. If that is not feasible or there is still a tie, a hole-by-hole play-off is recommended. ...
(c) If a play-off of any type is not feasible, matching scorecards is recommended. The method of matching cards should be announced in advance and should also provide what will happen if this procedure does not produce a winner. An acceptable method of matching the cards is to determine the winner on the basis of the best score for the last nine holes. If the tying players have the same score for the last nine, determine the winner on the basis of the last six holes, last three holes and finally the 18th hole. ...
Their method is very clear to understand. It goes from ideal forms to more and more abbreviated forms relative to circumstances. If there is a tie, and if you want to have a sole winner, use a long playoff. If there isn't time for that, use shorter playoff or a hole-by-hole playoff. (These translate to full-length playoff games as compared to rapid or blitz playoff games in chess). If there isn't time for playoffs at all, use "matching score cards" to break the tie. (This is comparable to mathematical tiebreaks in chess.)
Please also note that these guidelines are clearly labeled "recommendations." They are not absolute, and allow (as is necessary) an organizer to do something different as circumstance may require (as long as it is announced in advance). In that case, however, players and observers have these guidelines to help evaluate the fairness of any other tiebreak method the organizer wants to use.
Golf has several things in common with chess, and their method of dealing with tiebreak issues makes a useful comparison. Below I use a similar method to propose a set of guidelines specific to chess.
But first, let's analyze the many different proposals chess has had for breaking ties - or even the question of if ties need to be broken at all. There is a wide variety of opinions on all this.
Pros and Cons for Each Option
1) Co-Champions - "Tied players should just be declared co-champions. Don't bother to break ties."
"I like the old days when they would just be co-champions." (GM John Fedorowicz). This simple solution is advocated by many commentators, who point out producing a winner is a social convention that may not be necessary.
Pro - If a tie is an acceptable result, there is no reason to break it. This approach has big advantages in time, simplicity, and fairness.
Con - Unfortunately, it won't work in all cases. For example, it won't work:
a. When one player is scheduled to qualify from this event for a future competition, such as earning a spot on the Olympiad team. A method to break a tie in such a case is required.
b. When the public demands a champion - i.e., a world championship match, or any "championship" from national down to state or city levels which are desired to produce a clear winner for whatever reason.
c. When a sponsor won't abide what it might consider a ‘dud' result and wants a winner determined. Facetious example: CEO of Coca-Cola: "We have your chess sponsorship check here for $1 million, but there's just one thing. We don't like your idea of allowing co-champions. Our P.R. Department wants you to determine a clear winner." Chess Organizer: "Sorry, no thanks then. That's a deal-breaker. Chess commentators think co-champions is the way to go."
Conclusion - As GM Benjamin, IM Donaldson, and TD Mike Atkins (and others) point out, if co-champions are acceptable it should be the preferred method of dealing with a tie. Therefore the proposal below will list it as the "first choice" option. However, it is not a complete solution by itself.
2) Mathematical Tiebreaks - "If there must be a winner, mathematical tiebreak points are more fair than a playoff."
Math tiebreaks are often used, with the idea that their inherent logic tends to reward the player who played the more difficult schedule. "Why shouldn't the player who faces the stronger field be rewarded?" says Donaldson, adding that, "this might have more integrity than some form of rapid/blitz playoffs."
Pro - Simplicity and speed again favor this system over playoffs. Another nice advantage is that they efficiently handle multiple ties (3 or more players). The most popular tiebreak formats can generate instant results directly from the computer used to make Swiss pairings. Often players can estimate their tiebreaks before playing the last round and know from this whether or not they must play for a win.
GM Sergey Kudrin thinks math tiebreaks especially make sense in breaking ties further down the crosstable or in situations where there are multiple qualification spots up for grabs. He said, "I will distinguish between a Championship, where playoff might be desirable, and spots down the line. If we have 5 spots in the World Cup, I would not use playoff for those." For example, at the 2008 Frank K. Berry U.S. Championship Qualifier Open, there were seven slots in the U.S. Championship available in a 7-round Swiss. Four players tied for first, each winning a slot. Six people tied for the remaining three slots. These slots were decided by mathematical tiebreaks. But can you imagine trying to have playoffs to determine three winners from six players? It would be awkward and time-consuming
Con - First, it is worth noting that both the FIDE and the USCF rules, as sparse as they are, imply that a playoff - even a blitz playoff - is preferable to a mathematical tiebreak. This seems to be the opinion of many top players. GM Joel Benjamin: "If a tie must be broken, tiebreaks are much worse than a playoff (even in a Swiss). It may be okay to use tiebreaks to determine who gets a trophy, but not who gets the U.S. Championship title. If there is a multiple tie, you can use tiebreaks to whittle down to the top two for a playoff. Tiebreaks are ridiculous to even contemplate in a round-robin, when everyone plays the same field."
Although fine for many applications, math tiebreaks won't work well for all forms of chess events, and their underlying logic has been the subject of much debate.
· They don't work at all for a match. Nor do they work for "elimination" formats with "Finals" events such as many of the recent U.S. Championships. They don't work well for an event in which two people have perfect scores, which often happens in big scholastic events.
· They are dubious for round robin events, where the logic of why one certain type of score is given more weight than another is pretty flimsy, even for the common Sonnenborn-Berger system. (However, I should note that IM Greg Shahade has been developing some innovative new tiebreak systems for round robins.)
· Because the player has little control over the outcome of math tiebreak, it can lead to a feeling of helplessness and dissatisfaction as profound in its way as blitz playoff games have been accused of being. For example, in a typical state championship 6-round Swiss a characteristic problem is this: Players A and B tie for first place with 5½ points, drawing with each other and beating all their other opponents. Player A played the highest-rated player in round 3 and won, but that player, discouraged by his loss, dropped out of the tournament, thus screwing up Player A's tiebreaks. So B wins. Such things often seem like an unfair result, especially given the luck factor inherent in Swiss pairings to begin with.
· There are even more typical criticisms of math tiebreaks, which I will leave for another day. And it is worth noting that FIDE and USCF vary quite a bit on recommended tiebreak priorities to be used.
Conclusion - Though tiebreaks are functional for certain events they won't work for all cases, and thus are not a complete solution. Math tiebreaks are certainly a good option if there is no time for any playoffs, or for multiple qualification spots. However, most players, and indeed also the existing regulations, both prefer a playoff if time is available.
3) Chance Tiebreaks - "If there must be a winner, a method of chance is more fair than crazy rapid/blitz games."
Several commentators hate the spectacle of a time so much that they actually propose flipping a coin instead. Advocates of this idea believe that long playoffs are fine, but that chance would be fairer than rapid/blitz playoffs.
Pro - Chance-based tiebreaks have the advantage of being completely fair (in the sense that the odds of winning are equal). The argument here is that rapid/blitz playoffs involve a type of chess skill that is different from full-length tournament chess skill. A "blitz specialist" like Nakamura, for instance, has somewhat of an edge in blitz playoffs that he might not necessarily have in normal tournament play. Thus playoffs at a different time control than the tournament itself may have unfairness built in. Donaldson points out, "I remember an AF4C championship [in 2000] where Benjamin, Seirawan and Shabalov tied for first and five minute games were used to break the tie. Joel and Yasser were blitz specialists and Shabba knew going in his chances were much smaller than in regular chess. The final result confirmed it."
Con -A nearly universal principle in sports is that their tiebreaks are at least in some way related to the sport in some abbreviated fashion. Back in 1983 pure chance was used for a World Championship Candidates Match between Smyslov and Hubner. The tie was broken with a spin of the roulette wheel! There was naturally much criticism of the "absurdity" of this solution. Imagine if the Internet discussion boards had been around then! Our own GM Susan Polgar lost a Women's World Championship Candidates Match to Nana Ioseliani in a similar way. The two rapid playoffs were split, but an arcane rule stated that the match would be decided by lottery- Ioseliani won only to be crushed by Xie Jun in the World Championship match.
Conclusion - The current USCF and FIDE rules make no mention of chance playoffs. It would seem to be an organizer's last resort - if there were no time for playoffs and if math tiebreaks do not resolve the tie (a rare possibility).
4) Full-Length Playoffs - "Full-length playoffs are the best solution if a tie must be broken. Organizers should set aside the time and money for a proper playoff if one is necessary."
Many Internet commentators stated that the only appropriate form of playoff is one that is played at about the same rate as the tournament itself, as compared to using rapid or blitz. "Blitz is different than regular slow chess," says TD Mike Atkins, an expert on blitz event organization. "Managing the clock is at least 50% of the game in blitz, a much higher percentage than slow chess, and why it is called Blitz. Just the name conjures up images of lightning strikes in war..."
Pro - In terms of fairness, a full-length playoff logically uses the same player skills that were being tested in the main event itself.
Con - Unfortunately, long playoffs have several practical difficulties:
· Long playoffs require an extra day, and the organizers and players to plan months in advance for that day. The organizer must also budget for it- money that might otherwise go toward the prize fund.
· Scheduling for players is a hassle if they don't know for sure when they will be able to leave an event - i.e., it is typical that they need to be back to their job the next day, etc.
· Trying to build it into the tournament itself is also awkward. At the 2008 U.S. Senior Open the organizers took an event that used to be a 6-SS, one round per day, and changed it to a 5-SS with the sixth day set aside for possible playoffs. As it turned out, the playoffs weren't needed. Thus one whole round was taken out of the main tournament, which could have given a player a chance to come back from an early draw to catch the leader.
Second, a long playoff is not a full solution. These long playoff games can also be tied - so then what happens? Further full-length playoffs could last for days. Very few events could abide that. In this situation you have to resort to other tiebreak methods anyway.
Conclusion - this option should be preferred in the rare cases when time and money allow.
5) Delayed Full-Length Playoffs - "A ‘Delayed' full-length playoff match would be ideal because it is easier to schedule efficiently."
USCF President Bill Goichberg has advocated this idea, which is that the tournament can be structured from the very beginning to generate two "finalists" who will play a full-length multiple-game playoff match at a later date (perhaps months later) in conjunction with another event.
Pro - The advantage (as compared to immediate full-length playoffs, above) is that only the two finalists would have to participate, and they would have time to make their plans to do so, while the organizers have no more expense because the match is in conjunction with another event that is happening anyway. So, all the logistical and budgeting problems are minimized.
Con - The main objection is that such a system still doesn't solve the problem under discussion - what to do if the match is tied?
And also what happens if players tie for the right to be one of these two finalists? That tie first has to be broken before the final match is played.
Conclusion - This system is preferable to a straight full-length playoff in many ways and might be a good option for some major events. But its utility is limited and thus is no more than a partial solution to be used only if rare circumstances allow.
6) Rapid or Blitz Playoffs - "Rapid or Blitz Playoffs can be used to break ties, rather than leaving it up to chance or tiebreak points."
Many official FIDE events use blitz for playoffs, in various formats. It is not unusual as many critics claim. Benjamin points out they were used in the old World Championship tournaments he was in: "I had three overtime matches: Zjvangtsev in 1996, Neilsen in 1998, and Baklan in 2001. All matches went down to a sudden death blitz game, and I lost all three. ... Even though losing was painful, I thought all the matches were very fair and actually quite rewarding, though grueling, to play." FIDE regulations give blitz playoffs as the only example. Blitz playoffs of various sorts have been used in U.S. Championships since the late 1990s when there was a movement away from having Co-Champions.
Pro - Blitz uses chess as the basis for the tiebreak, which arguably is more natural (and more exciting) from a sporting point of view than math tiebreaks, assuming full-length playoffs aren't possible. It also is more likely to determine a winner within a short amount of time. Greg Shahade points out that if players know that blitz playoffs may decide a tournament, they have ample notice to hone up on their blitz skills: " Why should Hikaru Nakamura not be rewarded for playing hundreds of hours of blitz while another GM ignores it, knowing full well that it may be used as a tiebreaker, as it has been for a long time now?"
Con - Rapid/blitz involves a different set of chess skills than tournament time controls. Thus its use in playoffs might skew the result in favor of certain types of players. Also, even blitz doesn't always determine a winner quickly because of the possibility of continuing ties. The Zatonskih-Krush playoff went through four rapid/blitz playoff games (nearly two hours) with the score still tied, which in fact is not uncommon.
Conclusion - GMs Benjamin and Seirawan agreed that most GMs prefer blitz to break ties. Certainly blitz playoff games could form the backbone of any playoff system when longer playoffs aren't possible, playing them in sets of two until there is a winner. However, organizers do not like the "open-ended" timetable as there is no way to predict how long the playoff will take. Thus, time constraints can motivate organizers to set a limit for the number of rapid/blitz games and "force an end" to the proceedings through the use of a sudden-death or Armageddon blitz (as in Benjamin's example above). See next section.
7) Hybrid Rapid/Blitz, with Armageddon or Sudden Death - "Rapid or Blitz Playoffs should be used, with a final Armageddon game to force an end to continuing ties if necessary."
This format is perhaps the most common for the frequent situations in which the organizers have limited time for a playoff (perhaps 2-4 hours). A series of rapid or blitz games are planned, in sets of two, and it is hoped that they will determine a winner. If not, they force an end to the playoff with either an Armageddon blitz game or a "sudden death" blitz series.
"Armageddon blitz" is played with white having a few minutes more than black (the exact time amounts vary, but 7 to 5 is typical), and black having draw odds. With no draw possible, it forces a winner.
"Sudden death" blitz (used previously in many FIDE World Championship events) is a regular blitz game in which a coin is flipped to see who gets white. The winner wins the playoff. If the game is drawn, another one is played until there is a winner. (This also forces a winner, although it can take many games if there are a lot of draws, and so it is less common than Armageddon)
Pro - A hybrid format has a big advantage that, with a known maximum time required for playoffs, it can be efficiently pre-scheduled by the organizer. The first set of games can be a relatively slow time control, going then to faster rapid or blitz games or a final Armageddon game only if necessary.
Hybrid formats seem to be well liked by top players such as GM Joel Benjamin, who said, "I would recommend more rapid games w/increment before getting to an Armageddon game. It would make sense to start the last round in the morning to ensure more time for a playoff."
Con - This was the format used for the 2008 U.S. Women's Championship, which came under great criticism. After a set of rapid games and a set of blitz games that were split, a final Armageddon game was played, and unfortunately it ended in a wild time scramble that many felt was ugly . Many people object to the Armageddon blitz on principle, from a purist standpoint. They also object that it is a blitz game, and all blitz games have the above-mentioned problems. As John Donaldson said, in such formats there seems to be "a tendency to sacrifice a little bit of the integrity of chess."
Conclusions - First, those who were turned off by the video of Zatonskih-Krush need to keep in mind one thing - are you objecting to the Armageddon format itself, or to the time-scramble? Time scrambles can be mitigated (and almost eliminated) using a delay on the clock. Historically, Armageddon blitz games have not included a delay on the clock. We followed that practice for this Zatonskih-Krush playoff. But this experience makes it clear that at least a 3- second delay should be included for Armageddon blitz, (and all playoff games), to help mitigate the unsatisfactory nature of the kind of worst-case-scenario time scramble on display in the video of that game. Such scrambles are okay for blitz tournaments, but not for playoff games. (Delay clocks are better than add-on clocks in this regard, by the way, as Goichberg points out. With add-on the players are still tempted to move fast in order to accumulate time, but not so with delay.) Delay can be included in blitz Armageddon games simply with an appropriate alteration in the uneven starting times for the Armageddon players, perhaps 6 to 4 instead of 7 to 5. Thus, the time scramble possibility inherent in Armageddon play is not a argument against its use, as it can be greatly reduced just as in blitz playoff games.
Use of delay clocks for Armageddon games is one improvement that should be written into the regulations, if nothing else.
As for the question of Armageddon versus "sudden death" play, it isn't known how all players feel about it. Their opinions most likely would be divided. Organizers clearly would prefer Armageddon, because sudden death still has an indeterminate end point.
On the whole, hybrid rapid/blitz playoff systems (with delay clocks) like this do work satisfactorily in the sense that they meet all the needs of the players, sponsors, and organizers. Although they are not perfect, and perhaps are in much need of some standardization, they provide an "exciting" form of playoff that the general public can relate to, while at the same time allow for efficient organization.
8) Full-Length "Short Formats" -
Is an ideal system that attempts to get the best of both worlds - long playoff games played in a known time frame possible? Some think so, but it requires a way to deal with draws.
A relatively recent and interesting new trend in playoffs are various proposals aimed at creating one- or two-game short playoffs that will force a winner- while fitting the time controls to the total amount of time available, thus making the playoff as close to tournament conditions as possible. Unlike the Hybrid systems outlined above, which might use the entire time available or only a fraction of it, these systems use all the time to advantage. There are three basic kinds of proposals along these lines, which I dub "Total Armageddon", the "Two-Game Sudden Death Playoff" and the "One Game Showdown."
Two-Game Sudden Death Playoff/One-Game Showdown
A very interesting system proposed by IM Larry Kaufman calls for a match of no more than two games that uses all the time available. Here is how Larry describes it:
"One normal game is played using half (or perhaps 60% of) the available time, shared equally (color chosen by lot). If it is decisive, that's it. If it ends in a draw, a second game is played in the remaining time (again, time shared equally), BUT WITH THE SAME COLORS. However this time black gets draw odds. I believe that this rule gives roughly equal chances to black and white. GMs I've asked about it don't agree on which color they would choose, which confirms that it is pretty fair."
A logical way to do it would be to allocate 50% of the time for the first game, 40% for the second game, and 10% for a short break between the two. Both games would have a 3-second delay. For example, if 3 hours total time were available, the first game would be G/45 (half of the time = 90 minutes, split evenly for the two players), and the second game (if needed) would be G/36, with an 18-minute break in between. If only one hour is available, the times would be G/15 and G/12 with a 6-minute break.
I asked Larry why he thought this system was theoretically fair. It seems at first glance the first game favors white but the second game favors black by a larger amount.
Larry replied, "It is fair because if White wins the first game, there is no second game. Here is the math: I'll assume 40% chance of draw, 35% chance of White win, 25% chance of Black win (this is roughly in line with experience in master chess, though the percentage of draws rises with the level of the players). White has a 35% chance to win outright in the first game, plus a (40% x 35%) chance to draw and then win, for a total of 49%. That's about as close to even as we can ask for. I admit that for a really high level game, say the playoff for the overall US Championship, Black may have some edge."
IM Greg Shahade proposes a creative tiebreak, the One Game Showdown which also limits the amount of time needed for the tiebreak. He suggests a simple 60-minute game with an increment. If one player wins, the tiebreak is over. A draw simply leads to a reversal of colors-the clocks remain unchanged.
For example, suppose both players start with 60 minutes (+ 3 second increment) for the first game. Let's say the game ends in a draw with Player A (white) having 30 minutes left and Player B (black) 20 minutes left. Now another game is played, players switching colors, with the clocks left unchanged. (B will have white now, with 20 minutes, A will have black with 30 minutes).... and so on until there is a winner. Since it is possible for a game to end in a draw with only seconds left on the players clocks, a minimum start time could be used - say 5 minutes.
Clearly the player who starts with White has some advantage: Greg has two solutions for this:
1. The player with better tiebreaks begins with White so that the imbalance is dictated by the tournament standing.
2. Use a bidding system so the player with Black starts with more time (similar to the "Total Armageddon" system).
Pro - The Two-Game Sudden Death playoff has many advantages - it allows for two fairly long games (rather than shorter rapid or blitz games), and it forces a winner in the allotted time. If Kaufman's estimate of the balance is true, this system seems fairer than Armageddon systems (see below), and since the games are longer it minimizes the edge that blitz specialists have in rapid/blitz playoffs. It works for any amount of time available, although maybe one hour should be considered a minimum.
Note that it need not necessarily be used by itself but could also be combined in hybrid form with other systems. For example, if a set of two regular rapid games doesn't break the tie, the remaining time could be used for a two-game sudden death set.
IM Greg Shahade's idea arose toward the end of the publication of this article, so the GMs and TDs interviewed did not get a chance to comment on it. It has many of the advantages but may need some testing and tweaking.
Con - GM Joel Benjamin said the Two Game Sudden Death Playoff sounded gimmicky but that "It probably is fair, and if other players like it, that's fine by me." U.S. Chess League commissioner IM Greg Shahade disagreed with Larry's idea on principle, "I don't think an organizer should be the judge of what type of system creates a fair chance for both players. If one player feels that having White in the mentioned system is only 45% then that sucks that they should be forced to have it just because a poll decided it was close to a 50-50 proposition. As much power as possible should be put in the players hands when it comes to controlling their destiny."
Conclusion - The Two-Game Sudden Death format is worth a try for playoff situations, especially those who wish to avoid blitz. The One-Game Showdown idea needs further trouble-shooting.
Total Armageddon, (or Long Armageddon)
This format was originated by Don Shultz (as far as I've been able to determine) and was specified for use at this year's U.S. Senior Open (although it wasn't actually used there, as there was no tie for first). It basically uses the total amount of time available to play one big Armageddon game.
The details of how it might work could be tweaked in various ways. One approach is to give white and black proportional starting times on the clock (plus a 3-second delay). For example, if the total time available was 3 hours, white could be assigned half of that (90 minutes) and black one third of that (60 minutes), with black having draw odds in return for the time deficit. If players determine these ratios aren't balanced, they could either come up with better ones or they could use a bidding system (as Shultz originally planned): White will get half the available time (90 minutes in this case), and the two players then bid for how much time they would be willing to start with in order to play black and have draw odds. Let's say the bid got down to 64 minutes. The game would then be 90 minutes for white, 64 for black (plus draw odds). (It is possible to have the bidding be for who will play White, which is how Larry originally told it to me and how I explained it in the Chess Life article I wrote on the U.S. Championship, due out in August. But bidding for black is better.)
Pro - The system fits into the amount of time available, while generating the longest-possible game for that time.
Con - Players are naturally skeptical about long Armageddon play, and not just because it is untried. For instance, Joel Benjamin said, "I'm speaking only for myself, and other grandmasters may feel completely differently. But my feeling is, if I'm in a playoff for something important like the U.S. Championship title, I would want to play something that most closely resembles chess. To me, a game where players have unequal amounts of time and one player gets draw odds is far more bizarre than a thirty-minute game. I think Armageddon is a gimmick that should be a last resort and not a first resort." He adds that tired GMs facing a tough playoff would not like the bidding concept, even if it theoretically creates a fair situation. IM Greg Shahade agreed with Joel's line of reasoning, calling playing with draw odds "sort of anti-chess."
Conclusion - This idea didn't get much traction among the players I talked to. I personally think it would be worth a try, but at the least it needs refinement and testing before it is used in a major championship. It would be interesting to have an entire tournament played using this format, in order to get some experience with it. A big advantage is that such a format would lend itself very nicely to Wimbledon-style Elimination tournaments - a type of tournament very popular with the general public, but which normally doesn't work for chess due to the possibility of draws. But in this case, it would work very well.
Multi-Player Playoffs - One final issue: A weakness of "playoff" forms of tiebreak is that they have difficulty with ties involving 3 or more players. If time is available, such playoffs could take the form of a round robin between the tied players, as has been done in a few recent events, such as the 2003 U.S. Women's Championship that had a 3-way blitz tournament playoff. However, a simpler approach would be to use math tiebreaks to determine a final two for a playoff, and my respondents had no problem with that.
So, in view of the conclusions in the above analysis, a logical proposal for chess playoffs would be as follows:
Proposal for Chess Regulations-
Guidelines for Dealing with Ties
· A tie for first place can be an acceptable result for most chess tournaments. If it is a championship event, tied players may be considered co-champions.
· However, if for any reason there is a desire to determine a sole winner among tied players, the following guidelines are recommended:
o The organizers should publicize in advance the method of breaking ties that will be used.
o If there is a sufficient time, a full-length playoff using time controls comparable to those used in the main tournament itself may be used. Two-game mini-matches are played at this time control to determine a winner. If the score is still tied, more sets of such games can be played or one of the methods below can be used. Such playoffs might require an extra day in the schedule. (If feasible, the playoff could be delayed and held concurrent with a subsequent tournament.)
o If there is SOME time allocated for a playoff, (ideally at least 2-5 hours, but a minimum of 1 hour), there are two options for abbreviated playoffs:
a. Rapid/Blitz Playoffs - A series of two-game sets played at pre-announced rapid or blitz time controls. All such games should have at least a 3-second delay on the clock (delay, not add-on increment). If the first set is split, the next set is played, etc., until there is a winner. Ideally, organizers will schedule as many such sets as they can in the allotted time, perhaps starting with longer time controls and getting quicker in subsequent sets. A final playoff game should be scheduled to end the possibility of a continuing tie exceeding the total time available - using either an Armageddon blitz game or a sudden death series.
· A typical Armageddon blitz game is white 6 minutes (+ 3 second delay), black 4 minutes + draw odds (+ 3 second delay).
· A typical sudden death game would be 5 minutes each (+ 3 second delay). If the game is drawn another sudden death game is immediately played, switching colors, and so on until there is a winner.
b. Two-Game Sudden Death - The entire time available for the playoff is used to play no more than two playoff games. One player is given white (by any method of chance). If the first game is decisive, the winner wins the playoff. If it is drawn, another game is played, using the same colors, in which black will have draw odds. The time allotment for the games is relative to the total amount of time available for the playoff, using this formula:
· 50% of the time is used for the first game (divided evenly between the two players, rounded up to the nearest minute). Players will also have a 3 second delay on the clock.
· If necessary, after a short break of 10% of the time, the remaining 40% is used for the final game, again divided evenly, with a 3 second delay. (The organizer may adjust these percentages as required.)
· For example, if 3 total hours are available, the first game would be G/45 (+3), and after an 18-minute break, the second game would be G/36 (+3).
o If there are more than two players involved in the tie, mathematical tiebreaks (see below) can be used to reduce the players to two for a playoff, or if time allows a round-robin rapid/blitz mini-tournament can be played between the tied players (breaking further ties if necessary).
o If there is NO time for any playoffs, mathematical tiebreak systems should be used to determine a winner in a Swiss-System or Round-Robin event. The order of tiebreaks should be announced in advance. The recommended tiebreak methods are as follows: [refer to existing recommendations already in the rulebook].
This system is straightforward and helpful to organizers, while avoiding inflexibility. Notice how it uses the golf approach as a model - providing options according to circumstance. An organizer can use these options to decide how to schedule his event early in the planning stage.
Although the details here could be nitpicked by rules lawyers, the basic point is that a set of clear-but-flexible guidelines like this is useful and should be integrated into the chess regulations. I happen to think this proposal would work very well, but if the rule committees see fit to alter it, that is fine. The USCF and FIDE rule committees should seriously consider the proposal.
Remember also it is not set in stone. If further experience with long Armageddon games proves them to be fair and player-friendly, they could be easily added to the regulations. Similarly, if testing of the new Two-Game Sudden Death system shows problems, it could be either refined or dropped. The guidelines may evolve over time in many ways.
Conclusion - The Value of Guidelines and Consistency
One organizer told me the attempt to create playoff guidelines is useless, because in the end it all must come down to what the sponsor wants. But as long as the guidelines are "recommendations" and have flexibility, they can be very useful to help an organizer with planning.
There is also value in having some consistency and standardization among the systems that we have in use for chess events. An outside observer looking at international chess and the myriad tournament forms, time controls, and championship systems we currently use might easily conclude it is all chaos. The variety of tiebreak systems one sees in these events, and total lack of guidelines in FIDE or USCF rules, serves only to increase this feeling of anarchy. Having guidelines to provide more consistency would help reduce that appearance to the public and would be to the advantage of the sport of chess.
Tom Braunlich's article on the U.S. Championship appears in the August Chess Life which will be available on uschess.org tomorrow. Look for responses and feedback on this article, including possible reactions from delegates and committees at the U.S. Open next week (Dallas, August 2-11.)