Dropping down the wrong player from a score group

[ILfish]

Players need to know more that just chapter 1 of the rule book. A case in point was when as a player with 2.5 out of 3 I was the lowest rated in the 2.5 point group. Rather than me being paired with a player with a 2.0 score, another player was dropped down & I was paired against my brother instead. It was one of the few tournaments at the time [late 80s] where I did not have my copy of the rule book with me. I could not convince the TD that they had made an incorrect pairing, because I could not cite/point out in the rule book what was wrong about the pairing that was made.

Larry S. Cohen

[jwiewel]

@ILfish

Players need to know more that just chapter 1 of the rule book. A case in point was when as a player with 2.5 out of 3 I was the lowest rated in the 2.5 point group. Rather than me being paired with a player with a 2.0 score, another player was dropped down & I was paired against my brother instead. It was one of the few tournaments at the time [late 80s] where I did not have my copy of the rule book with me. I could not convince the TD that they had made an incorrect pairing, because I could not cite/point out in the rule book what was wrong about the pairing that was made.

Larry S. Cohen

If dropping the next to lowest (instead of the lowest) is required to even out all the colors then I’d generally do that as well if the ratings were close enough to justify it. Pairing two brothers would be a possible exception to “generally” but I have paired two brothers before the final round when they were strong enough to be likely to meet in the final round anyway (that way there is less perception of collusion in the final round result).

@GKarRacer

Considering how many club TDs I’ve seen have probably never even seen a copy of the rule book let alone read it, you’d think the answer to what you need to read would be “none”. But, that’s a different issue. Practically speaking the 3 chapters available online will be enough for him to get by for the time being, especially for small tournaments and doing pairings only by computer.

At some point you will want to read the rest of it, so that you fully understand the pairing rules, cash prizes, etc. This is more important for larger tournaments and for advancement beyond club TD.

A lot of TDs quite a bit higher than Club TD would be at a loss if trying to pair without using a computer.

There is a fair amount of mis-perception about the pairing rules. I’ve seen people assume that pairings were supposed to be 1vs2, like NCAA March Madness pairings in each scoregroup, use tie-breaks ahead of using ratings, are random within a scoregroup, etc.

[Smythe_Dakota]

@ILfish

… A case in point was when as a player with 2.5 out of 3 I was the lowest rated in the 2.5 point group. Rather than me being paired with a player with a 2.0 score, another player was dropped down …

@In response

If dropping the next to lowest (instead of the lowest) is required to even out all the colors then I’d generally do that as well if the ratings were close enough to justify it. …

The possibility of downfloating a player other than the lowest is entangled in one of those rules that Kevin Bachler would probably call “schizophrenic”. And in this case I agree with that description:

29D1a. In the case of an odd number of players, the lowest-rated player, but not an unrated player, is ordinarily [emphasis mine][b] treated as the odd player and is paired with the highest-rated player he or she can play in the next lower group. Care must be taken in doing this that the odd player can be paired in the next score group, that the remaining members of both affected score groups can be paired with each other, that the odd player has not played all the members of the next lower group, and that the color consequences are acceptable.

29D1b. If the conditions in 29D1a cannot be met, then try treating the next lowest-rated player as the odd player, or pairing the odd player with a lower-ranking player in the next score group. … However, switches to correct colors should stay within the appropriate limits (29E5) …

29D2. … The odd player is normally paired with the highest-rated player he or she has not met from the next lower group. It is acceptable to pair the player against a somewhat lower-rated player to equalize or alternate colors, but only within the rules for transposition as explained below.[/b]

And then:

[b]29E5a. The 80-point rule. Transpositions and interchanges for the purpose of maximizing the number of players who receive their due color should be limited to players with a pre-tournament rating difference of 80 points or less.

29E5b. The 200-point rule. Transpositions and interchanges for the purpose of minimizing the number of players who receive one color two or more times than the other color should be limited to players with a pre-tournament rating difference of 200 points or less.[/b]

At the end of 29D1a, the phrase “and the color consequences are acceptable” seems ambiguous. What constitutes “unacceptable” color consequences? Is it:

• (a) Only a major color disaster, such as assigning somebody 3 blacks in a row? Or
• (b) any set of color assignments that could be improved overall by downfloating the second-lowest player instead of the lowest?

If you say (a), you are a strict constructionist, like Larry Cohen (ILfish). If you say (b), you are looking a little deeper, like Jeff Wiewel (jwiewel).

To me, (a) presents several problems, one of which is asymmetry. You can transpose the downfloater only to avoid serious problems, but you can transpose the upfloater simply to improve colors overall (see 29D2 above). Why shouldn’t the same standards apply to both?

29D1 and 29D2 are a big mess, in many ways. Better would be:

Proposed 29D1. If there is an odd number of players in a score group, a player at or near the bottom of the group, but not an unrated player, is paired with a player at or near the top of the next lower score group. In general, the lowest-rated player in the group should be paired with the highest-rated player in the next lower group, but minor transpositions may be made with either or both of these players for all the usual reasons, such as to avoid pairing any player against an opponent he has already faced, or to improve colors overall. Such a transposition made to improve colors should respect the 80- and 200-point limits set forth in 29E5a and 29E5b.

What do you think of this idea, Baba Looey?

Bill Smythe

[timjust]

@Smythe Dakota

@ILfish

… A case in point was when as a player with 2.5 out of 3 I was the lowest rated in the 2.5 point group. Rather than me being paired with a player with a 2.0 score, another player was dropped down …

@In response

If dropping the next to lowest (instead of the lowest) is required to even out all the colors then I’d generally do that as well if the ratings were close enough to justify it. …

The possibility of downfloating a player other than the lowest is entangled in one of those rules that Kevin Bachler would probably call “schizophrenic”. And in this case I agree with that description:

29D1a. In the case of an odd number of players, the lowest-rated player, but not an unrated player, is ordinarily [emphasis mine][b] treated as the odd player and is paired with the highest-rated player he or she can play in the next lower group. Care must be taken in doing this that the odd player can be paired in the next score group, that the remaining members of both affected score groups can be paired with each other, that the odd player has not played all the members of the next lower group, and that the color consequences are acceptable.

29D1b. If the conditions in 29D1a cannot be met, then try treating the next lowest-rated player as the odd player, or pairing the odd player with a lower-ranking player in the next score group. … However, switches to correct colors should stay within the appropriate limits (29E5) …

29D2. … The odd player is normally paired with the highest-rated player he or she has not met from the next lower group. It is acceptable to pair the player against a somewhat lower-rated player to equalize or alternate colors, but only within the rules for transposition as explained below.[/b]

And then:

[b]29E5a. The 80-point rule. Transpositions and interchanges for the purpose of maximizing the number of players who receive their due color should be limited to players with a pre-tournament rating difference of 80 points or less.

29E5b. The 200-point rule. Transpositions and interchanges for the purpose of minimizing the number of players who receive one color two or more times than the other color should be limited to players with a pre-tournament rating difference of 200 points or less.[/b]

At the end of 29D1a, the phrase “and the color consequences are acceptable” seems ambiguous. What constitutes “unacceptable” color consequences? Is it:

• (a) Only a major color disaster, such as assigning somebody 3 blacks in a row? Or
• (b) any set of color assignments that could be improved overall by downfloating the second-lowest player instead of the lowest?

If you say (a), you are a strict constructionist, like Larry Cohen (ILfish). If you say (b), you are looking a little deeper, like Jeff Wiewel (jwiewel).

To me, (a) presents several problems, one of which is asymmetry. You can transpose the downfloater only to avoid serious problems, but you can transpose the upfloater simply to improve colors overall (see 29D2 above). Why shouldn’t the same standards apply to both?

29D1 and 29D2 are a big mess, in many ways. Better would be:

Proposed 29D1. If there is an odd number of players in a score group, a player at or near the bottom of the group, but not an unrated player, is paired with a player at or near the top of the next lower score group. In general, the lowest-rated player in the group should be paired with the highest-rated player in the next lower group, but minor transpositions may be made with either or both of these players for all the usual reasons, such as to avoid pairing any player against an opponent he has already faced, or to improve colors overall. Such a transposition made to improve colors should respect the 80- and 200-point limits set forth in 29E5a and 29E5b.

What do you think of this idea, Baba Looey?

Bill Smythe

ADM time.

[BobMessenger]

This topic is a spin-off from 7th Edition Rulebook Status? in the US Chess Issues forum.

[nolan]

The deadline for submitting ADMs to the office (they go to Jennifer Pearson this year) is June 3rd.

[ESchlich]

As long as a change is being considered, should limiting the number of up or down floaters a player receives also be considered in the change? I would find it interesting to learn how Tom Doan and Thad Suits implement the odd man rules. That information might help write a better rule if we agree with the implementation.

I am not a fan of the FIDE pairing rules but their philosophy that the rules should be clear enough that all TDs would do the same pairing is good.
Regards, Ernie.

[ESchlich]

I forgot to ask the following in my previous post. Should up/down floater info be a consideration if you want or need to change the lowest upper vs highest lower score group player.

[nolan]

Why should what the TD wants have anything to do with pairing? In more than a few ways, FIDE has the right idea with pairings that are far more deterministic, meaning a TD’s wants or personal preferences cannot enter into things. Yeah, there are times it produces pairings that some people don’t like, but at least it does so consistently.

[ILfish]

I also recall a 4 round 1 day event where after the first round there were an odd number in the 1 point group. There were 2 players rated in the 1400s that had scored upset victories, but neither were dropped down to the top 1/2 score group player [an expert]. Rather a 1700 rated player was dropped down due to the fact that both 1400s had a color conflict with the expert. It was many years ago, and I [tournament spectator] do not recall what pairing program was used at the event. I did ask the TD and was told that the pairings [1700 dropped to play Expert in lower score group] were due to colors. End of story is that the 1400s both lost their round 2 games and the 1700 beat the expert. The obvious question is why (even back in the 80s) was the 200 point transposition rule not enforced? Also, why did the TD not overrule the computer?

Larry S. Cohen

[nolan]

If it was in the 1980’s pairing programs were still not in common use. They didn’t start to take over most pairing duties until the early 1990’s.

[jwiewel]

@nolan

If it was in the 1980’s pairing programs were still not in common use. They didn’t start to take over most pairing duties until the early 1990’s.

Trying to analyze what might have happened more than 35 years ago, and without a wall chart, is “a bit difficult” but Harkness odd-man pairings (drop the middle) were still a valid variation back then.

[nolan]

It’s not that easy to reproduce pairing from an event held last week, either, because you just don’t know everything the TD (or the computer) knew at the time, or what settings the computer was using.

With FIDE pairing, you always use the published ratings of the players, not their most recent event. I suspect these days US Chess rated tournaments are about 50-50 as to whether they’re using published ratings or latest event ratings, at least for scholastic events.

[jwiewel]

@nolan

…I suspect these days US Chess rated tournaments are about 50-50 as to whether they’re using published ratings or latest event ratings, at least for scholastic events.

Even if you are using the published ratings from the current supplement there can still be cases where some of the players will legitimately end up with a different rating assigned.

[Smythe_Dakota]

@ESchlich

As long as a change is being considered, should limiting the number of up or down floaters a player receives also be considered in the change? …

FIDE already forbids (or at least strongly discourages) downfloating (or upfloating) the same player in two consecutive rounds. (I guess a downfloat in one round followed by an upfloat in the next round would still be OK.)

In the olden days, I would write a down-arrow or up-arrow on the pairing card, next to the round number. That way I could avoid subsequent consecutive downfloats or upfloats if I wanted to.

But it usually wasn’t necessary to avoid consecutive floats. My feeling was, that if the float pairing vindicated itself (i.e. the higher-rated but lower-scoring player defeated the lower-rated but higher-scoring player) then it was not necessary to avoid giving either of those players another float in the next round.

… I would find it interesting to learn how Tom Doan and Thad Suits implement the odd man rules. …

My guess is that WinTD is more likely than SwisSys to transpose a floated player.

… Should up/down floater info be a consideration if you want or need to change the lowest upper vs highest lower score group player.

As above, yes, I think it should.

Bill Smythe

[Smythe_Dakota]

@ILfish

I also recall a 4 round 1 day event where after the first round there were an odd number in the 1 point group. There were 2 players rated in the 1400s that had scored upset victories, but neither were dropped down to the top 1/2 score group player [an expert]. Rather a 1700 rated player was dropped down due to the fact that both 1400s had a color conflict with the expert. …

I assume that the 1700 and the two 1400s were the bottom three players in their score group, but as others have pointed out, it is pretty much impossible to figure out whether the pairing is correct without a complete crosstable. It does indeed sound like a 200-point violation, but who knows for sure?

If I had been TD’ing, and I ran into a situation like that, where there was a color problem one way and a 200-point violation the other way, the next thing I might try would be to upfloat somebody other than the Expert. Perhaps the second- or third-highest player could have been upfloated against the 1400 downfloater, giving both their due colors without any 200-point violations.

Obviously, some TDs are insufficiently willing to transpose upfloaters and/or downfloaters to avoid color problems. The rule should make it clear that transpositions involving floaters are OK.

Bill Smythe

[wintdoan]

@ESchlich

I forgot to ask the following in my previous post. Should up/down floater info be a consideration if you want or need to change the lowest upper vs highest lower score group player.

WinTD has a “keep in score group” option for a single player to try to prevent floats. The main intended use for that is where a player gets the forced bye, gets a game on (say) a cross round pairing, loses and thus will be likely to get the forced bye the next round as well. Other than that, there’s no provision for avoiding repeated up/downfloats in the US Chess rules side of the pairings.

[wintdoan]

@Smythe Dakota

@ILfish

I also recall a 4 round 1 day event where after the first round there were an odd number in the 1 point group. There were 2 players rated in the 1400s that had scored upset victories, but neither were dropped down to the top 1/2 score group player [an expert]. Rather a 1700 rated player was dropped down due to the fact that both 1400s had a color conflict with the expert. …

I assume that the 1700 and the two 1400s were the bottom three players in their score group, but as others have pointed out, it is pretty much impossible to figure out whether the pairing is correct without a complete crosstable. It does indeed sound like a 200-point violation, but who knows for sure?

If I had been TD’ing, and I ran into a situation like that, where there was a color problem one way and a 200-point violation the other way, the next thing I might try would be to upfloat somebody other than the Expert. Perhaps the second- or third-highest player could have been upfloated against the 1400 downfloater, giving both their due colors without any 200-point violations.

Obviously, some TDs are insufficiently willing to transpose upfloaters and/or downfloaters to avoid color problems. The rule should make it clear that transpositions involving floaters are OK.

Bill Smythe

The rulebook (through v5) had a rather bad example of doing a transposition involving a float where the rating of the opponent being floated was compared with the natural pairing in the original score group. I pointed out (during the 5th edition revision process) that that wasn’t correct—unlike a transposition within a single score group, you can’t look at both sides of the pairing when contemplating a change in the float or you can get some really bad pairings. Perhaps this was a case of someone taking that example as showing how to handle this and getting the really bad pairing as a result.

[wintdoan]

@Smythe Dakota

@ESchlich

As long as a change is being considered, should limiting the number of up or down floaters a player receives also be considered in the change? …

FIDE already forbids (or at least strongly discourages) downfloating (or upfloating) the same player in two consecutive rounds. (I guess a downfloat in one round followed by an upfloat in the next round would still be OK.)

That part of the rule is a bit clumsy. It does a two round lookback. Avoiding repeating the type of float from the previous round takes precedence over avoiding repeating the type of float from two rounds earlier takes precedence over color and position in the score group. However, an upfloat two rounds earlier and a downfloat last round doesn’t cancel the effect of the upfloat which I found rather weird.

[wintdoan]

@nolan

Why should what the TD wants have anything to do with pairing? In more than a few ways, FIDE has the right idea with pairings that are far more deterministic, meaning a TD’s wants or personal preferences cannot enter into things. Yeah, there are times it produces pairings that some people don’t like, but at least it does so consistently.

Unfortunately, the FIDE pairing “rules” are actually “what JaVaFo does”. The most recent rewrite of the pairing rules attempted to replace the description of JaVaFo’s algorithm with a statement of principles. Unfortunately, it doesn’t match the actual behavior of JaVaFo when it comes to messy backtracking situations—I have three different simulated tournaments where the WinTD implementation produces “better” pairings according to the rules as currently written than JaVaFo does (and that’s 3 out of 3 in tournaments with around 20 players and 10 rounds).