Greetings again and thanks in advance for the help.
Was surprised for a minute when my pairing program dropped a different odd man down from the top score group, but of course, after reviewing the rules - this is perfectly acceptable within the 80 point guideline for color alternation. But as I was reviewing the next score group - I was shocked when it did NOT drop a different odd man when it seemed called for within the 200 point guideline for equalization. Can someone take a minute to explain?
(clearly, the first 1813 player can be transposed with 1825 for color alternation on boards 1-2)
(and also clearly - dropping a player due black instead of white will help with color alternation -also because next score group top player is due white)
So far so good after 2 transpositions.
Score group 1.5:
1895 WB (paired against oddman)
1718 -B
1698 B-
1636 BW
1089 -B
1052 -W
The top player in the next score group has not played yet (2 half-point byes) and has 14 other members - 7 people due white and 7 due black.
So… here it seems as obvious as the first case. It is going to be beneficial to drop a player due white. The natural pairings are going to end up with an equalization problem:
1718 vs 1636 (ok)
1698 vs 1089 ( bad, leaves the 1089 with -BB)
Within 200 points, I should transpose 1052 with 1089 and drop a different oddman in this case as well, no?
I am leaning towards overriding the computer and releasing the pairings with that adjustment in about 5 hours. But I dont really want to do that because I’m going to start doubting everything the computer does - its a slippery slope!
As soon as I hit submit - I noticed a seemingly minor mistake in my pairing logic - and maybe it explains the difference in the pairing logic used by the computer.
This line:
1698 vs 1089 ( bad, leaves the 1089 with -BB)
should have read: 1089 vs. 1698 ( bad, leaves the 1698 with B-B)
because if I understand the rules correctly -B would get W before B-
So! now what does that change? … pairing a different oddman doesnt help the 1089’s color equalization as I stated, it would now only help the 1698’s color equalization!
And even though this doesnt seem logical – perhaps that is the issue?
So, now I guess I’m leaning toward receiving input on this issue - as perhaps I am misunderstanding the rules here I believe the swap between 1052 and 1089 would still be warrented in this case.
Do you have the 1052 and the top 1-1 on the same team or club? i.e. is there some reason they might be flagged as avoiding getting paired against each other if there is a reasonable way to do so?
Was the 1052 already paired down to a lower score-group and you have a FIDE-like setting that avoids back-to-back drops?
Giving B to Bx is almost as bad as giving B to xB, so it seems that pairing 1718 vs 1636, and 1698 vs 1052, and dropping 1089, as you suggested, would be indicated.
By any chance has 1718 already played 1636? That, of course, would create a major monkey wrench.
Also, what is the effect in the next score group down? You said there was 1 neutral, 7 due W, and 7 due B, so there shouldn’t be a problem – you just declare the neutral player to be either due-W or due-B, depending which way makes the colors better. If, however, there is some sort of problem with that score group, e.g. several players having already played each other, then maybe you would (somehow) end up with bad colors there too. If so, then the pairing program might balk at pairing down the 1089, if the net effect would be only to push the bad colors down to a lower score group. (This whole scenario seems unlikely, though.)
It might help if we could see the entire crosstable, including opponents and colors. Also, which pairing program are you using? And what pairing restrictions (father-son, brother-sister, teams, etc) are you facing?
At this point I think all of us are scratching our heads as to why the program wouldn’t pair down the 1089, as you suggested.
In a draw score group, who can play whom can be a factor even in an early round.
Also, what happens in the 1.0 group? Changing the downfloat could be an “out-of-the-fire-into-the-frying-pan” situation. As Bill Smythe suggests, a full wall chart table would be helpful.
Ok - this should answer the questions, but I will put up a wall chart with current pairings included.
There are no restrictions on the players involved. None of the players have played each other.
Well, the 1089 and 1052 are siblings - but that is irrelevant here - im not trying to pair them and they are in the bottom half of the score group.
The question about a FIDE setting to avoid back-to-back dropping doesn’t apply as the program is suggesting dropping the 1052. Neither 1052 or 1089 had to drop previous round (there were and even number of players who had taken byes in round 1)
1718 vs 1636 have not played - the computer (as well as me) suggests this pairing.
As mentioned in the OP - the 1.0 score group looks nice and clean. A neutral at the top and a even number of B and W throughout the rest of the group. And from what I can see - none of the 14 players with 1.0 out of 2 had played each other. I’d agree with “this whole scenario seems unlikely”.
I am using Swiss Sys 9.
The following URL points at the wall chart including the computer pairings. You might notice that I neglected to include myself in the 1.5 score group when describing the scenario- I am taking a bye to avoid giving a bye to one of the players.
Current pairings in question (#5 is the neutral player in 1.0 score group):
12 vs 17
26 vs 14
5 vs 27
My interpretation (drop #26 as odd-man instead of #27):
12 vs 17 (no change)
14 vs 27
26 vs 5
(Sorry. This is all wrong—I missed the fact that the second high rated player with two halves had withdrawn).
It’s possible that it’s not properly taking into account the two players who have no color. If you take out the bottom 1.0 (that downfloat is effectively forced since the top 0.5 is due B and is over 400 points above the second 0.5), then the 1.0’s have an excess of players due White, so you would (ordinarily) prefer to downfloat a player due Black. However, with two players with xx, you can get colors correct in the 1.0 score group no matter which player you downfloat so picking the one that improves colors in the 1.5’s is the better pairing. I see no reason not to override the program in this case.
WinTD would drop the 1089. Although the colors balance in the 1.0’s, you can’t fix all the problems because there’s a big gap between the 3rd and 4th boards on both sides, so you end up with two alternation errors. But the pairings from 1.0 down aren’t affected by the choice on the 1.5 downfloat.
Yes - even though I said the 1.0 group has 7 whites and 7 blacks - i did notice that i had to pair a few with alternation problems due to the rating jumps - but as you point out - the drop above does not impact.
So to close, I will override the computer… and slip down that slippery slope that leads me to question all of its decisions
This is just plain bizarre. I did the pairings by hand, and barring restrictions I don’t know about (such as teammates or other forced “do not pair these players”), I can’t possibly understand why SwissSys treated player 27 instead of 26 as the odd man in the 1.5 score group.
Please check that you are running the newest version of SwissSys (currently 9.33). If so (or if you upgrade to version 9.33 and SwissSys still makes these pairings), please take a few minutes to report this to Thad Suits (email suits at initco dot net). (You should include the tournament files with your bug report.) Thad is usually very good about responding to such reports.
I do know from using SwissSys the past 11 years that most of the pairing oddness seen these days involves the choice of odd player. SwissSys has gotten pretty good at pairing score groups with even numbers of players
(I can’t prove this, but my intuition from looking at weird pairings is that SwissSys pairs score groups from the top down, and once SwissSys is done with a score group with an odd number of players, it doesn’t seem to be willing to go back and revisit its choice of odd man if it would improve the situation in the next lower score group. It seems to set the completed score groups aside and doesn’t look back. In that sense, it does pair like some human TDs )
About your comment about your intuition - in this case it DID drop a different oddman (from the top 2.0 score group) when there appeared to be no impact for the next score group (1.5). It seemed to do this to improve the color of the 2.0 score group - (and an improvement for the dropped player as well!). If I click on the pairing logic button - it calls this exchange “better oddman for color or rating” in the reason/color coding. I don’t know what this means exactly within the software, I assume it to be “color” and specifically color alternation.
The surprise was why it didnt come to the same conclusion for the 1.5 group when there appeared to be an even better reason to apply “better oddman for color or rating” - specifically color equalization - within the 1.5 group itself.
Im on 9.2 i think… maybe 9.3 - I will update to latest version and send Thad the files.
It’s possible SwisSys was unwilling to trade one bad equalization for three or four bad alternations – even though that isn’t really what would have happened ultimately.
I said earlier that, as a human TD, I would be reluctant to make a transposition that merely pushes the bad colors down to a lower score group.
In this case, though, I would enthusiastically make this transposition, because in pushing the bad colors down it also replaces a bad equalization with bad alternations.
I’ve never been a huge fan of getting all het up over bad alternations anyway – especially in a small tournament because of the “camp” effect. However, 44 players is hardly “small”.
Other than to look up the number of players, I haven’t really examined your crosstable yet. But by now maybe everybody has all the issues figured out anyway.
Except that it doesn’t improve colors in the next score group. The upfloat from the 1.0’s has two half-byes, so can freely take either color. The alternation errors in the 1.0 score group are because of a large rating gap in the middle on both top and bottom halves.
One trick I learned at the Metrowest Chess Club is that, in a pinch, one can take a printed wall chart (no matter how it is sorted) and use scissors to cut strips for each player to make flimsy “pairing cards.” They’re not terribly convenient to work with, but it is possible.
When the club had a permanent cork board allocated to it in our meeting space, at the end of each week’s round we would print standings and crosstables and pin them to the board. While they were there for the players’ reference, they were also a last-case emergency backup for the club. If both of the club’s laptops failed, we could use the paper wallcharts to do the pairings as described above. We also made sure to print the wallchart showing all rounds in the event, including as yet unplayed rounds. That would guarantee we had a record of the advance bye requests.
I’ve started running the regular events at the Pittsburgh Chess Club again. We keep a printed copy of current standings/pairings on a wall of the meeting room. We also publish the same on our website. (Both sets of wallcharts show advance byes.) I pair the events on my Surface, so the files are there and automatically backed up in my OneDrive space as well. As a final backup, my DropBox gets the current SwissSys file after every round too (if, for some reason, my Surface dies, I can still access the file online).
I relate all this stuff solely to second what Mr. Ballou has said above. In a pinch, one quickly discovers it’s impossible to have too many fail-safes…especially if disaster strikes while tinkering with a file to improve pairings.
