Swiss Pairing Question with Swiss Sys

Running Chess Tournaments
1

During my first USCF tournament (23 players, spread out over 12 weeks), I did all the pairings by hand using a spreadsheet - following the rules as outlined in 6th edition.

I then followed up with Swiss Sys. I did this for 2 reasons - I don’t trust a computer until I know how it works. I don’t trust myself… most times :slight_smile: So with the exception of the random ordering of Unrated players (I chose a different order/pairing number than Swiss Sys), the pairings matched. I took this as validation I knew what I was doing and I could trust the computer.

Earlier this year I had a small 5-round scholastic end with an angry parent (surprise!) because one of the players ( a young girl rated 529) seemed to have an easy route to a larger trophy. Where as his child (rated 1075) had to play the top seeds. Against my better judgement, I engaged with the parent and explained “sometimes it happens.” He persisted and I explained that when you lose your first round match - you can sometimes get 4.0/5 without having a single game against the top seeds. He told me that I didn’t know what I was talking about, the pairings were wrong, and in all the tournaments he had ever taken his son to, he had never seen anything like this.

So - I said I would check and after the tourn I checked. I repaired the entire tournament by hand. In 4 out of the 5 rounds, I came to the same pairing as SwissSys - but there was a pretty big discrepancy in round 2 that saved this girl from playing a strong player. I would like a little input:

Round 2 score groups:

Score   Player  Rating  Color
1.0        A        1570    B
1.0        B        1570    W
1.0        C        1343    B
1.0        E        1242     W
1.0        F         1075    B
1.0        H          712     W
1.0        I          634     B
1.0        J         591     W
1.0        L         500     W
1.0        M          446      B
1.0        V         UNR     W
1.0        W          UNR     X
0.5        D          1309    H
0.5        G          865    H
0.0        K           529    B
0.0        N           286   W 
0.0        O           243  W
0.0        P           231   B
0.0        Q           201   B
0.0        R            191  W
0.0        S            Unr   B
0.0        T            Unr   W
0.0        U            Unr   W
0.0        X            Unr   B
0.0        Y            Unr   W

Note that A and B are brothers and I wanted to avoid pairing them unless +2.
Note that D and G are brothers who took 1/2 point bye in round 1 - I also wanted to avoid pairing them unless +2.

The 1.0 score group paired as expected. Since the 0.5 group consisted of brothers/teammates, I would have dropped them both into 0.0 score group. As they had taken 1/2 point byes, I would have paired them as follows:

K vs D
G vs N

Because I would be using the idea that those dropped from a higher score group are obligated to play the highest rated players of the score group into which they are dropped. I don’t know about dropping 2 players - but given no color problems, I’d have the higher rated of the two play the highest rated in the lower score group and the lower rated of the two play the second highest in the lower score group.

Further complicating the matter:

Of the 9 people left in the lower score group the following 5 were teammates: P,Q,S,U,Y (also teammates with N)

So - instead of allowing the lowest rated player to get the bye in this round (player R rated 191), I would have given the bye to player Q (rated 201). That leaves 4 players in the same team left in that score group of 8 - perfect.

U   O 
P    T
R    Y
S    X

Here i would a have X on BB and R and U on WW.

Cons: 3 color imbalances
S vs X is 2 unrated (just noticed now!)
Pros: No teammate matches
Dropped players into score group playing highest rated players.

Instead swiss sys in the bottom score group of 13 people (remember NPQSUY are teammates):
Gave the bye to the lowest rated player.
Avoided all teammates matching.
Avoided all unrated vs. unrated matchups

It paired those first 2 players who dropped down as:

S vs D
G vs N

Almost the same as I would have - except it gave K a very easy pairing intead of having to play 3. K was the player the parent was complaining about!

… I am going to stop there as I probably have answered my question - I had not noticed that if I took care of all the teammate matches that there would be an Unr vs Unr matchup - I am still going to post this though - just an an illustration for those curious. and to ask the following:

Before round 2 - seeing that there were 6 / 12 players in the bottom score group as teammates and seeing that the pairing program paired the dropping player against an unrated - would you trust the program? Would you repair that piece, avoid teammate matches and allow a Unrated vs Unrated? or would you repair that piece and allow teammate matches (avoiding Unr vs Unr).

Thanks!

2

What is the problem with unrated playing unrated?

Alex Relyea

3

Once the lowest rated (191) is given the bye, it depends on whether or not teammates are allowed to play each other. If not then D and G have to get paired with the two of the zero-point teammates and the 529 avoids D. Avoiding unr-vs-unr (no idea why you wanted to avoid that) is merely a minor variation on that theme that doesn’t change the 529’s avoidance. If the tournament is run avoiding teammate pairing if there is any way to do so then the pairings given were correct. Personally I would have preferred to pair the two 0.5s against the top two zeros (given that they were brothers that early in the tournament), but the teammate avoidance is quite reasonable and follows pairing rule 28N2.

4

There is a major difference between the two prominent pairing programs, SwisSys and WinTD, in how the bye is assigned when the number of players is odd.

If you read the rule absolutely literally, you would give the bye to the lowest-rated player in the lowest score group, before making the rest of the pairings.

However, it makes sense that you should be allowed to make a minor transposition in assigning the bye, just as you would in pairing the other players, if such a transposition will improve colors, or prevent teammates from playing each other, etc.

The author of SwisSys decided to follow the rule literally. That way, if the rule creates a problem, it is the fault of the rule, not the pairing program.

The author of WinTD disagrees, and feels that the absolute requirement to assign the bye to the absolutely lowest player is absurd, and inconsistent with the philosophy of the other rules.

This topic has been discussed in other threads, and has generated some heated arguments.

In general it is not always easy to say that one set of pairings is correct, and another is incorrect. Opinions vary, as do pairing algorithms.

Bill Smythe

5

Thank you for taking the time, guys.

So, I think I divulged which side of the argument I would be on - for consistency with the “spirit” of the other rules, I would have given the 1-point buy to someone rated 12 points higher in order to preserve the other aspects.

I dont know how I got sidetracked on Unr vs Unr - It appeared that the program was trying to avoid that and I invented a memory of reading a suggestion about avoiding that to fit what I thought I was seeing (human malfunction). My first read through the data (2 months ago) - I did not even notice.

If SwissSys gives the lowest rated person a bye and does that first, and also preserved the top 2 pairings, there would have been 5 teammates among 4 pairings - so it avoided that scenario by swapping player K (the player who was accused of getting an easy ride) with an unrated player.

Thanks again.

6

Upon reading your original post in greater detail, I’ve figured out that I like your proposed pairings better than the program’s pairings.

SwisSys essentially makes a huge transposition (1309 vs Unr) in 529’s opponent, just to rigidly enforce the bye assignment and to keep players from being paired against teammates.

Your version also avoids teammate pairings, and by relaxing the bye assignment rule slightly (by 12 points), also avoids the huge transposition.

Maybe it was from other, similar, advice from the rulebook or on these forums, suggesting that in a quad (round-robin sections of 4 players each) the unrateds should be spread out among sections so that each unrated has at least one rated opponent. Makes it easier for the rating system to initialize newbies’ ratings properly.

Bill Smythe

7

I have a question. How do you know that 1309 vs Unr. is a huge transposition?

Alex Relyea

8

I know that I’m not addressing the original question in this post, but I do want to commend Mr. Cassidy for learning the pairing rules and understanding how to do pairings manually. That’s what a tournament director ought to do, but I fear we now have many tournament directors who have never done so. Unfortunately, when these directors are asked why a particular pairing was made, their only answer is “because the computer says so.”

Having said that, I will also say that SwissSys is fairly reliable in making competent pairings. It has been a while since it has done anything that has left me scratching my head and submitting a bug report to the author. On the other hand, I’ve seen WinTD make pairings of which I am jealous because I would not have come up with pairings that good. I haven’t seen the source code for WinTD, but I’m fairly sure the algorithm it uses for pairing is “magic.” :slight_smile:

9

Considering that it is apparently a scholastic tournament for very young players, and considering the other players’ ratings, it is highly likely that regarding 1309 vs unrated as a huge transposition is quite reasonable. There’s no guarantee, though.

Bill Smythe

10

And don’t forget that the pairing settings on SwissSys (and WinTD) can be set by the operator. The TD can choose things like 3 colors in a row (or not), bye strictness, etc.