So this weekend, I was responsible for running 4 prerelease events, all between 26 and 32 players. (A tournament report on this will be posted today or tomorrow.) Upon arriving at the store, I determined that WER was not usable, on- or offline, and in fact would crash any time you clicked something. After this happened a couple times, it stopped opening at all, and I decided to proceed to run the event on notecards. Since my event paid out based on record, I did not need to calculate tiebreakers, and I used the following algorithm:

-Players with the best record are randomly paired against each other.
-If there are an odd number of players with that record, a random player with the next-best record is paired up.
-If any player is paired with another player who they've already played, that player is set aside and a new random player of the appropriate record is chosen.
-This process is repeating in descending order of match points.
-If there is an odd number of players, the last player left will receive a Bye unless they've already received a bye and there is another player also with the fewest match points who hasn't yet received a Bye, in which case a different random player with the fewest match points would receive the Bye.

I fortunately didn't have to deal with multiple pair ups/pair downs (can a player be paired up/down multiple times in a row?) or situations where the Bye couldn't both go to someone who hadn't had one and to someone with the fewest match points. That combined with no worries about tiebreakers meant that there were no player complaints, aside from the lack of printed pairings and match slips. I also managed to keep the turnaround of each round sub-60-minutes, which made me happy given the circumstances.

I have PPTQ's to run next week, and WER is simply an unreliable piece of software. Is there an “O” answer for the appropriate pairing algorithm? I know the MTR has some info on this, but it mostly only covers Top 8's, and it doesn't have answers to some relevant questions. Do tiebreakers impact pairings during any round, and if so, do they specifically impact pair up/downs? What is the priority order of the pairing rules? (Is it more important that players not receive multiple byes, that players not be paired up/down multiple times, there there be minimal pair up/downs, etc.?)

Ideally, WER will work as intended in the future, but I'd like to have a serious, viable alternative to handle this by hand when needed.

(I'm looking for an “O” answer, not guesses about how WER pairs based on circumstantial observation (e.g. “I once received two byes, so obviously…”).)

To my knowledge, answers to those specific questions have never been publicly available outside of WotC, and the specific algorithm, along with prioritization on “breaking a rule” are both non-public, and vary based on both the tournament software used, and the current round of the tournament.

There is an older DCI document called the "TO Handbook", which specifically mentions that, if a pairing is done by hand, to do it by:

pair players with the same match points
against each other randomly. (If there is an odd number, match one player from that
group with a person from the group with the next-highest match points.) Do not use
tiebreakers when pairing between rounds.

(entirely my opinion)
Honestly, if a situation occurred where I was running a WPN Premier Event on notecards, I would not worry about making everything 100% perfect here. The system you mention above sounds like you are covering all bases expected from players, and everything can be done efficiently for the majority of the tournament. I would probably calculate tiebreakers twice in total: going into the final round, and after Swiss.
(/entirely my opinion)

In all technicality, the algorithm used today is “the output of a pairing in WER/WLTR”, and that's about it.

> can a player be paired up/down multiple times in a row?

Yes. There was a time where the algorithm in WER seemed to pair by-standings in all rounds, and it was common for a player paired-down in R2 to get “stuck” getting paired down in every subsequent round if they were winning. I have not (anecdotally) heard of this over the past couple years, though.

Edited Jordan Baker (Jan. 16, 2017 09:09:56 PM)

I agree with Jordan, the system presented is covering all aspects that I can recall from running events by hand previously. The most important aspect is ensuring players don't play each other more than once until the Top 8. Calculating Tie-Breakers is going to be the hardest part and only needs to be done after the last round.

For calculating tiebreakers, Excel or Google Sheets are your friend.

The MTR tells us to use “the Swiss Pairing Algorithm” in section 10.4 which is fairly vague. There are a decent number of ways to pair a “Swiss” event and as long as you pair as many players of the same record as possible, it's Swiss.

When WER pairs, to my knowledge, it uses a different algorithm in the final round of Swiss compared to every other round. In most rounds it creates a bracket based on record and then pairs randomly within that bracket. In the final round it uses a “greedy” algorithm (how scorekeepers described it to me) where it does it's best to pair 1 vs 2, 3 vs 4 etc. (If players have played before they can't be paired, for example, if 3 and 4 have already played then 3 will play 5 and 4 will play 6). It is possible there is some additional work happening behind the scenes, but in general it's random within brackets until the final round, then it pairs 1 vs 2 etc.

The specific pairing algorithm is actually valuable to learn as a judge, as is tiebreaker math because it can help you quickly identify tables and players most at risk of collusion based on where they sit in the tournament.

Bryan, tie breaker math is well defined in the MTR, but that doesn't really relate much at all to the pairing algorithm, which appears to not be known to judges or the public.

I've never seen any actual documentation that WER pairs differently in the final round, and I would be interested in any real, non-anecdotal evidence on that. Certainly I wouldn't want to pair any way but randomly within records unless there is documentation telling me to do so.

I feel that since we as judges are supposed to be maintaining tournament integrity, which hinges quite a bit on pairings, it would only be helpful for us to have an “O” resource on how that is supposed to happen. WER is fine when it works, but given the recent problems, having a backup plan seems prudent.

There isn't any public documentation about how WER does pairings, but most regular users of it will have seen it do “odd things” in the final round. Most annoyingly, it will sometimes give the same player a bye twice in one event.

The question of whether, when WER does things like this, we should manually re-pair (because WER does not quite follow the pairing rules as laid out in the MTR, and of course getting two byes in a three round event is appalling), has been clearly answered with a “No”, for reasons of appearance of Tournament Integrity. WER takes precedence over the MTR.

So:
If WER is running: do what it says, no matter what.
If WER is not running: do standard Swiss pairing using cards, in accordance with the MTR (and don't sweat the fine details because we don't have an official source on what those details are, and they certainly don't follow the MTR).

I should have been more clear on what I meant by tie breaker math. I'm referring to the analysis that players will be doing to determine whether to play or draw and determine who still has a chance at the top 8. The “math” is in determining who is still in contention vs who is not based on tiebreakers. I'm not talking about the math to calculate tiebreakers.

I'm afraid I don't know where the documentation for WER is located. Nor am I certain on the details for how previous rounds are paired, but I am quite comfortable stating that WER always pairs 1 vs 2 etc in the final round. It is a big part of why players can “draw in” is because they can look at the standings, know who is playing who and determine outcomes from that information. I believe this priority is also why you get “odd things” in the final round. WER focuses on pairing the highest player with the next highest player they can legally play. Sometimes it runs out of legal matches before it runs out of players, but because it cares about 1vs2 more.

Edited Bryan Henning (Jan. 17, 2017 07:23:15 AM)

Eliana, I can't provide a lot of ‘O’ for you, because WER's algorithms are proprietary and not made public.

I can say that what you did sounds correct, and the best we could hope for in these circumstances.

Several months back, someone shared a web site that did Swiss pairings; it was created for situations when WER couldn't be used. (I wish I could search for that, for you…)

I will note that, in any case, the tournament still has to be uploaded via WER - or it didn't happen, and Planeswalker points are lost, and other BadThings™ could result.

d:^D

On Tue, Jan 17, 2017 at 5:56 PM, Scott Marshall <
forum-32813-5d3d@apps.magicjudges.org> wrote:

> Several months back, someone shared a web site that did Swiss pairings; it
> was created for situations when WER couldn't be used. (I wish I could
> search for that, for you…)


​Scott is likely referring to this:
http://apps.magicjudges.org/forum/topic/8248/​


//DLI

Sub 60 minute turnaround in that scenario is an achievement in the given scenario. Nicely done!

The only thing I'll add to the above (in terms of what you used, which is very sound and exactly what we were basically doing at our LGS) is that even though we don't care about tiebreakers, WER does ultimately care about how the round went in terms of W/L record for later reporting (To be clear, I'm not saying this because I don't think you know it, but more because it's something I forgot and it's how I learn). We were able to use the sheet to create pairings by assigning random number values to people and sorting by random values for each round based off of number of points the players had and assigning table numbers descending. Then just print the list and voila, you have pairings and table numbers.

Initially our LGS was just going to award points on the spreadsheet (3, 1 or 0) because they pay out to record too. This makes it a little difficult for reporting after the fact since it doesn't accurately report number of games won. The use of notecards could alleviate this (each player and DCI number at the top, and then Round number, opponent name, and the result for that round (W/L/D, # of games). In fact, you could use the notecards as the pseudo match slips (confirm results, enter them into a spreadsheet as you go along, etc) and just distribute them based off of table number round by round.

I wish I knew JS better (read: at all). I have all sorts of ideas for creating a dynamic excel spreadsheet that can be used in lieu of WER in cases like this weekend that would not only help reporting and organization of the tournament, but ultimately improve the quality of the tournament experience for our players, which is ultimately what keeps us going.

Edited Bernie Hoelschen (Jan. 17, 2017 11:19:07 AM)