Please keep the forum protocol in mind when posting.

Judging Technology » Post: WER rounds randomly.

WER rounds randomly.

May 23, 2015 12:13:14 PM

Philip Böhm
Judge (Uncertified), Tournament Organizer

German-speaking countries

WER rounds randomly.

Today I hosted a 7-man event with WER rounding..randomly.

Mathematically, 2nd and 3rd place should have the exact same opponent score. WER doesnt compute this well enough, so puts one at 66.6667, puts one at 66.6666.

Now the 2nd Tiebreaker (GW%) is higher for the player who got put 3rd by WER, so by MTR he should be 2nd!



What authority does the HJ have to “overrule” this problem of the program, for example if the event has an announced prize structure of packs for 1st and 2nd, but none for 3rd place ?

Edited Philip Böhm (May 23, 2015 12:15:33 PM)

May 23, 2015 03:51:09 PM

Aaron Henner
Judge (Level 2 (Judge Academy))

USA - Northwest

WER rounds randomly.

It isn't … random. I know of 2 bugs to do with WER computing the Opponent Match Win %.

1)
WER considers 1/3 + 1/3 + 1 to be different from 2/3 + 2/3 + 1/3.

2)
WER treats 1 win out of 3 possible differently from the MTR prescribed minimum 1/3 (get's rounded to a different decimal place).

Both of these problems are far more likely to occur in a 3-round tournament than any other number.


As far as to overruling WER…. that sounds like something that deserves an response.

May 24, 2015 06:45:31 PM

Gareth Pye
Judge (Level 2 (Judge Academy))

Australia and New Zealand

WER rounds randomly.

As I don't speak the language the pic is from I can't remember the
order of the columsn but those players do have different records
(3/2/0/0 and 2/1/0/1) so they should have differences. I was unaware
of the first difference in math that Aaron mentioned (but that sounds
like a plausibly a floating point rounding error) but the difference
between a natural 1/3 record and the enforced 33.33% minimum has been
around for a long time.

May 24, 2015 07:41:15 PM

Jordan Baker
Judge (Level 2 (Judge Academy)), Scorekeeper

USA - North

WER rounds randomly.

1) If the numbers aren't being stored to handle the use-case, then the rounding will get you:

1/3 = .3333… -> .33330
2/3 = .6666… -> .66670

1/3 + 1/3 + 1 = .3333 + .3333 + 1 = 1.6666
2/3 + 2/3 + 1/3 = .6667 + .6667 + .3333 = 1.6667


2) The MTR doesn't specify the minimum tiebreaker as 1/3; it specifies it as .33, i.e. .330. At one time, it was 1/3, but it was changed to .330 practically when WER was introduced.


> As far as to overruling WER…. that sounds like something that deserves an response.

This isn't an official response, but official responses in the past on this forum regarding things like this have pretty universally been “WER represents reality in terms of the event, not following it constitutes tournament fraud.”

Edited Jordan Baker (May 24, 2015 07:41:48 PM)

May 24, 2015 08:22:11 PM

Scott Marshall
Forum Moderator
Judge (Level 3 (Judge Academy)), Hall of Fame

USA - Northwest

WER rounds randomly.

As for overruling WER - no, you do not have that authority.

May 25, 2015 05:19:07 PM

Aaron Henner
Judge (Level 2 (Judge Academy))

USA - Northwest

WER rounds randomly.

I recommend an addition to the MTR stating that in cases of discrepancies between the MTR and the official tournament reporting software (whether WER, DCIR, WERS, etc), that the official tournament reporting software stands (both for this topic, and the frequently-occurring topic of WERs odd Swiss-Pairing logic).

May 26, 2015 08:31:17 AM

Rebecca Lawrence
Judge (Uncertified)

USA - Midatlantic

WER rounds randomly.

The record information is, in order, Played/Win/Draw/Bye, so it makes sense that the bye would be ranked lower I would think; this may be less a rounding error and more a way for WER to handle byes more elegantly