BBOtoXML - Incorrect ties
I've noticed that in the final EBU Pairs results - https://www.bridgewebs.com/cgi-bin/bwoo/bw.cgi?pid=display_rank&event=20210403_3&club=eburesults - two sets of pairs are shown as tied, but their MP scores are different (and there are no sit-outs). As I understand it, ties should be broken as long as there's a difference of at least 0.0001 MPs?
Comments
I think we will find when the results are combined for master points on My EBU they will be separated - I suspect it's a feature of Bridgewebs.
The USEBIO XML spec does not include an element for total matchpoints. Therefore the totals are calculated by Bridgewebs , by summing the matchpoints from individual traveller lines. The sum is displayed to 2 decimal places, but as the matchpoints in the traveller lines are already rounded to 2 decimal places in the XML file the sum of the values has a lower degree of accuracy than that. It would be possible to put more decimal places in the XML file, but Bridgewebs displays the values as they appear in the file. The Bridgewebs displays become less readable if a larger number of decimal places is shown.
However, the percentages shown are correct to 2 dp because they are calculated by BBOtoXML from the unrounded matchpoints.
I believe it has been stated that ties should be broken if there is a difference of 0.0001 MPs. However, it seems absurd to do this when different scoring programs sometimes show discrepancies of 0.01 in the computed percentage (some recent comparisons between Realbridge and EBUScore highlighted this). RealBridge calculates the results to full precision, whereas EBUScore does some rounding. I note that in the event you refer to, the matchpoint totals of the tied pairs were greater that 2600, so a difference of 0.0001 MPs is 1 part in 26 million !
BBOtoXML currently calculates the matchpoints and percentages to full precision but then rounds the percentage to 2 dp and regards 2 pairs as having tied if ttheir percentages agree to this level of accuracy.
On a separate issue related to the above event, I note that the uploaded event does not include contracts, lead cards, number of tricks, or play links. The EBU UMS upload cannot handle files larger than 2 MBytes, so for large events BBOtoXML produces two versions of the XML, a full size version for Bridgewebs upload and a "lite" version, omitting the above fields, for EBU UMS upload. It appears the "lite" version has been uploaded to Bridgewebs, which is why these fields are missing.
The difference in all the results was in the first decimal place of match points: luckily the close first place was 0.4 match points, and the percentage rounded up and down to split the 'tie'. The tie at 13= is ...3.1 v ...2.9.
I will split the ties when awarding master points.
The ties shown on BridgeWebs can be fixed by editiing the XML: replace the second 13 by 14, etc. Getting the right ranking master points to the right pair was harder.
Thanks Robin. It looks like the masterpoints for 22nd and 23rd have been swapped around? But hopefully this is an easy fix in the XML file.