Home Scoring and other IT questions

Robots and UMS file

I used two robots in a BBO duplicate pairs to make up a half table. They won with a hefty positive score - how do I correct it so that they play with no standing, and all the other players move up a rank and are awarded the right master points? I tried setting the Robots' score to a very low value (and adjusting other ranks/masterpoints manually), but that of course causes a 'bad scores' error when processed by the NGS system.

Comments

  • If you import the results to EBUScore, you can try marking them as missing in properties, or just add a large negative adjustment to the robot pairs score.

  • @Robin_BarkerTD said:
    If you import the results to EBUScore, you can try marking them as missing in properties, or just add a large negative adjustment to the robot pairs score.

    I thought I was doing that; I edited the XML file and changed the robots' score from +39.5 to -39.5 (an adjustment of -79 imps). I also changed the ranks of all the pairs and the master points to reflect the new ranking. But I left the other scores unchanged. That I suspect is what resulted in the 'bad scores' the next day.

  • @00022097 said:

    @Robin_BarkerTD said:
    If you import the results to EBUScore, you can try marking them as missing in properties, or just add a large negative adjustment to the robot pairs score.

    I thought I was doing that; I edited the XML file and changed the robots' score from +39.5 to -39.5 (an adjustment of -79 imps). I also changed the ranks of all the pairs and the master points to reflect the new ranking. But I left the other scores unchanged. That I suspect is what resulted in the 'bad scores' the next day.

    The problem solved on importing to EBUScore and marking the robots as missing. Thank you Robin. Easier than editing the xml file!

  • I just tried this with a 19 table set of results. A robot pair won the E/W and were awarded 70MP. It removed the robots Ok and recalculated the masterpoints. But now the top MP was 60 not 70 and the N/S awards were also reduced from 70 to 60. So not working as I thought it would.

  • Robin's suggestion of putting a large negative adjustment to the robot's score works much better.
    Everything else remains unchanged

  • @mickg said:
    I just tried this with a 19 table set of results. A robot pair won the E/W and were awarded 70MP. It removed the robots Ok and recalculated the masterpoints. But now the top MP was 60 not 70 and the N/S awards were also reduced from 70 to 60. So not working as I thought it would.

    This is the correct way of doing it. By removing the robot pair, you are reducing the field from 19 tables to 18.5 tables, which coincides with one of the masterpoint cutoffs, and therefore the program is correct to reduce the awards from 70 to 60. In contrast, if there had been 18 tables, removing the robot pair would lead to 17.5 tables and unchanged masterpoints.

    So technically...

    • If you take this valid approach, you have to be consistent for all of your sessions (you can't pick and choose based on the masterpoint awards).
    • If you want to give the robot a negative adjustment, the only legal way that keeps the masterpoints unchanged would be to give them lots of procedural penalties (falling short of disqualification I suppose). One good example would be the failure to disclose their leading and carding agreements (a warning first, and then 25% of a top or 4 IMPs per failure).

    You can't retain all the masterpoints for your human players and consider them without standing, but you can have a biased TD :) Anyway those are just the technicalities.

  • We would like to mark the Robot pairs as missing in our club. We can do this in EBUScore and load it directly to EBU. However, We upload our scores via Pianola so that we can report statistics. Pianola will not load results with a missing pair because the pair is still showing in the scores even if you delete all the boards played by that pair. Ideally the boards shouldn't be deleted so that the scores for the opposing pair stand. In other words all the human players below the robot pair move up one with no change to their % score but distributes the robot pair master points to the human players

  • I have now got this to work. I had inadvertently left a score in on my test. It will only work by removing every score so this actually changes the results because the opposing pair on the deleted boards obviously did play in the session if only with the robots to avoid the sit out. It doesn't feel right to change the course of events this way and I'm not sure it's legal.

  • I simply do what Robin Barker suggested some time ago which is to put a minus adjustment against the robot pair (something like -200 matchpoints ) in EBUScore which moves them down to last place. All other scores remain unchanged and NGS uneffected. Then the humans get the masterpoints.

  • @mickg said:
    I simply do what Robin Barker suggested some time ago which is to put a minus adjustment against the robot pair (something like -200 matchpoints ) in EBUScore which moves them down to last place. All other scores remain unchanged and NGS uneffected. Then the humans get the masterpoints.

    It does mean that sometimes the humans get more master points than they were entitled to.

  • By making any adjustment it changes the actual course of events without any real justification. As Gordon commented the result of the human players is artificially adjusted also. The legality of that is questionable. What would the EBU's stance be on clubs consistently awarding negative adjustments to Robots?
  • edited October 2020

    I think the legal justification for doing this is to make it a condition of the event that it is for human players and then have the robot pair play without standing to make up the tables.

    White Book 2.4.9 Stand-by players, players who do not qualify, and disqualified players
    A stand-by contestant may take the place of a contestant who is not present at the advertised
    starting time – see §2.3.3. A stand-by contestant may also be allowed to play in an event for
    which they do not qualify (for example, to avoid a sit-out – see §2.3.3). Any such stand-by
    contestant plays without standing and does not appear in the final ranking list.
    Any results obtained by the opponents of a stand-by contestant or a contestant playing without
    standing will count in full.

    But this does involve taking them out of the ranking list altogether, not just moving them to the bottom of it.

  • Could one argue that the robots are cheating by using a computer to track every card played, as a memory aid? In that way some sort of penalty would be applicable for every board they play? Say -10 masterpoint for each board played or whatever...

    Not that it bothers me - I see it as something of a challenge, lets try to win or at least lets beat the bots!

  • That's helpful advice thank you Gordon. Our problem is that Pianola can't cope with just marking the Robots missing. EBUscore leaves the scores in regardless to the robot pair being missing. Logic tells Pianola that it needs to assign those points to someone so will not accept the results. I don't know how the EBU database deals with this. It can only be by ignoring the sit out pairs results but awarding the human players results as they stand. It would be good if Pianola could handle it in the same way but this would probably involve an enhancement to both the EBUscore and Pianola systems, which in the current climate may not be considered feasible.
    Pianola holds our membership data and is invaluable for maintaining records, group emails & statistical reporting. Neither the EBU or Bridgewebs databases perform all of these functions effectively. So we have no other option other than to let the Robot scores stand. If it was accepted that Sitouts are inevitable with an odd number of players and not to substitute Robots it would solve the problem of course.
  • Someone asked me recently whether we could provide a way in BBO Extractor/BBOtoXML to move robot pairs down the ranking list because people don't like them coming top/getting masterpoints. I am currently in the process of implementing this by allowing a large negative adjustment to be provided (e.g. subtract 200 matchpoints) as suggested in one of the posts above.

    This moves the robot pair down the ranking list but does not affect the matchpoints or percentages for any of the other pairs. However, Gordon points out that it may mean that players get more masterpoints than they are entitled to, because the robots make up a whole table which would otherwise not be included in the table count.

    On the other hand, if the robots are removed from the ranking list completely, should their scores remain in the scorecards/travellers, so that boards played against them are counted in the overall results ? This is problematic for software that expects to find a pair in the ranking list if that pair appears in the travellers.

    It would of course be possible to remove the robots and all traveller lines in which they would appear. This would mean that all results obtained against the robots would be discounted, and it would be as if the players had a sitout rather than playing against the robots. The final percentages on the ranking list could be substantially different to those which appeared at the end of the event.

    So from my point of view, the first and third options are technically possible, and the second may not be, but which is the approved approach ?

  • As stated above, if the robots are considered to be playing without standing, all results against them do count.

    It all scores achieved by and against robots were to be removed, you might as well have a sitout, which solves all problems, except that people play less bridge.

  • @gordonrainsford said:
    As stated above, if the robots are considered to be playing without standing, all results against them do count.

    It all scores achieved by and against robots were to be removed, you might as well have a sitout, which solves all problems, except that people play less bridge.

    I understand, but I have to consider the practical problems of implementing that approach in software. LindaB has mentioned above that it would cause problems with Pianola, and possibly EBUScore and EBU UMS. It may also cause problems with Bridgewebs.

    One approach may be to move the robots to the bottom of the ranking list, and leave them in the ranking list, but calculate masterpoints as if the robots were a sitout pair.

  • On Pianola if a pair has scores on boards then they appear in the ranking list. If they are marked as missing in EBU Score then they are ranked last in the UMS file, and Pianola will show them as last.

  • @jgoacher said:
    it would cause problems with Pianola, and possibly EBUScore and EBU UMS. It may also cause problems with Bridgewebs.

    Of those it's only Pianola that has the problem.

  • NGS has a problem if you give a pair large adjustments and don't remove them from the ranking list - then the BAD SCORES result is invoked.

  • @gordonrainsford said:

    @jgoacher said:
    it would cause problems with Pianola, and possibly EBUScore and EBU UMS. It may also cause problems with Bridgewebs.

    Of those it's only Pianola that has the problem.

    It also causes problems with my Results Analysis software on Bridgewebs (part of Bridge Solver Online)

  • edited October 2020

    @gordonrainsford said:
    NGS has a problem if you give a pair large adjustments and don't remove them from the ranking list - then the BAD SCORES result is invoked.

    In that case we may have to generate a different copy of the XML file for UMS upload, if Robot pairs have been removed from the ranking list. We are trying to move away from producing separate files for Bridgewebs and UMS. In the latest release of BBOtoXML we generate a single version except when the file size is large than 2 MBytes (because EBU UMS currently can't cope with that). Looks as though we will have to do the same if robot pairs are removed

  • But they really shouldn't be in the ranking list if they are without standing.

  • ""Robin_BarkerTD12:29PM"
    "On Pianola if a pair has scores on boards then they appear in the ranking list. If they are marked as missing in EBU Score then they are ranked last in the UMS file, and Pianola will show them as last.""

    This isn't the case Robin. If the robot pair is simply marked as missing in EBUScore the UMSfile file doesn't show the rankings for the robots at all. However, the Traveller Lines and Board detail lines still show the missing pair number together with their outcomes on each board. The UMS upload to Pianola fails completely. With the following error message:-

    Upload
    These results cannot be uploaded because there are scores for pair 1 but there aren't any players defined for this pair number. Please check the settings in your scoring program and try generating the file again. If you're still having problems, please email your file to support@pianola.net.

    The only way to get the UMS loaded is to mark the robot pair as missing and remove all the boards played by the Robots, which removes the results for the opposing pair also. This distorts the results for the human players .
  • @gordonrainsford said:
    But they really shouldn't be in the ranking list if they are without standing.

    My previous post was intended to address that, though I may not have expressed it clearly. I meant the Robot pair would be omitted from the ranking list in the EBU UMS version of the XML file, if generated, but the traveller lines would remain and the masterpoints would be calculated based on the reduced number of tables.

    In the standard version of the XML, intended for Bridgewebs, Pianola, and others, we would include the robot pair in the ranking list but move them to the bottom, the masterpoints would be the same as in the UMS version. I can't see that it matters if the "without standing" pair(s) appear at the bottom of the Bridgewebs ranking list. I am not prepared to remove "without standing" pairs from the standard version of the XML because it causes problems for downstream software including Pianola and my own software on Bridgewebs, and it means that the XML is internally inconsistent (pair numbers in TRAVELLER_LINES that do not appear in PARTICIPANTS). I suggest the correct solution would be to add a field to the PAIR element in USEBIO to indicate that a pair is without standing, but this would require an update to the USEBIO standard.

    We would only generate the additional EBU UMS version of the XML if removal of the robot pair was explicitly requested via the BBOtoXML user interface .

    All this is for a future release of BBOtoXML, we don't do it at present.

  • "jgoacher2:10PM edited 2:11PM"

    "@gordonrainsford said:
    NGS has a problem if you give a pair large adjustments and don't remove them from the ranking list - then the BAD SCORES result is invoked."

    In that case we may have to generate a different copy of the XML file for UMS upload, if Robot pairs have been removed from the ranking list. We are trying to move away from producing separate files for Bridgewebs and UMS. In the latest release of BBOtoXML we generate a single version except when the file size is large than 2 MBytes (because EBU UMS currently can't cope with that). Looks as though we will have to do the same if robot pairs are removed

    I'm not confident that having a different copy of the XML file for UMS upload will achieve anything. We load results via EBUScore then Pianola. Unless both those platforms undergo programme enhancements the BBOconverted XML file upload to EBUScore will simply return error messages. The scoring side of EBUScore requires that the NS & EW MP's total the same for every board . It will not allow the missing pair side of the board to be 0 and just the unaltered opposing pairs scores to stand. What will the traveller lines show? This all comes down to the fact that computer systems are, in the main, logical beasts. In order to appease players who, understandably, get demoralised by seeing the robot pairs come in the top third often, we are trying to solve the issue by getting logical systems to cope with a situation where the input to the system is what actually occurred but to then convert data so that the output shows a different version of events. This will be extremely problematic for just one program to work with leave alone the BBO Extractor/BBOtoXMLBBO software, EBUScore, Pianola, Bridgewebs and EBU. On top of that there is the compliance with the White Book rules.

    I think it has to boil down to either living with the robots or allowing sitouts .
  • @LindaB said:
    "jgoacher2:10PM edited 2:11PM"

    ...
    I'm not confident that having a different copy of the XML file for UMS upload will achieve anything. We load results via EBUScore then Pianola. Unless both those platforms undergo programme enhancements the BBOconverted XML file upload to EBUScore will simply return error messages. The scoring side of EBUScore requires that the NS & EW MP's total the same for every board . It will not allow the missing pair side of the board to be 0 and just the unaltered opposing pairs scores to stand. What will the traveller lines show?
    ...

    I'm not suggesting that the missing pair side of the board would be 0 in the traveller lines in the EBU UMS version. The traveller lines would be unchanged in the UMS version, and everyone's total scores and matchpoints would be the same, just the robot pair would not be included in the ranking list. I don't know if this works with EBU UMS or not but I believe it is what @gordonrainsford is asking for. I don't know about the EBUScore issues, but it's not part of our processing chain: BBO Extractor -> BBOtoXML -> Upload to scoring website and/or EBU UMS

    I think it has to boil down to either living with the robots or allowing sitouts .

    Indeed, that would make life a lot simpler !

  • The robots have won at our club a few times, but that is because they have done better than the human pairs on the hands played. But their NGS ratings and finishing percentage also influence the NGS grades of all of the other players in the session. So will artificially adjusting the robots' score in other sessions also badly affect the NGS ratings for the clubs who accept the robots as normal players? I can't see how this will play out.

  • @PeterB001 said: So will artificially adjusting the robots' score in other sessions also badly affect the NGS ratings for the clubs who accept the robots as normal players?

    That is a really good question!

  • In theory it would, yes. In reality, robots are used so often at the moment that the grade is unlikely to be affected much as long as very few clubs make this artificial adjustment. It's one of the many reasons why adjusting the robots' score in this way is not a good idea, even if it's more straight-forward than the alternatives.

Sign In or Register to comment.