Tuesday, September 27, 2011

Ooops... a netbook

Tonight was the big night: I had a meeting with Marleen who will go to Tokyo to judge for the Belgian team. She will also meet with the WAG commitee to give a demonstration of the software. So I finished the setup applications and was ready to install it on her pc.

And there it was ... a mini, mini pc.

I thought: that screen will probably not be big enough for the GS Judge application. And right I was. So I did some modifications to the startup size of the applications, but it was clear that I needed additional functionality to deal with PC's with small screens. And fast. So this is what is on my feature list for the next days to come:

  • Screen parts can be resized as wished
  • Windows size, position and state will be stored when the application is closed and restored the next time it is started
  • The exercise canvas will be scalable so that large exercises can be scaled to fit the window
  • Scroll bars will be added to enable vertical and horizontal scrolling

  • By the end of the week I will put it all online. Come check it out.

    Saturday, September 24, 2011

    Friday, September 23, 2011

    New mini-language extension

    The mini-language had two possibilities to indicate that an element is devaluated:

  • Typing a minus "-" character after the symbol indicates that the DV value must be one lower than the original DV (D becomes C, C becomes B, B becomes A)
  • Typing a less-than "<" character after the symbol indicates that the lower element from the code must be awarded (Double pirouette on floor becomes pirouette)
  • These two situations cover all devaluation scenarios.
    So I thought.
    Until I encountered this on the judge score slip:

    It is clear what happened here (I checked with a video on Youtube): the gymnast intended to do a double pirouette with one leg above horizontal, so that is what is written down by the judge. However the double pirouette was not double, so it is devaluated to single pirouette with one leg high (D becomes B). But the leg was not the entire turn held high enough and then it gets devaluated to a normal, single pirouette (B becomes A) (now this sets me thinking: at least 360 of the pirouette was with one leg high enough, so shouldn't a DV=B be awarded?).
    Anyways, there is no way with the "-" or "<" syntax of the mini-language to devaluate a double pirouette with leg high to a normal pirouette. So I extended the mini-language with a rather universal solution: after typing the actual performed element, the intended element is added between brackets. The GS Judge application then displays and counts the performed element, but also shows the intended element - whatever it was - with a red circle around it to indicate it was not rewarded.

    All this will become available in next weeks release.

    What is meta-data?

    When I am coaching, I find pleasure in explaining the biomechanics behind an element. I find the same pleasure in talking about the mechanics that drive the GS software. This time I want to talk about meta-data.

    What is meta-data? Meta-data is an IT term and in short it refers to data about data. When you enter a symbol into the GS Judge application, that is data. The DV that appears above the symbol, is the meta-data. Every symbol on every apparatus requires a lot of meta-data. For example on balance beam, we need to know for every symbol:

  • Difficulty value (A-G)
  • Is scale (yes/no)
  • Is mount (yes/no/maybe)
  • Is dismount (yes/no)
  • Acro or dance
  • Split direction (cross split, side split, upward, diagonal)
  • Is salto (yes/no)
  • Is turn (yes/no)
  • Is roll (yes/no)
  • Is handstand (yes/no)
  • Is hold
  • Lower element
  • Has flight
  • Supporting hand count (0,1, 2)
  • Take off from both feet (yes/no)
  • Element direction (forward, backward, sideward)
  • LA rotation (180, 360, ...)
  • BA rotation (360, 720)
  • Landing type (on 1 foot, 2 feet, straddle sit, ...)
  • Can repeat for CV (yes, no, once, ...)

  • Having one of these values wrong for a symbol, can have a serious impact on the calculation of an exercise. Today I encountered such a problem. The acro line below was not recognized as a series with two different saltos. The reason was that the meta-data of the whip salto was incomplete: it lacked the BA rotation value and therefore was not recognized by the program as being a salto. The end result was one CR missing, which was a mistake of 0.5 points.

    When the new version will be released next week, I expect that more meta-data bugs will surface. If you detect one, don't hesitate to report it so we can improve the product.

    Regards, Tom

    Wednesday, September 21, 2011

    Train delays help speed up development

    Developing the software for gymnastics is a hobby (my real job is developing enterprise applications for an Italian energy supplier). Most of the development is done on my way to and from work. As I travel by train - 1,5 hour in the morning and 1,5 hour in the evening - I am able to make a steady progress. But with the WC approaching, I need more time than I have. Then it helps when the train has an increased travel time of 1 hour, which happened yesterday :-)

    Thursday, September 8, 2011

    Always read the small letters!

    In this case, it is actually: read the word between brackets. Last week Donatella Sacchi - WTC member - was testing the GS Judge software, and almost always the results were correct except for this one case. See if you can spot it! I give you two floor exercises. What is the difference in difficulty value?
    You notice that both exercises are identical except for the dismount, which is in both cases a simple A salto. Would you tell that the DV of the second exercise is 0.5 higher than the DV of the first exercise? The reason is in the definition of an acro line: An acro line may consist of a min. of one flight element without hand support and take off 2 feet (rebound). And there you have it - the word between brackets: rebound. A forward salto requires an approach run and therefore the salto starts from a rebound take-off. That means that the forward salto is an acro line on itself and being the last acro line means that the dismount is this A-element. So no CR! In the other exercise, the final salto is a salto backward. You cannot take an approach run and then do a salto backward. So you do it from a standstill: no rebound, no acro line. The dismount is the salto with the highest DV in the last acro line, which in this case becomes the forward double twist (D). That means a CR of 0.5. Here is the score slip for both exercises:

    Wednesday, September 7, 2011

    Can the software be used on a competition?

    Yesterday I received this question from an Italian judge and with his permission I answer the question on our blog.

    The question is not new and to be honest I would not recommend to calculate the D-score using WAG Judge.

    This is how we are used to judge:

    1) While the gymnast performs her exercise, the D-judge writes it down. More often than not the D-judge is also judging the E-score.
    2) After the exercise every judge counts the deductions.
    3) The D-judge then starts computing the D-score.

    How long does this take? After the gymnast ends her exercise it takes 2 or 3 minutes to get to the final score. The critical step is the computing of the D-score: many mistakes are made here.

    If you just want to replace step 3) with the WAG Judge application, you will have correct results but the time required to input all data (Gymnast name or number, the symbols, the execution value and saving the exercise to a file) will take more time than before. The reason is that the software as it is today, is not optimized for this task.

    The best way to use it today during a competition would be to have the exercises already saved on disk. This means that every gymnast must submit their exercises before the competion begins and some people are needed to put it into WAG Judge (or have the coaches use WAG Judge to submit their exercises). When the gymnast finished here exercise, the D-judge would compare the actual performed exercise with the submitted one. The amount of work is then reduced to just changing the symbols that were performed differently from the submitted ones.

    Optimizing the application would mean:
    * have a database linked to the application so that all gymnasts in competition are already available together with submitted exercises.
    * have input fields for all E-scores
    * have the results stored in the same database that holds the gymnasts info and submitted exercises.

    Oh, and of course, having an IPad where you write the exercise down would also help. The problem there is that the amount of technology required to recognise a written gymnastics exercise along with all deductions and other things a judge writes down, is not available.

    Some people are convinced that submitting an exercise prior to competition can give reason for prejudging. Do you have an opinion about this statement?

    Sunday, September 4, 2011

    Meeting with WTC member Donatella Sacchi

    During the World Cup in Ghent I had the opportunity to talk with WTC member Donatella Sacchi. We talked about the current code and the new code, and how gymnastics software could be of any assistance to the development of the new code.

    The result of the meeting was that I had a clear idea of what new peace of software would be of extreme value to the WTC. It is actually an existing tool that we use to test our software.

    To test WAG Judge we have many copies of written exercises from major competitions and we verify if the results calculated by the WAG Judge software equals the score that was written on the score sheet (yes, we do get these from the WTC). This test is automated and after every change in the software, it takes only two seconds to re-evaluate all 250 (and growing every week) exercises.

    Having these 250+ exercises organised in a database, enables the WTC to automate time consuming tasks such as counting the number of falls after a competition or counting how many times each element has been performed. We now have this information - and much more - in our databases and it is just a matter of defining what statistical information is required. The actual counting of elements/falls/connections/... is done by the computer and takes only seconds to complete.

    One example where this kind of information is useful for the development of the new code, is the rule on beam that says that you receive one tenth more for a jump with full turn that is landed in side position. After reviewing all exercises performed under the current code, only one gymnast attempted such a jump. Therefore, this rule is a candidate to be removed from the new code.

    Another area where our software can help, is by implementing the rules of the new code and let it run over all exercises from the current code. Than we can compare the differences in the final score, but also see what the effect is on the ranking in a competition. If for example a new rule is designed to favour execution over difficulty, it shows the effect of the rule on the ranking and whether the better execution ranks higher than before.

    The gain is in the speed. In the past all this work had to be done by hand.

    So what this means for us is that we concentrate on WAG Judge to get it ready for Tokyo, and then turn our attention to the new code which is expected to be finished early 2012, and help the WTC evaluating it.

    If you are in Tokyo and want to get in touch with us, find Marleen Van Dooren, the judge for the Belgian WAG team.