Adjustment of RGX calculation and introduction of division limits – English

++ HIER FÜR DEN KOMPLETTEN ARTIKEL AUF DEUTSCH KLICKEN ++

Hello dear community, 

a few weeks ago we conducted a big survey about our Roundnet Germany Index (RGX), which was introduced last year. We received a total of 133 feedbacks from you, which confirmed and complemented our own observations and analyses of the 2022 season very well. We were already aware that the underlying Elo system is experimental and still has some problems so early after its introduction. These inevitably arise from the small number of games compared to other Elo systems, such as in chess or table tennis. Since this problem was foreseeable, RGX was introduced in the 2022 season as a test phase. Due to the many tournaments in 2022, the number of games in the RGX has increased considerably, so that the introduction of initial measures based on the RGX makes sense. For this purpose, the working group „Ranking“ has intensively dealt with the results of the survey and carried out a data analysis on the current actual state of the RGX in order to further develop the RGX for the year 2023 and to derive measures to optimize the tournament system. The goals are on the one hand to develop a tournament system that is as fair as possible and on the other hand to use a ranking that reflects the actual playing strength of the players.

Today we want to present you the results and the resulting adjustments to the RGX, which are valid from now on: 

Important: All following adjustments in the calculation will be applied retroactively to all tournaments starting with the data sets from the year 2021 and not only for future tournaments! This will change the RGX values for all players.

TO THE TOTAL STANDINGS OF THE RG INDEX
TO THE INFO PAGE OF THE RANKING

ADJUSTMENT 1 – NEW CALCULATION BASIS FOR THE RGX

Problem 1: Good Intermediate players have too many points compared to Advanced players – new Advanced and Pro players (especially from abroad), on the other hand, are often undervalued too much at the beginning.

As one of the most common observations, this presents us with some problems: On the one hand, good intermediate players are ranked too high in the cross-divisional ranking, which at first merely contradicts the subjective assessment. However, this could also lead to the fact that these players would be assessed too well in the grouping, should they play an advanced tournament. The result could be unbalanced groups. In addition, there is the problem that foreign top players or players active in the German Roundnet Tour (GRT) for the first time are heavily underrated at their first tournament with the same starting value of 1,000 points, which has always been fixed as before, depending on the quality at the beginning, and thus disproportionate gains and losses occur.

In addition, it can be observed that in the Intermediate division there are more players who are actually too good for this division and thus distort the level because they play with a partner at a lower level, but dominate too strongly as a single player (e.g. by serving on a way higher level). The result is an understandable frustration for the remaining „actual“ intermediate players and a disproportionate increase in points for the actual advanced player. We see the main solution to this problem not exclusively in the adjustment of the calculation of RGX, but its use in personal division estimation, which is why we have defined an additional rule for this (see below Adjustment 2 – Division Limits). However, it is also a bigger problem from RGX point of view, because the Elo system works only if the points „move from the bottom to the top“, i.e. there is an exchange between the divisions. Otherwise, intermediate and advanced rankings are formed, but they are hardly comparable with each other.

Finally, in the current system, there is little exchange between divisions. Few successful Intermediate players play Advanced in the next tournament. Few „weaker“ Advanced players play Intermediate in the next tournament. This creates silos in which the ranking develops independently. Comparability is thus made more difficult. Accordingly, the division limits also come into play here – in addition to the following mathematical adjustment.

Adjustment to solve problem 1: Different starting values per division

The starting value of a player is determined by the division at the first tournament – so there are different starting values for Beginner-, Intermediate-, Advanced- and Pro-Divisions (qualification necessary). Players who start with a tournament without division (e.g. German Championships or tournaments before division introduction in April 2022) start with a default value of 1,300 points (average of all divisions). This mitigates problem 1: If you start new in Intermediate, you gain (or lose) about the same amount of points as before (because the difference to your teammates stays about the same). Due to the lower starting score, however, it takes a little longer to catch up with or overtake the players in the lower Advanced range. In addition, the special case of mostly foreign Advanced and Pro players, who play their first GRT tournament and would be clearly undervalued with the former standard starting value of 1,000 points, is handled.

Until now: In all divisions the starting value is 1,000 points.
New:

  • Beginner: 1.000
  • Intermediate: 1.200
  • Advanced: 1.400
  • Pro: 1.600 [currently only applies to (foreign) players who qualify for the Pro Division according to the tournament rules – see separate news on the Pro Division]
  • withot division: 1.300
    Info: The Beginner/Intermediate combo division is treated as Intermediate, Advanced/Pro as Advanced.
Problem 2: A 2:0 win in the sets has the same value as a 2:1 win

As further feedback from the community, our existing idea was confirmed: The current rule where the winning team gets all the points credited should be changed. In a match in „Best of 3“ sets, currently a 2-0 win results in the same point increase as a 2-1 win. Thus, losing a set does not have a negative impact for the winning team, and winning a set does not have a positive impact for the losing team, which does not reflect the actual performance of both teams. 

Adjustment to solve problem 2: Adjustment of the calculation formula

The calculation of the RGX will be adjusted so that the points in the „Best of 3“ mode are no longer distributed with p =[0, 0.5, 1], but with p=[0, 0.33, 0.5, 0.67, 1].

For clarification, here is a brief explanation of the general Elo principle: The points for winners and losers are calculated based on the difference between the „expected“ result and the actual result. This difference is then multiplied by the maximum score k. To calculate the expected value, the scores of the two teams are compared. The result is, for example: Team 1 wins the game with 70% probability. If only [0, 1] is used to calculate the actual result (excluding draws), there is only „all or nothing“. So the team gets k points for a win (1 – 0.7) and k (minus) points for a loss (0 – 0.7).

What changes now: The new levels lead to the fact that the actual result with a 2:1 is evaluated with 0.67, or 0.33 (instead of 1 and 0). So the „70% Team 1“ loses some (few) points in case of a 2:1. The actual result of the match (from team 1’s point of view) is 0.67, but the expected result is 0.7, therefore team 1 (0.67 – 0.7) gets k (minus) points. These inserted intermediate steps reduce the variance in the calculation, because game results are closer to the expected value more often than with the „all or nothing“ calculation.
In order to be able to understand the different cases more exactly, you can play through the most different cases in our RGX Calculator and understand the point allocations: RGX Calculator.

Summary: The rule generally leads to the fact that a set win in a „Best of 3“ match will be rewarded and the point distribution will more often be based on the expected outcome. This results in the possibility (also desired by you) that strongly favored teams could make minus points even with a 2:1 win, but an „underdog team“ can win points for a set win despite losing the whole match. A 2:0 and a 1:1 will be evaluated in the same way as before.

Problem 3: A 1:0 win has the same value as a 2:0 win

In line with the request from the survey for a weighting within a win in a „Best of 3“, the difference to a win in a 1-set match was also addressed. Currently, the same number of points are distributed for a win in a „Best of 3“ as in a 1-set match. As a result, on the one hand „lucky“ upsets, i.e. wins by „underdog teams“, are weighted too highly and on the other hand the achievement of beating a team by 2-0 sets makes no difference in the RGX compared to a win in a 1-set match. 

Adjustment to solve problem 3: Additional factor in the calculation formula

In addition to the adjustment to the calculation formula already described above, 1-set matches will be scored by a factor of 0.75. This means that for a win in a one-set match you get 75% of the points as for a 2-0 win against the same team.

In this move we also thought about adding a weighting between sets up to 15 or 21 points. We decided against such a weighting, because it would further complicate the formula as well as programming, we don’t have the corresponding data for some tournaments and therefore there is no added value.

Problem 4: The ranking is too close – The point values are too close to each other

The final aspect mentioned in the survey was that the ranking with the RGX values is still too dense overall, which was confirmed by our data analysis. The distribution of the point values is currently around 800-1,400 points, which makes it difficult to distinguish between the top and the basement, especially in the middle and thus largest player range. The gap between the Pro players and the Advanced level as well as between the Advanced players and the Intermediate level is perceived as too small by both you and us. 

Adjustment to solve problem 4: Change of the formula parameters

There are three adjusting screws in the formula for calculation: 
(1) the values for the actual game result (see above)
(2) the maximum points k achievable per game (previously 40, per game end 20)
(3) a constant for the team difference d (so far 400) to calculate the expected result.

Rule of thumb: If the difference between the teams is d, the expected result is 0.9, so the winning team gets 90% (or 10%) of the maximum points. If both teams have the same number of points, the expected value is 0.5, and if the difference is approximately twice the value of d, the expected value is 1, which means that from this difference on, the favored team cannot win any more points.

The parameters from (2) and (3) are now changed to 50 and 550, respectively. This leads to the fact that one can win/lose more points in a game (per person now maximum 25 instead of 20), but this maximum number of points is reached only with a higher point difference between the teams. Test calculations and a comparison with the current ranking have shown that especially the upper part of the ranking is pulled apart a bit as a result. The basic order of the ranking, however, remains largely the same, except for selective deviations – match finishers who were previously in the midfield/top/bottom are still located in this area.

The new formula as a basis for calculation

For those of you who are interested, we would of course like to communicate the correspondingly adjusted formula behind the calculation. This differs from the previously used only by the changed parameters. Only for the gradation of 1-set games a new constant b is inserted: 

With:
d = 550
k=50
b [0.75, 1] (1 if Bo3; 0.75 if 1-set game)
p [0, 0.33,0.5, 0.67, 1] (match result from team 1’s point of view; 0 if 0:1 or 0:2; 0.33 if 1:2; 0.5 if 1:1; 0.67 if 2:1; 1 if 1:0 or 2:0)
r1 = previous score team 1
r2 = previous score team 2
e1 = calculated expected value for team 1
x1 = alculated point gain/loss for team 1

Further adjustment: Later point deduction for inactivity

Furthermore, points in the RGX will be deducted proportionally from the players in case of inactivity. This is to prevent that a person who has not played in a tournament for a year, for example, and has presumably also lost in playing quality, is „overrated“ due to results that are far away in time.

Up to now, this procedure has been applied after six months. However, since this period was perceived by us as too short (especially with usually less tournaments offered in winter and/or in case of more severe injuries), the first deduction will be made in the future only after nine months of inactivity. Further steps are set after 15, 18 and 24 months. In addition, if you have not played a rating tournament for 12 months, you will be removed from the ranking.

Here are the exact numbers:
9 months of inactivity: 12.5% on the score above the mean starting value 
15 months of inactivity: 25% on the score above the mean starting value
18 months of inactivity: 50% on the score above the mean starting value 
24 months of inactivity: 100% reset to the mean starting score

Example: A person with an RGX of 1,500 does not play a tournament for nine months. Thus this person loses 12.5% of the score above the mean starting value of 1,300, i.e. 12.5% of 200 points = 25 points and thus drops to a value of 1,475. After one year, 25% of 1475 is subtracted, i.e. (rounded) 44 points to a RGX of 1431, and so on.

Conclusion: It gets more complex, but also fairer 

After many weeks of consultations, evaluation of the survey and trial calculations with different formulas and various data analyses, we have made an important step with the adjustments from our point of view to achieve the two goals „fair tournament system“ and „ranking that reflects the actual playing strength of the players“. We were aware that the Elo system „RGX“ still has some problems so early after its introduction. With the adjustments we have implemented your most frequently mentioned problems together with our own analysis results. Thus, we are one step closer on the association side to setting up a system that is as fair as possible for everyone.

However, we have to keep in mind that our system, as good as it definitely already is for our human and financial resources as a voluntary association compared to other sports and associations, is still far from perfection. In addition, many different subjective opinions always play a role in the evaluation, which we can not all reconcile. Especially among players with very few tournament participations we will observe outliers every now and then and even the best system can be exploited here and there by individuals. Therefore, we will continue to monitor the development in the coming year and make further adjustments to get closer to our goal. We would be happy if you would continue to observe the development from your point of view and give us feedback. If you have any questions, please do not hesitate to contact us. 

To get a feel for the changed RGX values, we have listed a few players from all divisions with the previous and new RGX values as examples:

Name„old“ RGXnew“ RGX
Lukas Eisenträger (PRO)1.3861.782
Mario Bürkle (PRO)1.3621.701
Julia Stadler (PRO)1.4071.786
Alexa Peusch (PRO)1.3951.786
Sascha Lang (ADVANCED)1.1501.476
Anne Metschies (ADVANCED)1.0601.400
Joshua Dülz (INTERMEDIATE)1.0081.276
Vivien Leß (INTERMEDIATE)9341.164
TO THE TOTAL STANDINGS OF THE RG INDEX
TO THE INFO PAGE OF THE RANKING

ADJUSTMENT 2 – INTRODUCTION OF DIVISION LIMITS

In addition, we would like to introduce another change, which is based on the (new) RGX and should improve the classification of the players into the different divisions. Up to now, this has been done at tournaments by personal self-assessment (exception: Pro-Division). We have identified – also through your feedback in the RGX survey – two problems in particular that we want to address with the new rules. These will apply immediately to GRT tournaments and will thus be applied for the first time to tournament registrations for Erlangen (25 & 26 February) and Opladen (25 & 26 February) on 15 January 2023. The registrations already made and possible late registrations for the tournaments in Bonn (07.01.), Göttingen (07. & 08.01.), Munich (21. & 22.01.) and Siegerland (28. & 29.01.) remain unaffected.

Problem 1: Players register for a division that is too low 

First of all it can be observed that in the Intermediate division more often players play who are actually too good and distort the level because they dominate too much as an individual (e.g. via serving). The choice of division could be due on the one hand to the fact that players deliberately play in the Intermediate, partly out of fear of losing too often and significantly in the Advanced and losing a corresponding number of points. In addition, due to the same starting values for all divisions, it was possible to win many games in the Intermediate division and thus collect points much more easily than in the Advanced division. The final reason could be that Advanced/Pro players play with a partner at Beginner/Intermediate level and believe that this would make the team „on average“ an Intermediate team. Of course, it is not possible for us to understand which reason was the decisive factor in the respective decisions.

However, the result – proven by the survey results – is often an understandable frustration among the remaining „real“ intermediate players and a disproportionate increase in points for the actual advanced/pro players. In addition, a problem arises for the calculation of RGX: the Elo system only works well if the pools of players mix naturally and thus compete repeatedly against similarly strong players. With almost separate Intermediate & Advanced pools no general ranking is created; good Intermediate players have too many points in comparison, the values are not really comparable among each other. In addition to adjusting the calculation of RGX and introducing different starting values, we would like to mitigate the problem with another measure.

Adjustment to solve problem 1: Introduction of an upper limit for the application

Upper point limits for divisions are introduced. Players with scores above this limit will not be allowed to compete in the respective division. The score at the time of registration for the tournament counts. Both players of a team must – independently of each other – fulfill the criterion with the respective personal RGX. An „offsetting“ via the average score is therefore not possible. The time of registration of the respective player counts.  If a partner swap becomes necessary after the initial registration, the RGX score at the time of the swap will count for the new player joining the team.

The concrete numbers are based on the starting values of the different divisions. The upper limit is the starting value of the next higher division with a buffer of 15 points – this results in the following values:

Beginner: 1.215
Intermediate: 1.415
Advanced: 1.615

Info: The Advanced value is only relevant if a Pro division is offered at the same time. If Advanced/Pro is the highest division, there is no upper limit for this division. (In this case this score is therefore also another possibility for the qualification to the Pro-Division) In the combination division „Beginner/Intermediate“ the upper limit of the Intermediate counts.

Problem 2: Players register for a division that is too high. 

It can also be observed that at some tournaments players register for a division that is (sometimes significantly) above their actual skill level. This has already led to teams that actually fit better into the division ending up on the waiting list and not getting a starting place in the tournament in the end. In addition, the tournament itself lowers the sporting level, the significance of (tournament) victories and the probability of sporting development of the players through games at the same level, which ultimately all players want. This problem can be observed almost exclusively in men’s advanced tournaments (more rarely also in mixed). The problem is compounded when a small tournament offers only an Advanced division and no Intermediate division. However, since we and the organizing communities cannot guarantee that several divisions will be offered at every tournament in the future, we have therefore come up with a solution.

Adjustment to solve problem 2: Introduction of a lower limit for registration

Lower point limits for divisions are defined. Only players above this limit are allowed to register for a tournament of the corresponding division. The score at the time of registration for the tournament counts. Both players of a team must – independently of each other – fulfill the criterion with the respective personal RGX. A „clearing“ via the average value is therefore not possible. The time of registration of the respective player counts.  If a partner swap becomes necessary after the initial registration, the RGX score at the time of the swap will count for the new player joining the team.

Since there are too few beginner tournaments at the moment, a limit for the intermediate division would de facto exclude some people from tournaments, so there is a lower limit for the advanced division only. This is also based on the starting value of the Intermediate Division:

Beginner: 0 (upper limir1.215)
Intermediate: 0 (upper limit 1.415)
Advanced: 1185 (upper limit 1.615)
Another exception is the Pro Division, where qualification is regulated separately.

The choice of upper and lower limits is intended to strike a balance between a sensible restriction of the divisions and the undesirable effect of players being „wrongly“ excluded from a division. The concrete numbers are chosen in such a way that only the really clear cases are affected (e.g. „weaker“ Intermediate players when registering in Advanced). For most players, therefore, this first introduction of lower limits does not yet change anything in practice and personal self-assessment will continue to play a major role. Due to the problems in RGX described above, there is a considerable number of Advanced players who still have relatively few points. We would like to continue to enable them to play in Advanced. For the future, however, it is planned to regularly analyze and, if necessary, adjust these upper and lower limits, so that the limits tend to become narrower and narrower and you have to „play“ your way up to the next higher division.

Specifically, the numbers mean: In Intermediate, approx. 85% of all active players are eligible to play, in Advanced approx. 75%. 61% of all players meet the criteria for both divisions, so they can still decide in which division they want to compete.

OVERVIEW OF YOUR ELIGIBLE DIVISIONS ON THE PERSONAL PLAYERZONE PROFILE
Using the RGX in seedings lists for GRT tournaments

The seeding list for GRT tournaments is created using the RGX. Teams are therefore sorted in descending order according to their summed RGX. Exceptions are teams consisting of players from a lower division or players without RGX. These can be set manually by the tournament organizers in consultation with the tournament department.

Conclusion: The pure self-assessment is supplemented by reasonable measures

Similar to the adjustments to the RGX, the working group spent a very long time discussing, analyzing your opinions from the survey, and weighing what measures to take to solve the problems. The result is the introduction of generously defined upper and lower limits. These will provide guidance to players on how to classify their playing strength into the appropriate division. This means that not everyone will be able to play in every division (as is already the practice in the Pro Division). This step is unavoidable from our point of view in view of the growth of our sport, the sometimes large discrepancy between the demand for tournament places and the corresponding supply and also the sometimes difficult self-assessment. The limits are chosen very generously for the time being, so that according to our first forecast the regulation will affect very few (but very clear) cases. The procedure also gives us the opportunity to gather and analyze empirical values so that we can further optimize the limits in the coming years. 

TO THE TOTAL STANDINGS OF THE RG INDEX
TO THE INFO PAGE OF THE RANKING
Scroll to Top