Player Statistics/Rankings

[Julian Dale]

I run the Cambridge and District bar billiards website (www.cambridgebarbilliards.co.uk) and have put all of the match data into a database so that the fixtures, league table and match results can be calculated and updated by the website automatically. This includes the individual player scores in each match.

Johnny George, who ran the previous incarnation of the website used to produce a player stats table listing things such as games played, number of wins, win percentage and average score. There was some debate within the league as to how the stats should be ordered; some preferring to use win percentage, some preferred using averages and the winner of the Under 30's Trophy was the player who had the most number of wins. This caused controversy in recent years as I narrowly won the Under 30's Trophy, having played two more games than my nearest rival (who happened to be Johnny George). In fairness, I had a higher win percentage, but it was not fair because Johnny could have had the same win percentage were he able to play in the same number of games. So then, how do you make it fair?!

I thought a ranking system could be the answer to the question of fairness. In tennis you have the ATP Tour rankings, golf has the PGA Tour ranking system and other sports have differing ranking/seeding systems to compare team and player standings. In bar billiards, I have seen there are already some tournament rankings that exist, but what I am looking for is a system where players can be compared on their performance in a league season, ignoring performances in knockout cup competitions.

My initial idea is to apply ranking points for each game played (with a differing number of points given depending on the circumstances of the game and the actual points result), then take an average of these ranking points.

An example would be that you would get more ranking points for winning away from home, you would get more for winning against the break and you would get more for scoring over a certain number of points. That would work fine after a little tweaking of how much importance to give to the ranking points (in our league, scores are relatively low and playing against the break isn't so much of a factor). The drawback is that it wouldn't be a fair comparison if you took a player who has played 5 games and won them all and compared that player with another who has played 18 games and won 15. In my mind, winning 15 out of 18 is more difficult than winning 5. So, the ranking points system has to take into account the number of games played... somehow!

So, does anyone have any ideas about how you could statistically rank the performances of players over the course of a league season in a fair and consistent manner?

[NigelS]

Hi Julian, website looks good.

Ranking systems do vary across leagues but tend to follow the same themes. There is no right or wrong method, but as long as you can justify why points are allocated a certain way and everyone knows how they get points no one can moan afterwards.

I calculate the Brighton League rankings, points are awarded for league wins and also competitions, extra points are awarded for high score and wins against the break. More points are awarded in higher divisions than lower divisions.

For your info here is a link to how the Brighton rankings work

www.brightonbarbilliards.co.uk/ranking_system.htm

[iang]

Hi Julian.
I'm fixtures secreatary for The Horsham League & we operate a ranking point system which is a score bonus with a starting point
of
0 - 2990 1 point for a win
3000 - 4990 2 points for a win & 1 point for a loss
so you can still gain a point if you lose this goes up with the amount of score.
that is used to calculate a masters list, but the player of the year is based on win percentage for both Divisions.

These decissions are usually decided at the AGM's & cannot be changed by one person.
Are you on the committee or just taking over the website ?

[davejones]

The first port of call for any changes should be to the person who donated the trophy to see if they have any objection to the changes you wish to make.
Some of the Worthing league trophies were donated by beneficiaries for specific competitions or cup competitions.

The Worthing league run two trophies alond the lines you mention.
First the performance trophies (separate for div 1 and lower div) which is based on most wins then total scores.

Secondly, is the ranking point system used for qualification for the masters.
League matches are as follows:–
3pts away win against the break
2pts away win with the break
2pts home win against the break 1.5 pts home win with the break
There is also extra points gained for the various individual comps.

[Chris_Sav]

The programme suite that I wrote for Dover Deal works on a simple ranking system

1 point per percentage of games won and one point per thousand scored (aggregate). Half that in pairs games.

That's also the standard for Kent rankings from interleagues.

We've been using the system since Windoze 95 days!

Sav.

Hi Jules

I have looked at your site with interest and can I say you have done a great job with the layout and the smoothness of the site.

I think the ranking system you show on the site is the best one, the reason being is that you will get no conflict with it, if you introduce technicalities and points for score then someone might feel more inclined to stuff their opponent.

I on Monday played as we only had 4 turn up so I played, I tapped up after a score of 6k however Liam can get 9,7k without tapping up as he is in a higher and more copetitive division. If I tried that I can just picture another AGM with my name on the agenda (again). So it would be unfair to give points to players who score more, unless people don't mind me going for top prize (which they do).

The ranking system you have is perfect, wouldn't change it. Also your website is looking great so keep it up and look forward to seeing the Nelsons mount a succesful challenge on the current champions (which included me shamefully :-[ :P) but i'll fight my corner after christmas against you ;D ;D ;)

In a way it was good to see SMB stood their ground with a 3-2 win gotta say, wouldn't want to predict any winners yet!

[Julian Dale]

Just to clarify, I'm not suggesting a change of the criteria used to award the Under 30's Trophy, it was just an example of why I think a fairer system was needed. I'm not on the board, but if I can use my computing skills to suggest a new way of comparing players then maybe that's something the league can introduce at a later date. For now, I'm interested in producing a new column on the player stats page so that people can choose to re-order the stats by things other than win percentage.

As for Johnny's point about encouraging players to stuff their opposition, I think the ranking system could be developed to discourage this sort of thing, or at least offer no incentive to score more than say 4000 (which in the Cambs league is enough to win the majority of division 1 games).

Using win percentage to rank players is good in that the better players will win proportionately more than a weaker player, but it has many flaws:

Firstly, if a player wins 5/5, is he better than a player who wins 17/18? I don't think so.
Second, if a player wins 10/10 but wins all his games in the second division, is he better than a player who wins 9/10 in the first division? No chance...
Third, if a player wins 10/10 in the first division but averages 2000 (thus he beat weak opponents), is he better than a player who has won 9/10 in the first division but has averaged 5000? Not in my book.
Fourth, if a player wins 10/10 but only plays home games (on a table he is familiar with), is he better than his team mate who wins 10/10 by playing all of the away games (on tables he is not familiar with)? No way.
Fifth, if a player wins 10/10 but is lucky enough to get the break in all his games, is he better than a player with an identical record who had to play against the break in all his games? No.

The win percentage is the best option we have at the moment so I will keep it as such for the time being, but I would still be interested in developing a system that compares players on more than just wins.

With respect, the thing I will say about the systems that other leagues have introduced is that they are not complex enough to adjust for the issues that I have highlighted above, and they are biased towards the players who play in all of the competitions. The Cambs league is not nearly as strong as those leagues, so some of our teams rotate players each week to give them a game, but many are short of players and have to draw a player to play twice on a match night, so I would also be looking for a system that doesn't put so much emphasis on the number of games played. I appreciate that the reason for this may be the volume of data that would need to be calculated if the ranking system was any more complex, but given the opportunity the Cambs league has with the results database, if there is a better formula out there for ranking players, I want it!

[iang]

A very good reply Julian.
Where have you been hiding ???

[Julian Dale]

Oct 16, 2008 16:16:20 GMT iang said:

Where have you been hiding ???

I've browsed the forums occasionally in the past but have never needed to post anything until now. Doing the Cambs league website has given me something interesting to work on, simply because I wasn't happy with the current options for sorting the player stats table!

I've got a couple of formulas that I'm working on at the moment that I'm hoping will persuade some of the purists that a ranking system is needed. The only trouble is, without a whole seasons worth of data to work on it's very difficult to quantify whether the formula is fair, so it may be a while before I have something worth showing.

[Julian Dale]

Just for info, here is what I came up with:
www.cambridgebarbilliards.co.uk/ranking_system.asp

It seems to be working well, with several county standard players either at the top or thereabouts. Time will tell if I make it into the top 10, as my season has been pretty poor so far!

[BB Warrior]

Having recently taken over the Sussex Website and helped to review the Ranking Point System we will be using this year, I was very interested to read about the way other Leagues operate their systems.

Julian, your system is obviously the result of a lot of thought and hard work. 8-) ;D

The only down side I would say is that possibly it is too complicated..... ::) ???

Among our main objectives in Sussex was to try to make the system both fair to players of all levels who play in Sussex Competitions and also easy to understand so that everybody knew immediately how many points they had earned from each match. ;)

Although, using the formula that you devised it works out the points earned by each player....... if I were a player I wouldn't have a clue as to how I had done until it appeared on the website. ::)

But, I will be very interested to see the end of season figures and see how your system would compare with "usual" ranking systems.

Good luck and keep up the good work. ;D ;D

[Julian Dale]

It does look complicated at first, but that's only because I've tried to cater for as many factors as I could quantify. You could strip it down to make it a lot simpler.

For example, you could specify a minimum number of games to be played in order to be ranked, rather than have the part that adjusts for the number of games played. That would be fine if most of the players in the league played the same number of games, but in the Cambs league there can be big differences.

You could ignore the part about what division your opponent is in. That would mean a good player in a lower division is likely to get a higher ranking than a good player in the top division, but then you could always split the rankings by division instead of comparing them all together. In our league that's not really an option because we have two divisions and yet we have league matches between division one and division two teams, so I had to find a way to factor that in.

I don't think you can leave out the factor that is playing against the break. The Cambs league is not so strong, so this isn't really that important, but I would imagine it would need to be worth a larger number of ranking points in stronger leagues, where the break is a fairly decisive factor.

So, if you stripped the system down, you'd be left with:
- Points as to the game result (win, lose or draw),
- Points as to how heavily the player scored in the game and
- Points according to whether you played against the break
- A larger amount of points for the same criteria if you are the away player.
You could then compare people according to their average ranking points per game. If people could see how many ranking points they had in total and how many games they had played, they would be able to easily tot up how many they had gained in their match and with some basic maths would be able to work out their new average per game. I think it needs to be the average ranking points per game rather than the total, as otherwise you would have the players who have played the most games at the top, rather than the players who have performed the best on average.

I'm not sure what a "usual" ranking system is, but I'd be happy to mock up some statistics tables using different formulas if you'd like to see them as a comparison. They might be fairly difficult to assess though without really knowing which players you would expect to be ranking highly, but it's easy enough for me to try. Let me know if you want me to do that, and obviously what formula/numbers to use in it!