History

The Grading System has been in operation now for over 25 years, is aware of over 1,750 bridge players, over 5,700 partnerships and has absorbed data from over 2,300 pairs tournaments played at four different Clubs and the biannual Worthing weekends. Originally, in order to provide a starting point, about 20 players were allocated a provisional grade, varying from 100 up to 170 for the better players. Since then the system has been entirely automatic - new players finding their own level - and the average grade has remained static at about 135. There have been small refinements added over the years, but the basic method has remained largely untouched.

The general principles

The system is broadly based on that used in the world of chess (where it attracts an enormous amount of interest), but is inevitably more complicated, since to judge a player’s performance at bridge, it is necessary to take into account the strength of his or her partner as well as the strength of the rest of the field. Furthermore, a chess game can have only three results – win, draw or loss, whereas in bridge a pair can achieve anything from 0 to 100% (well, perhaps 20 to 80% realistically). All of these factors are embraced by the system. The method adopted requires a result in percentage form and therefore only match-pointed pairs tournaments are taken into account – no teams or ‘Butler scored’ pairs.

The method in detail

Each time a pair plays, they are given a Partnership Performance Rating (PPR) which is based on only two factors – the percentage they achieve and the strength of the field in which they achieve it, which is defined as the average grade of all the contestants playing in the tournament. In a tournament where the average grade is T, the PPR awarded for 0% would be 0; for 50% it would be T; and for 100%, 2xT. The scale, however, does not slide smoothly between these three points. In order to avoid over-emphasizing the importance of freak scores a sine-curve is fitted. The actual formula used is:

where P is the percentage attained and T is the average grade of the tournament.

Having arrived at a pair's PPR, each of the players is awarded an Individual Performance Rating (IPR) based on the assumption that they have contributed equally to the result relative to their respective current grades. Thus, if a player graded at 120 plays with one graded at 150, they are expected to achieve a PPR of 135. If, in fact, they achieve a PPR of, say, 153 this is18 better than predicted and each player will be credited with an IPR equal to 18 more than his or her current grade, i.e. 120+18 = 138 and 150+18 = 168 respectively.

Regrading

Until September 2005, all grades were recalculated at the end of each calendar month.   Since then the system has operated on a quarterly basis.   If a player has played ten or more times within the system in the current quarter, then his new grade will be the average of all his IPRs for the quarter. If he has played less than ten times, each IPR forms a tenth part of the new grade, the remainder being an appropriate proportion of the old grade. For example:

Old grade: 175. Three IPRs in the quarter: 165, 154 and 192

New grade = (175*7+165+154+192)/10 =173.6, rounded to 174.

New players

When new players enter the system, different methods have to be used since initially they have no grade for the computer to work on. Without going into detail, suffice it to say that the method is still entirely automatic – there is no attempt to ‘guess’ a standard for them. The revised method is used until the player has put in three performances, after which he will be allocated a grade of his own (mid-month if necessary).

Quarterly Lists

If a pair rarely play other than with each other they will rapidly achieve a reliable grading as a partnership but the system cannot be expected to provide an accurate assessment of the players individually. For this reason, two individual lists, 'A' & 'B', are produced each quarter together with a pairs list.

The 'A' List

To be included in this list a player must have competed in at least twenty graded tournaments altogether (at least one of which was in the last quarter) and must have played with at least three different partners in the last six months.

The 'B' list

This lists all other players who have played at least 3 graded tournaments altogether (at least one of which was in the last quarter).

The Pairs List

For inclusion in this list, a pair must have competed together in at least five graded tournaments in total (at least one of which was in the last quarter).

In conclusion

No system is perfect – all have strengths and weaknesses.

On the plus side, there is no human intervention – all grades are based on unarguable, empirical evidence. Furthermore, simply glancing down the current list will show quite clearly that the system has ‘got it about right’.

On the minus side, only match-pointed pairs are taken into account. If you’re a ‘Teams specialist’, sorry! Furthermore, the grades are probably too volatile - my own grade varies alarmingly from one quarter to the next. To an extent, this is deliberate - everyone likes to see their grade show an upsurge after a good quarter.

Finally, it should be remembered that the system rates success rather than playing standard - not necessarily the same thing!. A very good player with a rather casual attitude to ‘the ordinary club duplicate’ might find himself with a surprisingly low grade, and Victor Mollo's Rueful Rabbit would probably top the list every time!

Roger Fellows

Email us with your opinion: brambledown@blueyonder.co.uk