While considering other issues around the UK Ratings recently, the question of whether 1- and 2-point matches should be included came up. We’re aware that there are some inconsistencies in the data we have – for example, some tournaments don’t routinely include one-pointers used as quick playoffs when they submit their results, some do. We would be interested to hear your thoughts!
First of all, how common are such short matches? Here’s a current breakdown:
1-pointers make up about 7% of the total and 2-pointers are pretty negligible. (We were a bit surprised to see tiny numbers of other even match-lengths appear as well!).
Where do they come from?
And if we took them out, what would be the effect on our top 20 players? Here’s before:
… and after:
As you can see, there are very few changes and noone moves by more than one place.
Possible reasons to exclude 1- and 2-pointers:
- We are not sure they are not reported consistently.
- The ELO formula is believed to over-state the value of 1-pointers a little…
- … and 2-pointers rather more significantly, given that a 2-point match played perfectly should almost always be played as a single DMP game. FIBS used to, and probably still does, bar 2-point matches on these grounds.
- The lack of cube action reduces the skill required to win.
- The short match length increases volatility and reduces the skill required to win.
Possible reasons not to exclude 1- and 2-pointers:
- It’s a rather arbitrary exclusion – the effect of skill doesn’t suddenly start at 3 points, it increases progressively with match length.
- The ELO formula does compensate for the increased volatility of such games.
- In actual practice, their presence doesn’t appear to distort leading players’ ratings at all, so there’s no gain in accuracy to be had.
- We shouldn’t make assumptions about theoretical best play for 2-pointers.
- The best solution to inconsistent reporting is to be absolutely clear as to the policy for 1-pointers and make people aware of it.
- The point of the ratings system is to gather as much data as possible on games being played. Excluding shorter games disproportionately affects some clubs and certain players. It also makes it harder for players to build up sizeable experience totals, which are necessary to make rankings more accurate.
There are probably other arguments in either direction – please do make them in the comments!
We’re not going to change our policy right now, which is that we welcome all match lengths in the ratings system, but we would like to sample people’s opinions as part of our ongoing review of ratings in general. So, we have set up a Surveymonkey page – we’d be grateful if you’d take a minute or two to fill it in so we can gauge members’ feelings. Thank you in advance for your help!