Russell Sherwood Saturday, November 4, 2023
ICCF New Rating System
Recently, ICCF moved to a new rating system. Several documents have been issued explaining the new system, but these tend to be difficult to understand for those without a Statistical background.
Below is a brief attempt to give a simple explanation of how it works: If you want to know more, it's back to the more detailed specifications!
The new system was designed with several objectives in mind – some explicit and some indirect benefits – including:
- Recognizing the increasing levels of draws within correspondence chess, especially at higher rating levels.
- Recognizing that the Elo formula utility is flawed within correspondence chess – What I mean by this is that with a typical Elo formula, the WLD statistics in a game between a 1900 and 2000 rated player are the same as those between 2400 and 2500 rated players!
- Recognize that not all results should be measured in the same way.
- That new players are an unknown quality – an 1800-rated player in an event could be very different things in terms of playing strength. This leads to the idea that games should be treated differently, dependent on these kinds of considerations.
- That a process of “leeching” rating points had been observed – players simply aimed to draw out games quickly to gain rating points.
- Several ways to game the rating system have been developed by players
So how does the new rating system now work…..
A player’s strength is now measured using two numbers, Rating and Rating Deviation.
Rating Deviation – What is it?
Rating Deviation is a concept that many players struggle with – it is listed as a measure of the uncertainty of a player’s rating – I prefer to consider it as a measure of a player’s activity.
In simple terms – the more games you play – the more games you finish – the lower your Rating Deviation will be.
When we think about this, the more games we complete, the more accurate the rating will be as a measure of our actual playing strength. Going a little deeper, the more games we have completed recently is also a better indicator of our playing strength – so the games we played 10 years ago have very little relevance on our rating today.
A new player will start out with a very high RD (250), as they have completed zero games, but this will drop as they complete games and their rating becomes more stable.
Rating Deviation – How does it affect rating calculations?
If we examine the “Beginners Guide” we see that the level of RD has an impact on the benefits a player gains from a result.
The lower your RD, the less benefit (or loss) you will gain from a result as your RD increases.
The converse of this is also true.
The higher your RD, the more benefit (or loss) you will gain from a result as your opponent’s RD decreases.
What does this mean – if you play a new starter or an inactive player returning, those results will have less impact on your rating but more on their rating than those against players with established ratings.
Now we know the impact the Rating deviation has on the rating formula, let us look at the formula itself.
- The new formula calculates different percentages for Win/Loss/Draw
- The benefits for these results are different depending on the ratings of the players – In simple terms, the higher the ratings involved the higher the benefit of a positive result.
The percentages for these results were based on several, recent, year's results.
What does this mean? Below are the charts used in the “Beginners Guide”.
The first shows the likely outcome based on the player and opponents’ results for a 1500-rated player, and the second shows that for a 2500-rated player.
Without going into any mathematics, we can see that the results curves for Win/Loss/Draw, whilst similar in shape, are pegged at different levels. The main takeaway from this is that draws become more common the higher the rating involved, and the consequence of this is that those draws have less impact on a player's rating. Decisive games become less common as ratings increase, so the benefit of these results also increases.
The Impact of the Rating Deviation is to modify these benefits as outlined in the RD section.
ICCF have two useful resources – the first is the details of the rating formula – If you really want to delve into the mathematics of the system, read this. If you do this, I suggest putting your own results in to work through. The second is the Calculator, which allows you to plug in numbers – do beware, this is for illustration only, rather than definitive.
As a final note - it is now incorrect to refer to an ICCF rating as an ELO rating (Capitalized or not) as we have moved away from the ELO system - it is now simply ICCF rating.