1/a predicted EGD | 1/a observed EGD | 1/a predicted revised | 1/a observed revised

p predicted EGD | p observed EGD | p predicted revised | p observed revised

rating point contibutions EGD | rating point contributions Revised

rating points for game results EGD | rating points for game results Revised |

The horizontal axis shows the go rating scale used by the EGD and the revised rating system.

The vertical axis shows points change.

The upper and lower lines bound the maximum rating points change for wins and losses. It is determined by the **con** factor and the **epsilon** term.

The middle line is the contribution of the **epsilon** term. You may have to zoom in (at the lower ratings) to see its contribution.

The FIDE chess rating system uses a **K** factor, which basically works the same as the **con** factor. They mostly use 20, but for youth players and new players they use 40. For high ratings they use 10. (also see the FIDE rating calculator)
Other chess rating systems use **K** factor 24 with 16 for high ratings and 32 for low ratings.
So the EGD uses a rather high value for low ratings. That causes strong rating oscillations in the lower ratings, which in my view is only noise that obscures deflation effects at the lower end of the rating range.

Perhaps it was hoped that this increased noise would somehow absorb deflation?

Anyway, I chose to reduce the noise by reducing the **con** factor for lower ratings to allow detection of deflation signals.