Lies, damned lies, and gradings
The other day I found myself messing around with grading calculations for my games from the last few years. You can massage the figures to draw pretty much whatever conclusion you like - especially if, like me, you don't play all that many games. When the law of large numbers hasn't had a chance to assert itself, it's just a question of picking out whatever randomness appeals to you. For instance, since taking up chess again (after about ten years out) my grades have gone: 139, 153, 160. (And, albeit after very few games, my performance this season is a touch over 170). What a marvellous story of continuous improvement - at this rate I should be giving young Carlsen one hell of a beating by 2020 or so. Well maybe, but I notice that if rather than considering the grading season you instead start the year at January 1st then you get a completely different story. Then my annual performances have gone: 125, 170, 155, 165. Perhaps Magnus shouldn't be too worried after...