It's not you, it's me

A week ago I participated in my first over-the-board tournament since the initial wave of COVID, a period of almost two years. During that time I worked very hard on my chess and my online ratings at slow time controls improved by a couple hundred points, so I was looking forward to some better OTB performances. I even signed up for the U1800 section rather than the U1300 that I qualified for because I wanted to play other players who were better than my USCF rating of 1265, not ones who were worse.

Of all the ratings you can have these days, with different websites and different time controls and whatnot, I value my official USCF over the board regular time control rating the most. It's the only one where you can be reasonably certain that your opponents are treating the game seriously and not cheating or sandbagging, and besides that it's the standard (in the US, at least) against which online ratings are compared. If I tell someone that my rating is 1900-something on Lichess at Classical time control, the question that usually follows is "How does that compare to a USCF rating?"

The answer to that is apparently "Very poorly." There are a lot of different ways people have tried to compare online ratings to official USCF ratings, and depending on whose method you choose to believe my online ratings on both Lichess and Chess.com in slow games should translate to a USCF rating anywhere from 1400 to 1700. So I was looking forward to a chance to raise my rating. Yeah, I would be playing people higher rated than me, but I had four games to play and the numbers were telling me that I should be higher rated than I was anyway. If I could even win one game out of the scheduled four it would be a nice confidence boost and my rating would probably increase even if I lost the other three.

I ended up getting a bye in one round so I actually only played three games, and the pairings gave me a reasonably soft schedule; I only had to play one opponent with a rating over 1500.

It didn't matter. I lost all three games. I was doing well in the first two, but got into time trouble that caused my game to go downhill, the same problem that has been affecting me since I returned to tournament chess a couple of years ago. In the third game I played horribly against an opponent better than me and I resigned on only my 16th move. My rating plummeted another 41 points to 1224.

My spirit was broken. How can I be losing all these games? Every outward sign tells me that I should be getting better, but the results just keep getting worse. When I first started playing tournament chess in 1992 I couldn't understand why players rated around 1100-1200 weren't better. I wondered how anyone who devoted any time at all to this couldn't be better than that. I had only started playing a year before, and did no training other than reading chess books, and yet I could already beat them. But now here I am, 30 years later, and I'm basically at the level for which I once had contempt. But I don't just read chess books now. I train. I do tactics exercises and complex puzzles and I spend hours and hours analyzing all my games. I spend more time on chess than any other activity and have devoted more than an hour a day to chess for literally years, and I'm worse than I was before.

I don't feel worse. I feel like I'm a lot better than I used to be and my online results support that. But in official tournaments it seems like I can't beat anybody at all.

So I started thinking about this. Maybe I really am a worse player than I used to be, but it seems like it's the opposite; my opponents seem like they're better than they used to be. Is there a way I could test that and find out? It seemed like in the old days I used to have no trouble beating people below about 1300 and sometimes I'd beat people higher than that. Now I struggle to beat anybody. But maybe it's not me, maybe it's them. So I decided to try to find out.

I don't have a database, but I did keep all my old tournament scorebooks dating to sometime in 1994 (I had been playing tournaments for two years at that point but nothing prior to 1994 survives). I decided I would have Stockfish run a quick analysis on all those games and then compare the results by era. I've had three different periods in my life when I played tournament chess: 1992-1995, 2002-2009 and 2018-present. It was quite a bit of work putting all the games in, and a quick computer scan is limited in usefulness but unless I wanted to spend months on this project (and I didn't) I had to make some compromises.

One of the things I discovered from assembling this data is that at least at this level, Inaccuracies and Mistakes are meaningless. How often they occur doesn't affect the Average Centipawn Loss or the result very much if at all. But I tracked them, so I included them in the table. Blunders and Average Centipawn Loss tell you a little bit more, but I found a lot of cases where they made no sense either. Games with a lot of sharp tactics tend to have a lot more blunders and higher Average Centipawn Loss than games that don't, and accordingly the system will regard those games as of lower quality when that is not necessarily true. I think in the aggregate some of those things will be smoothed over by the averages if there are a sufficient number of games.

Here are the results. I grouped the games not just by era but by rating range. Other than the record, these numbers are all for the performances of my opponents, not myself. I also tracked my own numbers, but the blog is getting cluttered with all the numbers so rather than reproduce those numbers in a chart I'll just mention some of them in the text below. Note that when I list records I put them in the order Wins-Draws-Losses.

As mentioned above, the Blunders and the Average Centipawn Loss tend to decrease as ratings get higher, which they should.

You can already see the difference in my results. In the 1994-95 period I was very good against any opponent rated under 1600, losing only one out of nine games, but couldn't beat anybody above that. By the 2002-09 period the level at which I couldn't beat anybody (with one exception) had dropped to 1200. The numbers suggest that opponents between 1200-1599 were stronger than they had been in 1994-95, at least in the games I played. The Average Centipawn Loss for the 1200-1399 level had decreased from 62.7 to 38.5; for the 1400-1599 level, from 52.0 to 43.4.

It's also noteworthy that by then a lot higher percentage of my opponents were rated below 1200.

Not that different than the chart for the 2002-2009 period. Lost every game against opponents rated over 1200, won most against under 1200. Of the two games I lost against U1200s, one was on a blunder for which I have no excuse; I just hung a piece for no reason. In the other I had an easily winning position but did not realize that my clock had ticked down to one minute left. In my attempt to force the win before running out of time I committed four blunders and went from being +10 in the computer evaluation to getting checkmated in four moves.

The high number for Average Centipawn Loss for the 1200-1399 range resulted from two games with a lot of tactics. I had a winning position in one and a drawn endgame in the other; I lost both games on time forfeits.

Time trouble has been a persistent problem for me since 2018, which is strange because it had never been an issue prior to that. In 23 tournament games played since 2018 I've lost 13 games, 6 of them by time forfeits and many of the others because of blunders in time trouble. This is not a revelation, obviously, and I know I need to move faster. I've been working on it in my online games and have made a lot of progress, but in OTB games it just seems like things get complicated early on in every game, so I slow down. Other than trying to match my online time controls with the ones I'm likely to have in OTB tournaments I don't know what to do about it.

But back to the point about whether I'm getting worse or whether my opponents are getting better. What did the numbers say about my own performances by era? I'm most interested in the performance numbers against opponents in the 1200-1399 and 1400-1599 rating ranges. In 1994-95 my Average Centipawn Loss was 56.3 against the former and 38.4 against the latter, which was good enough for a combined record against those two groups of 5-2-1. After 1995 I didn't win those games anymore, going a combined 0-2-5 from 2002-2009 and 0-0-6 from 2018-present, an almost complete reversal. Yet my Average Centipawn Loss against those groups wasn't as dramatically different as you'd think: 49.5 and 69.2 in 2002-2009 and 86.0 and 39.0 from 2018-present. In both cases one of the numbers was higher (meaning worse) but the other was essentially the same or even slightly better. Objectively speaking my performance was somewhat worse, but it's hard to believe that difference would be enough to go from winning or drawing almost all those games in 1994-1995 to losing almost every one (and winning zero) after that.

What the numbers seem to be telling me is what I already knew: it's not that I'm making worse moves than I used to, it's that I'm taking a lot longer to make them that's the problem. I get good positions in most of my games but it takes me too long to get there, and I either end up losing on time or blundering at the end trying to avoid it. In that sense, my falling rating is a product of the fact that yes, I actually am getting worse. But that doesn't have to be the whole answer. While I don't think the above analysis proves it, I still think that my opponents are in fact better than they used to be at the same rating levels in 1992-1995. I'm less convinced that there's been much change since 2002-2009. I think my understanding of the game has improved since then, but the results haven't followed for me because I haven't kicked the time trouble problem. Until I do, my rating will continue to fall.

It's not easy to deal with it when you spend an enormous amount of time and effort on something and not only do you not improve but you actually regress. Fortunately the regression has been confined to my official OTB tournaments and I'm doing fine in my online games. If it weren't for that, I don't think I could cope with the failure. It's bad enough even as it is; I've been moping around for a week. The more I lose, the less I want to go to another tournament and lose again, which will make it hard to ever get my rating back up.