How to get 97% accuracy

You're going to hate me for this: It's actually not that hard. With the buffer of some context: there are variables within the accuracy calculator that are not our gift to control. Most of the time, of course, that works against us, but just occasionally, it works for us. If you want a 97% accuracy game, you can play for it without the added pressure of trying to get it, because that would be pointless. You can play a game very well, and end up with 80%. The complexity of the position is a determining factor. For example, while I was trawling through my archive fact-checking myself for this post, I came across this game:

My hypothesis was that I had probably scored more than 97% before; probably quite a few times. 97% no - not that I had found in the games as far as I scrolled back. But I have had a couple of 100%s. This is not as impressive as it sounds. I scored 100% on the above game because most of my moves were theory - if you are following Game Review's "book", you cannot lose accuracy for those moves - and once my opponent walked into the opening trap, all of my moves were the best ones. Not hard to find, as they were pre-learned. There were only 9 moves in the game; 6 Book + 3 Best = 100%.The test results (disclaimer: possibly too low sample size) disproved the hypothesis, at least to an extent. I frequently broke the 90% threshold, but never quite got as high as 97% - except for a couple of games where I actually scored 100%, one being the game above. But - oh yes, I should probably explain why I'm talking about 97% in the first place. It wasn't going to be today's topic. I had one in mind (because, just occasionally, I plan my topics in advance rather than rushing at the last minute to hit a self-imposed deadline which, as my wife will attest, is nuts). One reason I had a topic in mind was that I was away for the weekend - I have (had) been in Aviemore since Friday, playing in the Highland Bridge Congress. I returned home today, taking away over two hours' worth that I would normally spend writing this. This would have been fine if I had written the post in advance like a normal person, but since I am anything but normal, I hadn't. The topic I wanted to write about I thought I could flesh out a bit more (stay tuned for an introduction at the end of this post!), so that's coming next week. This week, I have time for a post where I'm only posting about one long game.

This was the analysis summary of the game. Shoutout to wintrcat for making his own chess computer analysis website, so that anyone can analyse as many games as they like for free. (There is nothing wrong with chess.com's Game Review, but this game was not played on this site). Hopefully he gets enough donations to keep the operation going. Since I was trying this out for the first time, I set it to a high depth review. I was Black, scoring 97% accuracy, and I think my opponent deserves a lot of credit too - 87.5% is a great score, and between us there was only ONE inaccuracy the entire game, with no mistakes or blunders either side! It takes two to tango, as the saying goes, and my opponent lived up to his end. More on that later. But there is the catalyst - I had scored 97%, not in a miniature but in a proper-length (32 move) game! 97% over 32 moves is, I put it to you, MUCH harder to get than 100% in 9. Want to see a game where someone scored 100% and LOST?

Again, that was only 12 moves, but well done (and hard luck!) to my opponent. But I expect he was still playing from preparation here; what would his score have been after another 20 moves? Bold of me to assume I could last that long! As I said, 97% in 32 is much harder than 100% in 9 (or even 12!) - of course, 100% in 32 would be even harder, but if I ever score that, I'll report myself to Kramnik! Since that is so far out of my reach, how did I score even 97% in such a long game? Well...

Step One: The transition from Theory to Non-Book Position is key.

You're not losing even a modicum of accuracy as long as your moves are "book", even if they are sub-optimal ones. For example, in the first game I showed, the computer doesn't like the Smith-Morra gambit, evaluating the resulting position as slightly better for Black, until he falls into the trap. But in most games, at some point, the game is going to leave book - the when and how is important. If it is you, then your chosen move will have to be better than what the book suggests - unless of course, there isn't another "book" move, because you've reached the limit of the computer's pre-programmed knowledge. But a lot of the time, it will be your opponent, and you'd prefer there was a concrete way forward to combat their inaccuracy. I'd show how that played out in the feature game, but why spoil things? Instead, let's look at another example:

Step Two: You probably need some assistance from your opponent

I don't mean that in a sarcastic/disrespectful way. I mean, what type of game is it? Is it a tactical extravaganza, where the smallest imperfection is going to be largely inaccurate? A deep, dark forest where there are many winding paths? A cut-and-dry position where you are not asked a lot of questions, and the way forward is clear? If the latter, you have a higher chance of accuracy. Allow my student to demonstrate - he got - you guessed it - 97% accuracy for this game, which for someone who recently achieved his ambition of breaching 1500 Rapid ELO, is pretty darn good!

Step Three: Your rating ought to reflect your actual strength

Game Review adjusts how harshly (or otherwise) it judges your moves based on your ELO. If your rating is higher than it should be, then you will find it difficult to get good accuracy scores, because the engine will mark you down for not finding moves it thinks you should. Lower ELOs - while capable of collecting question marks by the bucket on a bad day - are treated more leniently; advanced positional mishaps are not given question marks. This may not seem like objectivity, but if two 800s are going at it, the chances are the game isn't going to be decided by someone capturing towards the flank when they should have captured towards the centre. So it does make some sense.

TANGENT TIME! Do NOT get obsessed with numbers!

While we all crave that 90%+ accuracy, it simply isn't the be-all-and-end-all. 97% in a game where your opponent blunders frequently and it is pretty obvious what to do will probably not make you feel as good as 85% in a real tough grind of a game where you and your opponent were throwing tactical haymakers at one another, with you maybe not making the best moves, but consistently keeping your edge and coming out on top. I have not checked with him, but I'm pretty sure the student who played the game above will have many games he can think of that he enjoyed more, or is prouder of, than that one. And his GR Rating for that game was 2250. That's always way off base, I'm afraid. As happy as I am with how my student played that one, a 2250-rated player would have played that differently.

Step Four: I mean... you do actually have to play well

The above is an example of some features that high accuracy games have in common, but there is no magic formula. Which now I think about it, does make the title of this post somewhat misleading. The key ingredient is you - but you knew that anyway, surely? Nobody's out there thinking they could blunder their way to a 97% accuracy score by making a certain sequence of moves that breaks the code? (And if you could, would you really want to?)

OK, so here it is...

Without further ado, the game that inspired this post. 2 moves of theory, followed by 30 moves where a real game of chess broke out, but there was a through logic that helped me navigate. I hope that it has been worth the wait.

Next week: Something a little bit different, as I introduce you to another mind sport, and explore the similarities and differences!