Something I’m missing here after a game is finished is having an accuracy score, let me explain, and sorry, I will be comparing to chess applications
In Lichess they use a system call average Centipawn loss. Which to my understanding is the average of how many points each of your moves are below the optimal move ( 0=perfect game). I see in the analysis that you are shown several potential moves with one at 0 so I assume there at least theoretical a way to derive this.
ChessDotComm has a percentage accuracy score. Which is similar but you are given an overall percentage i.e. a good game could be above 90% depending on your level etc. ChessDot has other things baked in but is not public I think.
So, what this does is it allows you to see if you won because you played well or lost because you opponent played poorly and vice versa. You can also benchmark your accuracy over time. i.e. Me as a noob would be happy getting 70% accuracy or say a 1.5 point average CentiStone loss. From there I can track how I improved to 75% or say 1.2 Centistone loss.
I doubt I’m the first to bring this up so if there is a previous thread kindly let me know.
AI Sensei can do that. Here is a report from one of my games.
(What is called “inaccuracy” or “blunder”, etc. depends on the level you determined ; for me I set it at 1d.)
Great! Just what I’ve been looking for thank you
Katrain can also do performance analysis
This kind of feature was already requested in the past.
Main argument against it, is that cheater could use it to better hide from the cheating detection.
I understand the point and don’t want to question the moderators judgement here. But if a cheater is going through the trouble to cheat and wants to avoid detection, they are probably already using Ai Sensei or Katrain for this. Plus, the cheater that uses AI for every move will be caught, its the smart one’s that only use it a couple of times in a game at a critical point that are the real issue and this performance report with have limited impact.
It would be a cool feature, but I’ll analyse my games on another page which seaways into the ongoing conversation of GO needing a ChessDotCom. I play Correspondence on OGS, Rapid 9x9 on GO quest, study on GO magic, analyse on AI sensei, read material on Sensei library. Ok its an ecosystem, but thats a lot of different places to go. I’m not complaining here, just that to grow the game in the west convenience is key.
I don’t think cheating is a real reason not to add it. Like showing analysis doesn’t necessarily expose mods’ methods of detecting cheating.
I think the harder problem is showing these numbers in a way that is actually helpful. I think AI Sensei’s is probably a good Proof of Concept though.
For those interested: https://github.com/online-go/online-go.com/blob/4ed60176f8fe21960376b663515f4721dad22ab1/src/components/AIReview/AiSummaryTable.tsx. a big piece of this exists, i think it would need more polish before exposing it to users
Human players are never accurate, they make a lot of mistakes. If they want to reach an accuracy level of 70 or 75 percent accuracy, you have to read deeper and remember that chess is easier than Go, even at its accuracy level. If a chess player has an accuracy level of 97-90 percent at best, the accuracy of skilled players can reach up to 80 percent, and that does not mean that he is an AI player. Players who are used to AI can be detected in Go. If a player does not drop to 75-80 percent accuracy level at all stages of the game, then it is possible to assume that something bad happened in that game.
This topic was automatically closed 91 days after the last reply. New replies are no longer allowed.