Cheating! Or not!

Posté par northlecale le 30 mars 2018

Heh! That’s not a nice topic to blog about Mister. Yes, you’re right on that, but it is a topical one. There is an accusation that some cheating has gone on in the PGETC. An observer from one team, who was using Leela to analyze all the games, noticed that one guy in the other team was always playing Leela’s move. It is one of the inevitable things about having really good software that you can easily use for live game analysis, and indeed for live cheating.

  1. Take the new Lizzie program, still in 0.x development stage but very powerful given that you can live sync to LeelaZero.
  2. Then you can just as well use classic Leela 11
  3. Or indeed hook up any other AI to Sabaki, GoGUI, or whatever.

In short, if you fancy cheating, you’ve got everything going for you.

In Chess a guy called Ken Regan came up with a powerful detection model. It’s not a very easy task to prove that somebody cheated. You have to show (obviously) that somebody is playing at a different level to what they normally would. You cannot simply say that so-and-so played 5 moves in a row like Zen. Ken specifically warns us against the belief that a move that is given a clear standout evaluation by a program is much more likely to be found by a strong human playerSEE HERE . Where he talks about how certain forced games can have as their characteristic a higher level of AI choices.

Using the GoR calculator on the EGD website, I did find that the player in question had an increase of about 50 points (half a rank) if it were a level A event – which is half a stone. Looking a bit further, I found that if I moved his rating up to 2600 (+ 2 ranks), I got roughly that his performance was as it was expected to be. That might look like a good piece of evidence. Well actually it alone doesn’t prove anything other than that he played well on the internet for that PGETC season.

In his defense you have to note two things:

  1. The player is clearly improving in over the board play.
  2. They openly admit to studying with Leela (Classic). Not just studying with, extensively studying with for 2 years, with witnesses to back them up.

Whilst we wait for the statistical analysis to play out, I repeat what I’ve said earlier on here. If you want to spend time studying Go with a computer, then there’s never been a better time to do it.

Publié dans Baduk, gossip | Pas de Commentaire »

 

Dylansiloret44 |
Mistymil |
Godemon Games |
Unblog.fr | Créer un blog | Annuaire | Signaler un abus | Freshbad
| Pokémon Rush
| Mechanicat