Online chess is broken, I assume you’re cheating until you prove me wrong

@AlexiHarvey said ^

Nowadays I only play in a paid federation tournament, which brings a more human feel to things.

Indeed OTB tournaments have a cool atmosphere.

Really what's the difference between an anonymous player and a Bot?

cut

PS: I suspect chesscom is thinking in this direction as well, given the amount of money they are throwing at Bot development.

Nope. Btw 'bot development' is simply pasting a cartoon photo onto varied levels of drunk Stockfish lol.

The bots are there to extract money from children.

Should have made clear I meant a paid on-line federation tournament - roughly 30 players across 50+ games in a Swiss format.

Bot development is surprisingly quite different from Stockfish. Essentially the standard move search engine is coupled with position evaluations determined only by a neural network trained on a subset of human games (apparently LiChess Blitz ones!) - each Bot is then given a selected and varied opening repertoire. The interesting thing is that the strength and playing characteristics of each unique Bot produced is completely emergent, so the Bots have to play against each other to get there relative strengths and subsequently there is backward benchmarking with human play to get approximate human elo ratings, on top of everything else there is a whole categorisation system required with regards to play style. The details can be found on the Chessiverse platform, suffice to say you're not looking at detuning of a single 'engine' optimised for playing strength. Based on my own limited experience the results of the whole process would easily pass a chess-based Turing test - in fact, as I hinted above they are even more human-like than humans!

As for 'extract money from children' the platform is indeed very colourful*, but I think this is a little unkind - you can play a limited set of Bots for free and the subscription is a lot less than say chesscom. It's also important to note that as you can only play Bots with no chat service, there is absolutely no safety concerns for parents - nor kids being nasty to each other! The only real niggle in this area is that they are struggling to produce Bots in the sub-800 elo range, currently there are no obvious kiddie features.

*The colour variations can be detuned in the settings, I have left as-is for now.

They are not human like whatsoever. If you had played them you would know. Bots make mistakes because they are programmed to. Humans make mistakes because you are putting them under pressure.

I can assure you I have played many Bots.

Regards deliberate mistakes: I have detailed how the Bots are constructed, there is only one area when mistakes could be 'programmed-in'. When a unique opening repertoire is assigned to each Bot the base is taken from typical LiChess human players for the given elo level - clearly this is likely to incorporate popular human opening mistakes. The reality is humans have a higher expectation of Bot gameplay than that of other humans!

A good test would be to seed the player base with a few 'hidden' Bots and see how many accusations of cheating they gather. From a commercial point of view the site could then ignore any further accusations from such human players! Which does make you wonder a bit.

Data from chesscom indicates that the likelihood of humans detecting real cheaters is as low as 2 to 4% - as such chesscom doesn't even look at a human flagged games.

If humans are rubbish as spotting cheaters what would that suggest regrading their ability to spot or even assess Bots.

I built a plugin that scans the status of all the players I have met. Does not look good.

9823 Chess960 Games
4401 Unique Opponents
77 ToS Violations
339 Closed Accounts
1.7% % ToS Violations
7.7% % Closed Accounts
9.5% % Flagged Combined (ToS + Closed)

@Lumbis said ^

I built a plugin that scans the status of all the players I have met. Does not look good.

Extremely interesting @Lumbis - Is there a chance you can share this plugin?

@Lumbis said ^

I built a plugin that scans the status of all the players I have met. Does not look good.

9823 Chess960 Games
4401 Unique Opponents
77 ToS Violations
339 Closed Accounts
1.7% % ToS Violations
7.7% % Closed Accounts
9.5% % Flagged Combined (ToS + Closed)

You literally assume that all closed accounts got banned. That's very much wrong. Also a good share of ToS violations is due to rating manipulation and not cheating.

@Cedur216 said ^

I built a plugin that scans the status of all the players I have met. Does not look good.

9823 Chess960 Games
4401 Unique Opponents
77 ToS Violations
339 Closed Accounts
1.7% % ToS Violations
7.7% % Closed Accounts
9.5% % Flagged Combined (ToS + Closed)

You literally assume that all closed accounts got banned. That's very much wrong. Also a good share of ToS violations is due to rating manipulation and not cheating.

I made no specific claims, just posted stats. What you says about ToS is true in that all are not cheaters.

"9.5% % Flagged Combined (ToS + Closed)" reads like you considered all closed account flagged. Which is again wrong if you use "flagged" as "banned for being guilty of cheating" or "banned for being guilty of breaching fairplay"