Here's a simple algorithm that should help contribute to bot detection. You guys might already use it, but if it not...
The move times for a bot are rather predictable because of the nature of the algorithms used.
What one notices quite often is that when the opposite player takes a while to move, the next move of the bot will be quick. This is because it "thinks" while waiting. Much like a human player, but different.
What we also will note that for bots of significant strength, they will tend to play much better moves the more time they get.
Hence, if there is a strong trend along these variables, it can overwhelmingly suggest a bot. Humans also play in similar ways, no doubt, but they are not as meticulous as a computer and computers never make mistakes, unlike humans.
Other indicators are when they very quickly make moves that are dead on and are "booked".
Basically, one can make a pretty good bot detector by looking at the innate characteristics of how a bot works. The algorithm is key and figuring out characteristics of it is what helps distinguish it from a human. A human may apply similar principles but there are differences, and for most players, the differences are significant from a bot(if not over one game, at least many).
Of course, it gets harder and harder to detect bots when humans become better and better chess players only when they use a similar algorithm or the chess algorithms adapt to look more human.
I imagine with a look work, bot detection could be 90%+ accurate and grow near 100% over time. Each game a player makes could give a "bot rating" for both players.
e.g., it is 83% likely that white is a bot with a std of 3%.
No truthful player would mind being compared to a bot. If you play legit and someone thinks you are a bot, that should be a compliment.
So, it's usually pretty easy to tell who uses a bot by communicating with them. Not 100% but if they have a 98% bot rating and get pissed when you accuse them, then chances are they use a bot.
The move times for a bot are rather predictable because of the nature of the algorithms used.
What one notices quite often is that when the opposite player takes a while to move, the next move of the bot will be quick. This is because it "thinks" while waiting. Much like a human player, but different.
What we also will note that for bots of significant strength, they will tend to play much better moves the more time they get.
Hence, if there is a strong trend along these variables, it can overwhelmingly suggest a bot. Humans also play in similar ways, no doubt, but they are not as meticulous as a computer and computers never make mistakes, unlike humans.
Other indicators are when they very quickly make moves that are dead on and are "booked".
Basically, one can make a pretty good bot detector by looking at the innate characteristics of how a bot works. The algorithm is key and figuring out characteristics of it is what helps distinguish it from a human. A human may apply similar principles but there are differences, and for most players, the differences are significant from a bot(if not over one game, at least many).
Of course, it gets harder and harder to detect bots when humans become better and better chess players only when they use a similar algorithm or the chess algorithms adapt to look more human.
I imagine with a look work, bot detection could be 90%+ accurate and grow near 100% over time. Each game a player makes could give a "bot rating" for both players.
e.g., it is 83% likely that white is a bot with a std of 3%.
No truthful player would mind being compared to a bot. If you play legit and someone thinks you are a bot, that should be a compliment.
So, it's usually pretty easy to tell who uses a bot by communicating with them. Not 100% but if they have a 98% bot rating and get pissed when you accuse them, then chances are they use a bot.
