Do bots have an accurate rating?

As people have said, BOT ratings are not accurate. It's best just to use them as a fun challenge to see how high a BOT you can beat. But always come back to playing live chess as people will always play a little bit different and this will make you better at chess

I am a 200 rated player, and i have defeated the nelson (1300), and also a IM, he is my frined i have defeated in offline not online

That could be luck.. you need to beat 10 times in a row the same bot.. 
I did that until 1300 so far. It is a good practice. Because beating 1 time a bot could be luck. 
To prove you can beat the bot you need to beat 10 times in a row. 
If you lose you reset, and if it is a draw as well. 
I beat all 10 times. Trying 1400 now.

I can beat 1200 bots and probably would lose to an actual 500 so yes its inaccurate

Not accurate at all in my experience.

I'm pretty sure I could have dusted Martin when I was 250. Low rated bots are unrealistic because their mistakes are contrived.

They just play very differently from human players. Even the strong bots (2000+) sometimes make mistakes which a 1000 rated human is unlikely to make and then other times 1300 rated bots play tactical sequences far beyond what I would expect from players rated 1800 rapid. All in all the mistakes seem very artificial and when they're not "due" to make their mistakes, they will sometimes just keep playing the best moves, no matter how complex they are

It would be very interesting were the bot ratings determined the same way as human ratings, by their successes and failures in correctly timed games against humans. It wouldn't skew the rating system because there are so few bots that they wouldn't inject a large amount of incorrect rating-points into the pool. And like any player, they would quickly level out at a rating that was directly comparable to the human ratings. Of course it would make it difficult to sort them into permanent categories as they are at the moment. And since there is a certain repeatability about bot-style, there will be humans who are good bot-splatters for whom the ratings will be biased one way, and humans who are good human-splatters for whom the ratings will be biased the other way. But it would be better overall.

Incidentally, I agree they're really skewed. I have won against bots up to about 1500 and my actual rating is about 380. So they're out, for me, by a factor of about 4.

No, when I was starting this account as beginner, I could already beat about 1400

When I was not really familiar with chess and openings specifically