A bot among us?

jere · 2015-02-12 22:23:26

Some of you may have suspected we have a bot in our midst. The one that shall be named CLOCK FORM. It's an intriguing thought. I figured we would have bots eventually, but I didn't expect to see them so soon. Besides, Clock is losing money at a steady clip. Why run a bot that reliably loses money?

Anecdotal evidence seemed to support the suspicions. Clock plays very poorly and somewhat predictably. They've played a lot of games, though this isn't a smoking gun (since my twitter bot has been collecting data, Jeopardy Alcohol has played more).

So today I realized that I had data that could help answer this question. Not definitively, but it's pretty convincing when combined with the other data points. I'm able to look at the times when people play games. It's useful in several senses, but let's apply it to the question at hand.

This is a graph of Jeopardy Alcohol's games versus the hour played (again, only picking on them because of their high number of games):

Now here is the same graph for Clock Form:

Notice anything strange? JA sleeps, but Clock Form doesn't. There are lulls throughout the day, but I'm betting those are just times when hardly anyone else is playing. CF is an enigma, but I'm at least fairly confident they're not a person. It opens up more questions. Who created the bot? Do they also play manually? If so, do they use CF or a separate account? Are they gathering data to improve their operation? Is practically every person playing Cordial Minuet a programmer?

I don't have anything against bots, but I did find this interesting. I hope you did too!

P.S. Isn't CLOCK FORM a particularly great name for a bot?

shallow_depths · 2015-02-12 22:57:53

I definitely suspected the same, especially when they still raised when the graph showed clearly they had lost that round. Mistakes happen but it was pretty suspicious. This was a while ago, mind, I think they have tweaked their strategy a bit since then.

It's certainly intriguing.

Last edited by shallow_depths (2015-02-12 22:59:28)

Cobblestone · 2015-02-12 23:45:08

This is an excellent analysis, and I absolutely agree. I've played several games with Clock Form over the last few days, and its strategy rarely changes, and is very predictable. Its betting seems to follow a very narrow set of patterns based on what you give it and what you bet. For example, last night I played a game where I was able to make it fold at the first betting stage every single time. It folded itself down to 0.

Also, ya, I had what shallow_depths described happen several times.

AnoHito · 2015-02-13 03:02:36

Interesting... Well, when you think about it, right now the cost of training a bot pales in comparison to what you might make with a bot that wins 51% of it's games. I'm not really surprised that someone is attempting it. I considered doing it myself early on. But the more I looked into it, the more I realized what a difficult task it would be. Playing Cordial Minuet well is really more about human psychology than anything else. It's about finding behavior patterns, misdirection, and manipulation. None of which are things any existing AI can do very well. With the current state of AI, I think it would be next to impossible to write a Cordial Minuet bot that could constantly beat skilled human players. I'm not sure, but maybe they are thinking that when the influx of new players comes after the game is publicly released, they bot will be able to perform well enough against to inevitable flood of unskilled players to make money. Though personally I think they may just be programming the bot for the sake of the challenge, since realistically they would have a very difficult time making enough money with it to justify the time they'd have to spend on it.

Incidentally, doesn't "Clock Form" sound like the perfect name for a bot? Cordial Minuet's name generator is remarkably astute sometimes.

computermouth · 2015-02-13 03:35:06

Pretty sure I've lost to that guy. #ECONOMYMAKER

.. · 2015-02-13 05:20:20

Cool! Nice detective work.

It's not my bot. I did some game-theoretic investigations and started work on a bot but haven't touched it since last year. It would play better than that : ) I was also doing it just for the challenge.

Well, when you think about it, right now the cost of training a bot pales in comparison to what you might make with a bot that wins 51% of it's games.

Right now? Regardless of when you're running the bot, you can train it for peanuts on 1 cent games. But potential for profit will go up as more players join the game and number of higher stake games increase.

Last edited by .. (2015-02-13 05:22:13)

jere · 2015-02-13 16:43:57

Well, when you think about it, right now the cost of training a bot pales in comparison to what you might make with a bot that wins 51% of it's games. I'm not really surprised that someone is attempting it. I considered doing it myself early on. But the more I looked into it, the more I realized what a difficult task it would be.

I agree. My intuition is that writing a good picking strategy, one that can pick up on the opponent's patterns, would be a good deal of work and the betting side is likely intractable. And you can't just get a slight edge (51%) because you have to consider the rake. I'm very happy that it's not so simple and Clock Form currently seems to be a losing proposition. Makes things more interesting for the time being.

Incidentally, doesn't "Clock Form" sound like the perfect name for a bot? Cordial Minuet's name generator is remarkably astute sometimes.

Ha. I said that exact thing in my post.

Pretty sure I've lost to that guy. #ECONOMYMAKER

Plenty of people have. Speaking of names, with the upmost respect, ECONOMY MAKER is also a very appropriate name.

AnoHito · 2015-02-13 18:22:50

Incidentally, doesn't "Clock Form" sound like the perfect name for a bot? Cordial Minuet's name generator is remarkably astute sometimes.

jere wrote:

Ha. I said that exact thing in my post.

Oh, lol, I must've missed that. Sorry, I was skimming through your post a bit when I read it the first time.

jere wrote:

Plenty of people have. Speaking of names, with the upmost respect, ECONOMY MAKER is also a very appropriate name.

Yes, yes it is... Do you think it's possible the bot is actually Jason's attempt to make sure someone is always playing the game? It would make sense, because he wouldn't want there to be times when no one was available to play right after the game launched. It would make perfect sense for him to create a bot that lost small amounts over time to help get people into the game. And it's doubtful anyone would complain about him running a bot if it was losing money. Hmm...

jere · 2015-02-13 19:00:37

In some ways it would make sense, but it also would be a pretty bad PR move. Even if the bot mostly loses, someone could lose against it once and cry foul. And then there's this:

Well, what I'm "fine" with is different than what I want to actually build into the game.
There's a certain aesthetic experience that I'm shooting for, and playing randomly isn't it. It's very important that the game, as an object, embodies the target aesthetic experience exactly.
So, for example, I'm not going to put a chatroom in it. But third-party chatrooms are fine---and even exciting---to me.
I love the idea of people developing bots for this game, but I'm not going to build a bot into the game, because this is a game for humans at it's core, and I wouldn't want the average player using a bot (or random picker) as their core experience.

.. · 2015-02-14 00:43:43

Jason has also said more directly that he wouldn't consider doing that. The quote you gave was about building more capabilities into the client, such as a 'random pick' button.

Last edited by .. (2015-02-14 00:44:34)

jasonrohrer · 2015-02-14 02:29:36

Baby, it's not my bot.

ClockForm · 2015-02-14 10:44:37

You guys called it! Glad that somebody finally noticed.

After seeing Jason's post regarding bots (from this thread), I thought it would be a fun/interesting/challenging project.

I actually haven't done any work on it in the past week or two. I just recently turned it back on for fun and to drain out the rest of my account balance.

As everybody's discovered my bot is pretty terrible. It just calculates the odds of winning and picks/bets accordingly.

jere wrote:

Why run a bot that reliably loses money?

It makes for a pretty screensaver

jere wrote:

Who created the bot?

Me

jere wrote:

Do they also play manually?

I've played a bit. Especially when I was just starting and creating the bot. I haven't played much since then. I'm much more interested in improving the bot. I have, however, watched my bot play a lot. It's mesmerizing. Except when you're agonizing over its terrible decisions.

jere wrote:

If so, do they use CF or a separate account?

I've done almost all my play on the clock form account. I did at one point create another account to play against my bot and debug some specific issues, but I only played a handful of games with it.

jere wrote:

Are they gathering data to improve their operation?

I'm definitely collecting data, albeit in a pretty haphazard way, throughout the games. I probably won't ever use it for anything.

jere wrote:

Is practically every person playing Cordial Minuet a programmer?

Probably most

shallow_depths wrote:

I definitely suspected the same, especially when they still raised when the graph showed clearly they had lost that round. Mistakes happen but it was pretty suspicious. This was a while ago, mind, I think they have tweaked their strategy a bit since then.

I've gone through many iterations and tweaks. The bot actually started out just making random moves and always calling! No matter what! It's definitely not good now, but it shouldn't be making that obvious of a mistake any more.

Cobblestone wrote:

For example, last night I played a game where I was able to make it fold at the first betting stage every single time. It folded itself down to 0.

Good strategy! Right now it won't call any raises if the calculated chances of winning after the first pick are below 50%.

AnoHito wrote:

Though personally I think they may just be programming the bot for the sake of the challenge, since realistically they would have a very difficult time making enough money with it to justify the time they'd have to spend on it.

This

jasonrohrer wrote:

Baby, it's not my bot.

Thanks for providing me the opportunity for this fun little project!

I may or may not put more work into this depending on whether I get motivated. Unless Jason has changed his stance on bots.

And yes, I'm very happy with the name 'Clock Form'. It's perfect.

Dan_Dan84 · 2015-02-14 11:00:25

Well, I'm not a programmer. It's probably why the only person I'm able to beat is a money-losing bot.

I started suspecting something was up with Clock Form pretty early on, but it's not as convincing as Jere's graphs. For me, three things: 1) very bizarre betting patterns, even folding on the first round when I bet only 4 coins. 2) As someone else already pointed out: meeting bets when the graph showed an obvious loss. 3) I don't know about anyone else, but when I finish a game, I take a moment to look at the summary screen to see the details of my win/loss. Then I look at any other browser tabs I have open (e-mail, Facebook, etc.) to see if there's anything new. And then I get up and get a drink, maybe, or use the washroom. And then I consider starting a new game, if I feel like it. But with Clock Form, the game is already open and waiting for an opponent. Every time. I suppose he could just be strangely obsessed, but...

So yeah, totally a bot. Unless someone wants to get all Werewolves/Mafia here and claim that he/she's clock form...

Anyway, I suppose bots might be inevitable, and I'm sure you programming types are interested in the challenge of building and using one, but for me, I'd rather be guaranteed that I'm playing a human.

Dan_Dan84 · 2015-02-14 11:06:58

Oh great, now I look stupid posting that while the human Clock Form was finally claiming the bot as his own.

The bot's choices for how many chips to bet was also a giveaway. I mean, I don't always put in an even amount, say, 10 or 20, but the bot always puts in a, shall we say, robotically determined amount: sometimes even just one chip.

One more thing I will say: playing against a bot is kind of annoying. It's mainly the constant folding, which means the game drags out! But when I finally coax it out and get it to bet a lot on a losing hand, well, that's quite satisfying.

jere · 2015-02-14 13:00:39

Thanks for coming in here and solving the mystery once and for all, Clock! As a fun programming challenge, I definitely get it. If you ever get totally bored of it, I'm sure we'd all be interested in seeing your source code.

Last edited by jere (2015-02-14 13:00:54)

Dan_Dan84 · 2015-02-15 04:05:57

Since many people start games in the 1-4 c range, sometimes it's hard to know at first if you're playing a human or a bot. One time I was sure I was playing Clock Form, but it turned out to be a certain well-known human player... Eventually it becomes clear (one giveaway, beyond just playing better, is the fact that humans often pause to think things through, while bots just pick), but yeah, it feels a little bit like a modern Turing test...

Playing Clock Form also reminds me of this article from The Onion: "Area Grandma Enjoys Flourishing Correspondence With Mailer-Daemon"

http://www.theonion.com/articles/area-g … wit,35100/

.. · 2015-02-15 08:58:45

Ah ha!

If I ever put my bot up, I was planning on having it avoid certain publicly announced stakes including games < ~5c, since those are likely to be new players, and they're probably better off learning how humans play. Maybe it would also hold back on joining a game until it's been open for at least a minute or two, so humans get preference.

Cobblestone · 2015-02-15 21:08:30

ClockForm wrote:

Good strategy! Right now it won't call any raises if the calculated chances of winning after the first pick are below 50%.

Thank you! That totally makes sense then based on what I was doing. I noticed after a few games that it always seemed to reliably give itself the 36 row on the first pick, I'm guessing since getting the 36 was the best odds of winning. So I just gave it the worst number on the 36 row, bet, and it would fold. every. time. I'm not sure if that's what was really going on, but it worked at least!

Illuminati · 2015-02-17 20:54:23

Oh, so there we go. Haha....

...

Didn't see this before. Bots are cool.

Last edited by Illuminati (2015-02-17 20:55:24)

CORDIAL MINUET ENSEMBLE

#1 2015-02-12 22:23:26

A bot among us?

#2 2015-02-12 22:57:53

Re: A bot among us?

#3 2015-02-12 23:45:08

Re: A bot among us?

#4 2015-02-13 03:02:36

Re: A bot among us?

#5 2015-02-13 03:35:06

Re: A bot among us?

#6 2015-02-13 05:20:20

Re: A bot among us?

#7 2015-02-13 16:43:57

Re: A bot among us?

#8 2015-02-13 18:22:50

Re: A bot among us?

#9 2015-02-13 19:00:37

Re: A bot among us?

#10 2015-02-14 00:43:43

Re: A bot among us?

#11 2015-02-14 02:29:36

Re: A bot among us?

#12 2015-02-14 10:44:37

Re: A bot among us?

#13 2015-02-14 11:00:25

Re: A bot among us?

#14 2015-02-14 11:06:58

Re: A bot among us?

#15 2015-02-14 13:00:39

Re: A bot among us?

#16 2015-02-15 04:05:57

Re: A bot among us?

#17 2015-02-15 08:58:45

Re: A bot among us?

#18 2015-02-15 21:08:30

Re: A bot among us?

#19 2015-02-17 20:54:23

Re: A bot among us?

Board footer