I welcome our great overlords from SkyNet and look forward to serving them at their leisure. 00100100

Bots can duplicate human decisions but presently I don't think they are playing that way. In this game after three people fold, there are a billion different decks left. Bots would, I think have to be programmed a lot differently than they are now to deal with that.

As to colluder detection, sites don't evaluate every hand. They only do that if something is suspicious. If colluders already play well, don't always play together, and follow the precept "play normally unless more than one play is reasonable in which case pick the one that gains more (due to another colluder being in the hand or knowledge of the colluder's folded cards)" they will rarely be caught.

That being said, I must admit that the idea that the folder show only one of his two cards might be an even better idea.

Good thing I didn't take that advice before I thought about inventing a poker game you play against the house, writing the first book about holdem, or writing a book about the underlying theory of poker. Those thinking sessions were work about $10,000 an hour. I understand that algorithmic multitablers would have trouble continuing to grind out three dollars an hour per table with these rules. But I have already made whatever money I was going to, off them.

David, I still (seriously) don't understand why you think this will "thwart bots" to any extent. Can you ELI5? The only rational explanation I've seen ITT is that human opponents will be better at adapting to shown holecards than will bots. Is that the full extent of your rationale or is there something else?

I think currently the best bots essentially learn from super-computer test sets, where they play ridiculous samples of hands and define strategies based on the outcomes of samples a human could never achieve in a life time.

I believe you would be correct in stating these strategies would be difficult to adjust on the fly to account for new in-game information.

However - I think people would actually be able to program much simpler bots that just did equity calculations based on all the tabled information and that would be enough to give them an edge over humans.

EDIT: That said - I always like new ideas. It definitely provokes some thought. And the anti-collusion aspect is interesting. And as with all ideas, the initial 'flaws' that are discovered are really just brain-storming, so you can adjust the idea and fine-tune it until you've got something that works (or doesn't!)

the technology has to stop being put into the game. it is a person to person game.
instead it is being setup just to quickly fleece the unaware. sooner or later the unaware becomes aware.

one way to make it so, is to have your camera on the front of almost every computer nowadays shows that you sitting there playing while at the table in real time and talking in the mic if needed. that would be a partial start and certainly make most players more comfortable.
soon the site that did this would get all the rec. players and those to follow.
anyone not using outside help likely would welcome this.

that could be one of many things that might work along with rule changes as david suggests

Libratus [$10M supercomputer developed by top talent] took 10 minutes for some decisions (iirc) and did most of its adjusting overnight...?

---

I agree collusion is woefully easy to practice, and not that challenging to execute effectively. I wouldn't be surprised if clusters of players have their own spoken code to give away their exact holdings to friends.

Any chance augmented reality could help online poker make sure theres human eyes looking at the game

The reason they aren't being programmed that way is that there is no reason to do so. But thats the cool thing about programming, you can change the code. So If this actually happened, you can surely bet that would start being written in, especially since it represents an opportunity to increase their edge over their human counterparts.

David as others have said, the rules you suggested will be massively helpful to bots, rather than detrimental. From a basic perspective - chess is a full information game, and bots make even the best humans look silly these days. You are proposing to take a game with much hidden information and move it closer to full information. That, from a basic perspective, has to be helpful from bots.

I think (and I mean this in a factual sense and not to be an insult) you are extremely uninformed about how poker bots are designed these days if you think they are just programmed with a set of rules that they follow. Sure, there are basic programs that are little more than a big collection of "if/then" statements, but the ones playing for serious money are using learning, and giving them full information about their opponents holdings and how they played them will accelerate their ability to decimate the opposition to light speed.

As Josem says, this idea will also do nothing to deal with collusion because sites either care about collusion and employ sufficient resources to detect and enforce it and have all this information anyways, or sites don't care and ignore valid player complaints anyways.

Small computers will eventually fit under your skin and send signals directly to earphones / ear drums / brain cells.

Glasses could read small imperfections on cards and give you notice of what cards have been dealt / burned.

Face reading software will make it possible to analyze expressions for tells.

These technologies are less than two decades away, and perhaps less than one decade away.

In the abstract this might be wrong. What humans do well is filter. Our brains avoid combinatorial problems. Still in practice all the games we play are fairly simple. To be clear, I mean "more total information" and not "fuller information". The latter clearly favors bots.

the additional information proposed here clearly favors bots.

Poker rooms will require a full body scan before entering!!!

Agreed. My point was just to be careful about what one means by "more information". What is relevant here is "fuller information", and the proposed rule change seems to give that.

This is typically not how bots work, they play a style as close to GTO as they can. Even if you understand and mimic their strategy they will be better at it and you won't understand poker well enough to know what tiny holes in their strategy they have for you to exploit. Even if you can see every hand you will have no chance vs future bots. How many cards are shown is irrelevant, really.

I mean, it's pretty clear that David started this thread thinking of bots programmed to exploit population, whithout realising that they are completely irrelevant (since they can be exploited to death) to the current gto bots discussion. The fact that both of those types of programs are called "bots" can be misleading, since they are completely different things.

Still waiting for an answer to this. Surely David's thinking of a more rational nexus than some marginal edge humans may have in adapting to shown hole cards.

Suggested name for this game: Face up Fold 'em.

A further variation could be that each of the first 3 players to fold face up has to choose one of their face-up folded cards as a flop card. The dealer then constructs the flop out of these cards that players have folded. If less than 3 x players fold pre-flop then the dealer can finish off the flop from the remaining cards in the deck after the BB has acted, dealing all 3 cards the same as now in the event that every player was in the hand. This would add further complexity - potentially in a 6 max game the first 3 players fold and the flop comes 345 with A67 also exposed and this is before the opening round of betting has been completed...now the player on the button with AAis free to open the betting but has a lot more to consider than if this was a standard NL Holdem game...the sheer number of alternative scenarios / game tree size would surely keep the bots at bay at least for a few years? Other cool stuff...you pick up 23in the BB and you might just have a monster depending upon the flop that gets constructed from folded cards...all sorts of deeper strategy considerations around whether to create a connected / disconnected flop if you are the second or third player to fold....is it possible to benefit a weaker player at the table somehow in the hope that they win the pot and you can then target their chips in a later hand?

Further variation suggestion. Players get to play 2 x hands (or more) at once. So they can raise from seat 5 and then 3-bet their own initial raise with their other hand at seat 4. Again, this would create all sorts of additional complexity to make for a more interesting game that can maybe withstand the march of AI just that little bit longer.

The idea that game complexity can be significantly increased so as to at least stall bots mastery of the game is not absurd and makes total sense.

Of course it makes sense...a game structure where fewer cards remain unrevealed=less potential for collusion to occur in the first place. This has nothing whatsoever to do with what a site's security team might do to investigate possible collusion. It is a completely separate point.

Last I checked, it wasn't possible to be openly on your phone during a live poker game telling 4 other people at the table your hand. Team viewer and other software where you share screens with your friend...they use that in Live poker though don't they? oh wait..no..that is actually online. WhatsApp and Facebook messenger might be used at a live table but at least you can make it difficult for anybody attempting this by making a point of looking at their screen every so often and I have even seen people using jammers to (illegally) block phone signals at live tables...not possible online.

OP convinced me that online poker will be dead to bots in 20 years, and that it's already in danger to colluders.

I don't think he gets how this actually aids bots, and bots can also figure-out GTO w/ collusion taking potential of discovery into consideration. The British kept Enigma being cracked a secret for what? Like 50 years?

Also, your hole cards are already in the HH.

Probably because it wouldn't and David's idea makes lots of sense.

If today you have a rule about what hands you open if it's folded to you in UTG+1, you'd need 1326 rules depending on which two cards were folded before you acted.

If today you have a rule about what hands you open if it's folded to you in UTG+2, you'd need 1 million 624 thousand 350 rules depending on which four cards were folded before you acted.

etc.

So the current hypothetical bot has 19,600 rules depending on which 3 cards are on the flop and 921,200 rules for the turn?

edit:
I'm pretty sure that starting with a bot that plays a decent game of poker using equities and ranges would take approximately 1 minute to add dead cards to the calculations and zero additional processing time in use. Updating ranges would also be easy, and exposed folded cards would help define these faster and narrow remaining players' ranges.

If the hypothetical bot is "1. I haz AA, raise. 2. I haz KK, raise. 3. I haz QQ, raise. etc." No counter is necessary.

The complexity added by exposing dead cards makes pre-calculated game trees impractical. Any game decision path followed by a human can written and followed by a bot, the difficulty is distinguishing relevant information for consideration.

20? More like 2.