Could human beat GTO bots in Fast Fold poker?
Fast Fold poker obviously plays different in long run strategy. You can select the hands you want to play and adapt this selection to your style.
Let's say a professional player wants to play 10 000 hands session against 200 GTO bots at Fast Fold tables.
Is it possible for the human player to end up this session with a reasonable profit ?
My instinct tells me that a crazy player could possibly do it by finding a "golden path", but that's just my intuition.
Fast fold poker does not have a different equilibrium than regular poker, so long as all the other constraints are the same (namely rake/blind structure/etc.)
It is possible for a human player to end positive in a 10,000 hand session against a perfect bot, but it would be due to variance.
I thought there must be some difference from the human player's perspective. I was imagining human vs bots match as making a sculpture - you only need to get rid of the unnecessary part to get a perfect shape in the end (big profit).
Does a GTO bot focus only on the current hand or perhaps it considers the current session result also ? I mean, will bot consider the size of the current pot in comparison to the current session result ? In Fast Fold, these pots can accumulate into big ones if human plays well and narrows the hand selection to playing mostly against big stacks.
For example - if there's a big pot between big stacks in hand number 9 974 that can change the session human result from loss to profit, will the bot know this and adapt its strategy to a more cautious one and avoid making this pot even bigger (check or fold) ?
Good luck prising out actual Bot internal strategy from these boffins.
This is the chap you need to find:
I thought there must be some difference from the human player's perspective. I was imagining human vs bots match as making a sculpture - you only need to get rid of the unnecessary part to get a perfect shape in the end (big profit).Does a GTO bot focus only on the current hand or perhaps it considers the current session result also ? I mean, will bot consider the size of the c
No.
...Does a GTO bot focus only on the current hand or perhaps it considers the current session result also ?...
A GTO strategy doesn't care about prior results or ending your session on an upswing. These are emotional considerations. They matter to lizard brain but they don't matter in theory. A GTO strategy simply maximizes the EV of every hand in every spot.
Fast fold poker does not have a different equilibrium than regular poker, so long as all the other constraints are the same (namely rake/blind structure/etc.)...
At the risk of feeding the trolls, there IS actually a difference between regular poker and fast-fold poker: Time EV. If you're using an hourly utility function instead of a per hand utility function, then you change the equilibrium. A winning player would be incentivized to fold break-even hands.
Imagine you have a winrate of 3bb/100 hands. You've got 86s on the CO. Playing this hand is worth 0bb and costs you time. Folding does not cost time. You might as well fold, save time, and take your +3bb/100 edge to the next hand. Now obviously at some point opponent should fight back against your nitty opens and call tighter. But you would, in theory, reach new equilibrium.
Intuitively we could imagine it as folding being slightly +EV/hour, which has to do with how much time you're saving and your overall winrate. So you need greater than 0EV in bb/100 in order for playing some hand to be better than folding.
The implications of this are really weird though. Like for you to be winning, some other players should be losing. And if the incentive is for winning players to play tighter, then presumably the opposite is true for losing players: they have negative time EV so they should play wider, and defend some slightly -EV stuff.
(side note - I am 95% sure Brokenstars already knows all this)
Imagine you have a winrate of 3bb/100 hands. You've got 86s on the CO. Playing this hand is worth 0bb and costs you time. Folding does not cost time. You might as well fold, save time, and take your +3bb/100 edge to the next hand.
Isn't the blind cost relevant though?
Or are you saying it doesn't reach the levels we make replanting that time/energy into higher EV opens?
Sorry I didn't explain that well.
There are two ways to increase your hourly in cash games:
1) Increase your edge
2) Increase your hourly volume
- In zoom games you can fold breakeven hands (which add nothing to your bb/100 winrate), and therefore play more hands in total (increasing your hourly).
- In regular games you do not usually have this luxury. In most cases the hand will take as long to play out regardless of whether or not you're playing breakeven hands like 86s above.
Isn't the blind cost relevant though?
Or are you saying it doesn't reach the levels we make replanting that time/energy into higher EV opens?
It's irrelevant in cash games. From the perspective of holding 86s, all that matters is what decision makes the most money - folding or raising. In zoom games folding is slightly +EV (using an hourly utility function), so you can edge-pass more.
M Ratio matters in MTTs because survival is baked into the utility function. But survival is irrelevant in cash games because you can just rebuy or top up.
I still think that accumulation of some stacks into very big ones in Fast Fold poker may give the opportunity for the human to jump from one huge pot to another (in the ending phase of the match) and have a chance to win the whole match with very good profit. Luck would be necessary here to make it happen though.
HUGE POT #1 (won) >>lucky jump>> HUGE POT #2 (won) >>lucky jump>> HUGE POT #3 (won)
If you get lucky, you can win a lot of money. Amazing.
My point was that those huge pot combos can let you win much more $ than in normal cash game match. 200 players make a difference than one 9 players table.
Theoretically you could make massive winnings in just 10 hands, in the ending phase.
You could also lose a lot.
Yes, but during a 10 000 hands match you'll sometimes experience downswings and sometimes upswings. What if the downswing parts mostly occur in the 1st half of the match and the upswing parts occur mostly in the 2nd half where the chip accumulation among players is much bigger ?
Yes, but during a 10 000 hands match you'll sometimes experience downswings and sometimes upswings. What if the downswing parts mostly occur in the 1st half of the match and the upswing parts occur mostly in the 2nd half where the chip accumulation among players is much bigger ?
How is that not luck?
I guess ITryDeuces' point is that zoom features more big stacks. More big stacks → more variance → better chance of winning a 10k hand match as the underdog.
He's right of course. And by the same logic, you're more likely to win a 10k hand match against GTO bots if you're playing 1000bb deep. More variance + same edge = higher chance of luckboxing victory.
But it's not a particularly interesting conclusion.
Yes, someone can go on a heater and win against a superior opponent. This is news?
I just wanted to rectify the term: "unbeatable GTO bot".
One of the great things about poker is that anyone can win any hand.
ITryDeuces did you think the term "unbeatable GTO bot" meant you can literally never win against it?
ITryDeuces did you think the term "unbeatable GTO bot" meant you can literally never win against it?
Yes, that's how I understand this term when it comes to human - computer interaction. Adding infinity to explain "unbeatable" is strange to me, because the infinity itself is "unbeatable". Adding infinity decreases the value of the exceptions.
If someone said he developed an unbeatable video game, it would mean to me that nobody has ever finished it and never will.
Let's say there's 0,5% chance that a human can "luckbox" a GTO bot in Fast Fold poker match. If there was a one-time human vs GTO bot match organized and the human player would win, we could then say that the bot is definitely "beatable". And as you know, 0,5% could be enough to win.
As a conclusion, I would say that GTO bot is "known as an unbeatable" at the moment, but things could change in the future.
Here's how you easily beat a GTO bot:
Shove 72o, bot calls with AK -> Hit a 7 or a 2
It's actually a matter of a precise number of hands you'd have to play against a bot to state if the bot is beatable. No one lives forever, so we can't play infinite number of hands. And since it's impossible to determine this number precisely, it's kind of not fair to glorify the bot as UNBEATABLE.
Bring Ivey or Dwan to a match vs 199 GTO bots 10K hands in Fast Fold and it could happen.
