Garrett Adelstein Was Sure His Opponent Couldn’t Bluff. He Was Wrong.
By Duncan Smith · Online Cash Game Pro
Garrett Adelstein had already become one of the biggest winners on Live at the Bike when he played one of the most consequential hands of his career in March 2018.
The game was $100/$200 with a $200 ante, and Garrett and regular opponent Art Papazyan were roughly 700bb deep. What followed was a heads-up 4-bet pot, a massive river decision, and a bluff Garrett was convinced Papazyan simply was not capable of making.
He was wrong.
Years later, Garrett views the hand as a turning point. It exposed holes in his understanding of deep-stack poker, but also something broader: the danger of becoming too confident in your model of an opponent.
“No single hand, I think, played a bigger role in the trajectory of my career.”
Deep Stacks Change the Preflop Rules
Garrett opens to $600 from the Cutoff with Kh Ks and Papazyan 3-bets from the Small Blind to $3,000 with Ac Kd. Garrett responds by 4-betting to $9,000, and Papazyan calls.
At ordinary stack depths, players are accustomed to relatively small 4-bet sizes. Garrett argues that those conventions become less attractive when hundreds of big blinds remain behind.
Papazyan’s 5x 3-bet is reasonably sized, especially at this stack depth, while Garrett’s 3x 4-bet follows the same logic. A smaller 4-bet might be reasonable 100 or 150bb deep. With stacks approaching 700bb, Garrett wants more money going in before the flop.
The more interesting lesson involves what happens next—or, in Papazyan’s case, what almost never happens.
Garrett believed Papazyan essentially lacked a 5-betting range. This may not be critical at shallower depths e.g. 100-150bb, but again, 700bb is a whole different story.
That has major strategic consequences. Garrett can 4-bet hands such as QQ or even JJ without worrying about being pushed off his equity by a huge 5-bet bluff. The same is true at the other end of the range. A hand such as K8 suited can 4-bet, happily take down the pot when Papazyan folds, and still reach the flop with reasonable equity when called.
The absence of a credible 5-bet strategy lets Garrett expand in both directions.
“Many of my opponents, especially Art in this situation, did not have 5 bet ranges.”
A 5-bet range consisting only of AA is not really functional. If your opponent knows another raise always means the nuts, he can react accordingly. At extreme stack depths, Garrett argues that players need enough value and bluffs to prevent aggressive opponents from 4-betting without fear of another raise.
Condensed Ranges Require Precision
The Qc Ts Qh flop checks through, with Garrett explaining that Papazyan’s range actually interacts better with the board despite Garrett being the preflop aggressor.
Because Papazyan called the 4-bet out of position, he can retain hands such as AQ suited and TT at high frequency, while Garrett may have called the initial 3-bet with some of those holdings.
Garrett also needs to protect his flop checking range. AA, KK and AK cannot always bet; some strong hands and bluffs must arrive at later streets through the same line.
After Papazyan checks again on the 8h turn, Garrett bets one-third pot with his Kh Ks. Papazyan calls.
The 8d river is close to a complete brick, double-pairing the board. Papazyan checks again, and Garrett bets roughly 70% of the pot.
At this point, the narrowness of both ranges makes small details matter enormously. Garrett notes that AA and KK are no longer effectively the same hand. Holding AA blocks Papazyan from having AA; holding KK does not.
Garrett also needs bluffs.
If his flop check always means he has an overpair or AK and never contains air or a slow-played monster, Papazyan’s river decisions become far too easy. Garrett therefore needs some preflop 4-bet bluffs that can check the flop before betting the turn and river.
“If you check the flop, if that always means basically you never have air and you never have a monster hand… you are not playing good poker.”
Does Your Opponent Actually Have the Bluff?
Papazyan responds to Garrett’s river bet by check-raising to $51,000.
In theory, Garrett can identify hands that could serve as river check-raise bluffs for Papazyan. JT and T9 suited are two possibilities because they can block some of Garrett’s strongest value combinations.
But simply finding a bluff in a solver is not enough.
“If the assumptions that make that bluff good are not true against our human opponent, then of course we need to throw it out the window.”
That is the real opponent-modeling problem.
It is not enough to ask whether someone is “capable of bluffing.” You need to determine whether his range can arrive at this exact node with enough bluff candidates, and whether his earlier actions allow those hands to exist at all.
Garrett believed Papazyan was unlikely to find them.
He had played with him regularly for months and had rarely seen him make an enormous river play of this kind. Papazyan was also playing significantly bigger than usual. Putting more than 250bb into the pot with a river check-raise represented a major commitment, and not one Garrett expected his opponent to do without the goods.
The problem was the price.
Garrett was getting better than 3.4-to-1 and needed to be right only around 23% of the time. Folding therefore required tremendous confidence that Papazyan’s range was overwhelmingly value.
Garrett eventually folded, and Papazyan pulled off a legendary bluff against Los Angeles live poker’s end boss.
What Happens When Your Model Is Wrong?
Garrett’s frustration was obvious during the original broadcast. He complained throughout the decision and remarked that Papazyan must have “19 queens.”
Looking back, Garrett dislikes the attitude that his reaction revealed considerably more than he dislikes the fold itself.
The bigger problem was how certain he had been.
He had constructed a model of Papazyan in which this river bluff essentially did not exist. Papazyan was capable of making a play Garrett had effectively removed from his model, which meant the model itself needed to change.
“Being convinced Art had 19 queens wasn’t just a bad look. It also cost me a ton of EV.”
The hand ultimately became valuable because Garrett did not leave the experience there.
Papazyan and another regular Live at the Bike opponent, Dan Zack, had been getting the better of him in enough difficult spots that Garrett began questioning what was missing from his game. This hand, more than any other, convinced him that the answer required much more work.
Garrett dove into PIO and eventually spent thousands of hours studying solver outputs. He was already winning big on stream and off it, but winning was no longer sufficient evidence that his understanding was complete.
There is an additional irony. When Garrett finally ran this particular hand through a solver years later, Papazyan’s Ac Kd check-raise was not especially GTO-approved.
This just strengthens the lesson.
Garrett did not need Papazyan’s specific play to be theoretically perfect. He needed the hand to demonstrate that something he believed with enormous confidence was false. Papazyan was capable of making a play Garrett had effectively removed from his model, rendering his model obsolete and in desperate need of an update.
The response was not to dismiss Papazyan’s bluff because a later simulation disliked it. It was to reconsider the assumptions underneath his game.
Improvement does not only mean accumulating new strategies. Sometimes it means discovering that a belief underneath years of successful poker was wrong, then being willing to tear into your game until you understand why.
Garrett was already winning.
After this hand, that was no longer enough.
“But I refuse to be outworked.”
More Garrett Hands
Breakdowns of Garrett Adelstein’s Biggest Hands
Garrett’s Payback Hand vs Dylan Gang →
4 Lessons From Garrett Adelstein vs Tom Dwan →
