poker.yoga

Corrections

Answers we published, then changed. Each entry says what it used to claim, what it claims now, and what changed our mind.

A verification pipeline that never overturns anything is not a verification pipeline — it is a formatting step. This is the list of times ours did its job, kept in public because a claim to check our work is worth less than the receipts.

Changed:

Two postflop answers overturned by a realistic defence range

What it said

One bet-sizing lesson taught the big size on a dry board; one c-bet lesson taught betting. Both had been solved before shipping — against a big blind that defended only about 22% of hands versus a button open.

What it says now

Re-solved against a defence range of about 58%, which is what the equilibrium actually defends, the solver preferred the small size in the first spot (0.61) and a check in the second (0.64). Both lessons were rewritten.

Why it changed

The solver was never wrong; the opponent model we fed it was. A defence range that tight quietly removes the weak Kx, Qx and junk that pay off value bets, which makes big sizes and c-bets look better than they are. The same review examined five contested spots in total and refuted three of the objections — those three lessons stayed exactly as they were.

Changed:

Three short-stack hands flipped from fold to push

What it said

On a four-left bubble with a 50/30/20 payout, 66, 77 and A8s were taught as folds from the small blind.

What it says now

All three are pushes. The lesson and the ICM regression tests now pin the corrected answer.

Why it changed

The ICM engine had been modelling the big blind as a chip-EV caller — the range someone would call with in a cash game. On a bubble the caller is far tighter, because busting costs them real money, and against that tighter range the shove is clearly profitable. The engine had verified the folds with full confidence; it was answering the wrong question.

Changed:

One pot-odds convention, applied everywhere

What it said

“Pot” meant the pot before the bet in some places and after it in others. Call spots compared two-card equity from the rule of 4 against a one-street price without saying so, and two blog posts reached opposite verdicts on the same flush draw.

What it says now

“Pot” always means the pot before the opponent's bet. The rule of 4 is only compared against the price when the call closes the action for both cards; otherwise the comparison uses the rule of 2 for the next card, and implied odds are named as what covers the rest.

Why it changed

Two conventions, both defensible, had been used side by side, so adjacent lessons contradicted each other and a student learned that the arithmetic was arbitrary. Picking one and enforcing it was worth more than either. The fix went through the lessons, the calculators, the blog posts, the glossary, all five translated overlays and the generated cheat sheet, and a build-time check now holds every affected call spot to it.

Changed:

A gutshot lesson that taught the wrong hand

What it said

The lesson on folding a gutshot to a pot-size bet used J♠T♠ on Q♦8♥3♣.

What it says now

It uses T♥9♥ on K♠Q♦6♣, a pure gutshot the solver folds 99% of the time.

Why it changed

The original hand is one the solver actually defends: on that board it has backdoor equity and pair outs well beyond the four-out gutshot the lesson was describing. The teaching point was sound, but the example contradicted it, so the example had to go.

Found one we've missed?

If a number here disagrees with the repository, or a lesson disagrees with itself, tell us and it goes on this page. support@poker.yoga

How we verify answers