Super System vs the solver: six pieces of Doyle Brunson’s 1979 advice, checked

Super System came out in 1979 and taught a generation how to play no-limit hold’em. Doyle Brunson wrote the no-limit chapter himself. Almost fifty years later I can do something he could not: take each piece of advice and check it against a solved game tree.

Six claims from the book, each one against real solver output, with the spot named every time. I paraphrase the book rather than quote it, and I don’t give page numbers I can’t verify.

Everything preflop comes from one Poker Academy solution: NL200 on GGPoker, 100bb deep, 6-max, no ante, 2.5bb opens. EVs are in big blinds and measured against folding, so a fold is worth 0. The flop numbers come from a saved 25-flop report on lojack against a big blind call: 5.5bb pot, 97.5bb behind, in the fast-fold solution of the same format. Each flop is weighted by how often it actually comes.

1. Preflop: twelve combos against 166

The book’s preflop advice is famously tight on re-raises. Aces and kings go back in; queens and jacks come along without raising. Nothing in the chapter is built around re-raising as a bluff.

Aces and kings are 12 combos. Out of the 1,326 starting hands you can be dealt, that is 0.9%.

Here is the same decision in the solution. The cutoff opens to 2.5bb and the button is next. It 3-bets to 7.5bb with 12.5% of hands, 166 combos, calls 3.0% and folds the rest.

Forty-five of the 169 hand classes 3-bet here at least 1% of the time. Only a third of those 166 combos are premiums: 55 combos come from TT+, AQ+. The other 111 combos, two thirds of the 3-betting range, come from hands nobody in 1979 would have re-raised. A3s 3-bets 99.1% of the time, T9s 86.7%, 65s 84.7%.

A lot of places will tell you that modern strategy treats AQ+ and TT+ as the standard value 3-bets. That sentence leaves out the thing that decides it, which is which seat you are sitting in.

2. The same hands, one seat over

Move the same hands to the big blind against a lojack open and the answer flips. The big blind 3-bets to 11bb with only 4.5% of hands, 60 combos, and calls 16.4%.

In that seat TT and JJ are pure calls, 100% of the time. QQ 3-bets 30.9%. AQo 3-bets 15.2% and calls the rest. “TT+ is a standard 3-bet” is untrue against the tightest opening range at the table, from the seat that has to play every flop out of position.

A third version, the hijack against that same lojack open: 3-bet 7.4%, call 1.5%, fold 91.1%. AQo there 3-bets 55.3% and folds 43.6%. One hand, three seats, three answers, and a hand-by-hand rule can only carry one of them.

3. AK suited, AK offsuit and the two flushes

The book argues that AK offsuit is the better hand because it can make a flush in two different suits. It’s a nice thought and it’s wrong, and you can settle it with a deck of cards instead of a solver.

With AK suited you need three more cards of your suit among the five board cards:

With AK offsuit you hold one card of each of two suits, so you need four of a suit on the board. Two suits to try it with, and the two cases cannot both happen in five cards:

Twice as many suits to try with, and three and a half times fewer flushes. They are also not the same flush. When AKs gets there it holds the ace of its suit, so it has the nuts. When AKo gets there the board shows four to a suit, so everyone holding one card of it has a flush too, and the king only wins if nobody has the ace.

The solver agrees, and it agrees in EV. From the lojack both hands open 100% of the time, and they are worth different money:

0.60bb a hand is a real gap, and it is smaller than “far more profitable” makes it sound once you count combos. AKs is 4 combos, AKo is 12. Across one full deal of all 1,326 starting hands, the four AKs combos are worth 4.24bb between them and the twelve AKo combos 5.58bb. The suited one is the better hand. The offsuit one still pays more of your rent, because you are dealt it three times as often.

4. The 90% continuation bet

The book’s flop rule is to bet almost every time you raised before it. Modern advice usually calls that a leak. Both are too blunt, and the report shows why.

This is the lojack opening, the big blind calling, the big blind checking the flop, and the lojack deciding. Twenty-five flops, each weighted by how often it comes:

Weighted over all 25 flops the lojack bets 63.5% of the time. Brunson’s 90% is not far off on the boards his range owns: 98.8% on K♣︎K♦︎4♥︎, 98.4% on K♣︎J♦︎8♥︎, 95.6% on Q♣︎Q♦︎J♥︎. It falls apart on the boards the caller likes: 44.8% on 8♣︎5♣︎3♦︎, 48.0% on 7♣︎4♦︎2♥︎, 49.5% on 9♣︎7♦︎5♥︎.

The sizes changed more than the frequency did. Of that 63.5%, 21.9 points are a quarter-pot bet and 35.6 points are about half pot. Only 6.0 points are a pot-sized bet or an overbet. In 1979 nobody was betting a quarter of the pot with a fifth of their range. There was no reason then to treat the flop bet as a size you pick.

This heavy betting is the early-position raiser, which often gets said backwards. A lojack opener against the big blind is in position, and in position it bets most flops. The range that checks nearly everything is the one out of position, so when the lojack opens and the button calls instead, the lojack ends up checking a lot. What drives it is position, whatever the seat is called. That last sentence is my reasoning; the report only covers lojack against the big blind.

5. Leading into the raiser

Super System tells you to bet into the preflop raiser when you flop something big. For decades the standard answer was that you should never do it: check to the raiser, always. The solver says both sides overshot.

On average the big blind checks 85.7% of flops and bets 14.3%. That looks like the old rule. Then look at 9♣︎7♦︎5♥︎: it checks 20.8% and leads 79.2%, most of it for a quarter pot. On 8♣︎5♣︎3♦︎ it leads 42.2%, on 7♣︎4♦︎2♥︎ 47.7%, on 6♣︎6♦︎4♣︎ 55.1%. Those are the flops a big blind calling range hits and a lojack opening range misses.

So leading is real, it is rare, and the board decides which. Brunson’s instinct was right and the rule he wrote around it was too narrow. When the solver leads, it leads with a whole range on the boards that suit it, and big hands are only part of that range.

6. One hand, more than one right answer

This is the part of the book that aged best. Brunson makes the point that the same hand can be played several different ways and none of them is wrong. In 1979 that was an instinct about disguise. It turned out to be a description of equilibrium.

TT on the button against a cutoff open 3-bets 68.0% of the time and calls 32.0%. Both actions are worth 0.96bb. That is indifference: the two branches pay the same, and the solver needs the split so the cutoff cannot exploit either one.

Forty years before anyone could compute that, he wrote down the observation, which is the most impressive thing in the chapter.

What does not survive at all

Super System also tells you that when you are running good you should keep playing every pot until you lose one, to stay on the rush. There is nothing to check here: no solved tree contains the pot before it. Each hand is dealt from a full deck and the last one has no say in it.

The honest version of the idea is about the player, not the deck. If winning makes you sharper, playing more is fine. If it makes you looser, it costs money. That is a note about tilt, and it belongs in a different chapter.

The other line that gets quoted is the book’s claim that no computer will ever play world-class poker. Libratus took 120,000 hands off four top heads-up professionals in 2017, and Pluribus beat professionals six-handed in 2019. Libratus played heads-up only, which is the two-player zero-sum version of poker, the one where a single strategy can guarantee it will not lose in the long run.

How I’d read it today

• Keep the instincts. Aggression, position, disguise, the same hand played more than one way. All of that is in the solver’s output.
• Drop the fixed numbers. 90% of flops, only aces and kings, always lead with a big hand. Every one of those is a frequency that depends on the board and the seat.
• Check the spot before the hand. TT is a 68% 3-bet on the button and a 100% call in the big blind, in the same solution, on the same day.
• Read it as history. It is a book about how to think at a table full of people who had never read one. That part has not dated at all.

Cheers! 🙂