Access rec.gambling.poker thru dejanews. Find a message posted by abdul --
any message. Click on the thing that says information about author. Get a
list of his rgp postings and look thru them until you find something of
(This version is dated April 27, 1999. The most recent version can
always be found
and that version is nicely formatted too. Permission is granted
to reproduce for netnews archives, netnews articles, and personal use.)
Due to the effect of community cards, hold'em is a game of
"domination," a term coined by Roy Hashimoto. A hand is dominated if
it has 3 or fewer outs against another, like AJ against AQ. Second
best offsuit hands are what make you money in hold'em - when *other*
players play them. With the flip of a card, pairs and suited hands
can transform from dominated to dominating.
Big and small pairs, suited hands, and offsuit hands play differently.
Small pairs, suited cards, and zero and one gap hands (examples: 22-66
A6s, and JTs and QTs respectively) thrive on "implied odds", a term
coined by David Sklansky, meaning they will frequently be folding
after the flop unless they flop big, and so they normally want to see
the flop cheaply.
Offsuit hands have "reverse implied odds", since they cannot usually
bet and raise with confidence towards the end of the hand. Normally,
an offsuit hand likely to be best should make it expensive to see the
flop, in order to harm the hands that would have good implied odds to
see the flop cheaply. A strong offsuit hand is still strong when
facing several opponents, between its chance of making a AKQJT
straight, two pair (usually using a low pair on the board) or a top
pair that holds up.
Big pairs have reverse implied odds as well, but they are much more
robust, since they can win unimproved, or by making two pair with a
low pair on the board, or by making a set or full house.
The flop is the nexus of the hand; limiting raises preflop goes far
towards disguising your hand.
The next sections detail strategy for opening, playing against
limpers, playing against raisers, and defending the blinds. Six
representative hands, namely QQ, 55, ATs, 76s, AQ, and JT, will
illustrate preflop strategy for each of these situations. However, if
you think you know better for your particular situation, you probably
do, as "it depends." In any case, this simple desert nomad does not
claim to be always correct, only always thought provoking.
When no one has yet entered the pot, the following chart shows you
conservative minimum opening hands for various positions to the right
of the button. The farther off the button, the tighter you have to
play, as you need a reasonable chance of having the best hand.
OPENING HAND RANKINGS AND MINIMUM OPENERS
Button Pairs Suited Hands Offsuit Hands
TT AQs AK
99 AJs AQ
6/7 88 ATs KQs
5 77 A9s KJs QJs AJ
4 66 A8s KTs QTs JTs AT KQ
3 55 A7s K9s T9s J9s KJ QJ
2 44 A3s K7s Q9s 98s T8s A9
1 33 A2s K6s Q8s 87s 97s J8s A7 KT QT JT
22 K4s Q6s 76s 86s T7s A6 K9 T9
0 K2s Q4s 65s 75s A3 K7 Q9 98 J9
Q2s 54s 64s 85s Txs A2 K6 Q8 87 T8 J8
43s 42s 63s 7xs K5 Q7 76 T7 J7
32s 62s K4 Q5 65 86 96 J6
K3 Q4 54 75 85 J5
K2 Q2 43 53 63 J2
Note: The hands are sorted horizontally to make them easy to locate:
aces, kings, queens, zero gaps, one gaps, two gaps, and "other."
When a hand is "missing" in the table, it belongs with the lower
hand; for example, Q3s belongs with Q2s since Q4s is just above Q2s.
Jxs fits in the same spot as Txs. Jxs refers to baby suited jacks
lower than the zero gappers, one gappers, and two gappers, in other
words J2s-J7s. A handy way to refer to the rank of nonpairs is by
the connectors 32-KQ and then big aces AJ-AK. So for example
"77/QJs/AJ or better" is a short way of saying 77-AA, QJs, KJs-KQs,
A9s-AKs, and AJ-AK. On the button the minimum openers are 22/65s/98
caliber hands. Weaker hands are listed for when you wish to steal
with more hands against tighter opponents in the blinds.
Adjustments: You can open one level looser when your opponents are
properly tight, at least two levels looser when they are too tight,
and two levels looser on pairs and suited hands when your opponents
are too loose. You can also open an additional two levels looser on
pairs and suited hands when your opponents are passive. A normal rake
will move you back up one level, and a harsh rake will move you up two
levels, more in late position.
In early position you have to play fairly tightly, even in loose
games, since you don't know how many raises there will be. Consider
how likely you are to be raised by weaker (or stronger) hands if you
limp, how likely you are to be called by weaker (or stronger) hands if
you raise, and how likely you are to steal the blinds if you raise.
When opening in tight games in any position or loose games in late
position, your attention should be on getting heads up with a blind or
outright steal the blinds. Most hands are worth less than the blinds
and so for most hands stealing the blinds is a coup; hence, raising is
correct for most hands. AA is worth about four times the blinds, so
stealing the blinds with it and your other very strong hands is a
major disaster. Without other concerns, in a tight game you should
raise with marginal hands, and limp (and usually reraise if raised)
with your strongest hands. This advice contradicts Sklansky
and Malmuth. Balance your hands that you could have in various
preflop scenarios, mixing strong with weak and weak with strong, so
that you do not give too much information away by your actions, yet
strive to still play most hands appropriately.
Here is one way to balance your opening strategy for a tight game
where you are fairly likely to steal the blinds if you open-raise:
Tight Game Opening Strategy
Raise and call 2 QQ JJ TT KQs KTs JTs
Raise and call 1 77 QJs KJs AQ AJ
Limp-reraise / raise & call 2 AKs AQs AK
Limp & call 1 66 55 A9s A8s A7s KQ
Limp-reraise AA KK 99 88 AJs ATs
Note: "Call 1" means call one raise back, fold for two, and similarly
for "Call 2." When two ways to play are listed, separated by a slash
(/), do them each 50% of the time or adjust depending on the texture
of the game. In general, you should mix up your play a bit on all
hands. Because players tend to put you on AA or KK when you limp-
reraise, often refrain from doing so heads-up. Some plays are
"sacrifice plays" for the sake of balance, such as limping with KQ in
early position when in isolation raising would be better and folding
would be best. Other plays are profitable only in context, such as
being able to play 55 early under the cover of the limp-reraising
hands. Beware reraising when the raise comes from the blinds, as few
players will raise from the blinds without holding QQ-AA or AKs.
Example: You are in early position, 6 off the button, in a game that's
so tight that an early raise often wins the blinds. The pot is not
yet opened. How do you play your hand?
QQ Raise to add support, but limp-reraise is more immediately profitable.
55 Limp if (and only if) you limp-reraise often with other hands.
ATs Limp-reraise to profit from opponents folding AQ & AJ to limp-reraises.
76s Fold. A raise would be better than a call, though, to steal the blinds.
AQ Raise, for win share and to get heads up.
JT Fold. Dominated. Even KQ is played up front only for balance.
In games where a raise generally gets 1 or 2 callers, but rarely
steals the blinds, open-raising with any playable hand is very
reasonable and helps avoid leaking information.
In a loose game, where you will gets lots of callers if you limp and
almost as many callers if you raise, proper play is more
straightforward and includes playing more suited aces. Here is one
way to balance the hands for loose-aggressive games:
Loose-Aggressive Game Opening Strategy
Limp-call 2 / raise & reraise 99 88
Limp-reraise / raise & reraise AK AKs
Raise & reraise AA KK
Raise & call 2 A5s A4s A3s KQs AQ
Raise & call 1 AJ KQ
Limp & call 1 QJs JTs QTs 66
Limp & call 2 ATs A9s A8s A7s A6s KJs KTs 77
Limp-reraise QQ JJ TT AQs AJs
Note: For loose-passive games and extremely loose games, replace all
limp-reraises with "raise & reraise."
Example: You are in early position, 6 off the button, in a game that's
so loose that you always see a flop, usually 5-8 way for 1 bet or 4-6
way for 2-4 bets. The pot is not yet opened. How do you play your
QQ Limp-reraise to punish them, except raise in very loose or passive games.
55 Borderline call/fold. Play if you can see flop cheaply.
ATs Limp and call all raises, fearing that raises indicate AK, AQ, or AJ.
76s Fold, but it's close for very loose-passive games.
AQ Raise to destroy the implied odds of the fish and narrow the field.
JT Fold. Dominated. KJ and QJ suffice in very passive games with no rake.
In middle position, you will be raising with more weak hands to steal
the blinds, so you can raise with most of your strong hands too,
especially since limping is unlikely to induce a raise.
Example: You're in (late) middle position, 3 off the button. How do
you play your hand?
QQ Raise. No one is likely to raise for you. Provide cover to steals.
55 Raise if you can get heads up, call if you can get 4 callers, else fold.
ATs Limp-reraise if you are limp-reraising with AA and KK, else raise.
76s Fold. Likely dominated downstream. Cannot count on enough callers.
AQ Raise, for the same reason as early position.
JT Fold. You'd need a minimum offsuit of close to AT or KJ to open here.
On the button, you should be open-raising with a lot of hands if your
opponents defend the blinds properly, and if they are too tight you
can raise with any two cards at least until they start adapting.
Example: You're on the button. How do you play your hand?
QQ Raise. It is too conspicuous to limp here.
55 Raise. Your pair is quite strong here, if you get heads up.
76s Borderline raise/fold. Laying odds. Fold versus loose small blind.
AQ Raise. This is a monster. A3 would suffice.
JT Raise. Finally, on the button or one off, it is likely best, barely.
A rake seriously reduces the number of hands with which you can steal,
as you will be paying a lot for a crapshoot against the big blind.
With a Draconian rake, like where the big blind gets dropped once the
flop comes, you would need about JJ or better to open on the button!
Even with a modest rake, JT and 76s should be folded.
You should raise an opened pot when you will win the pot more than
your fair share of the time or your hand would play better without
additional players in the pot. Consider whether calling would lure
dominated hands to call after you (or additional hands period to give
you odds for your draw), or whether raising would drive out dominating
hands after you or allow you to get heads up (or almost so) versus a
hand you dominate.
Most people think that you should play looser after limpers compared
to opening. If a tight player limps, you have to be careful. Even if
the limper raises with his best hands, versus his weak limp you have
to play about as tight as if you were opening in his position, as you
have no chance to steal the blinds, though you should still raise if
you suspect you might dominate his hand. On the other hand, if the
limper would limp with his best hands, then you must play much
tighter. After several tight players limp, you can play hands that do
well multiway (any pair, any suited ace, big suited kings and queens,
and medium to big suited zero and one gappers), but the only offsuit
hands you can play are AQ and AK, partially for fear of domination,
partially for fear of the big cards being "dead." (AJ and KQ are okay
after just one tight limper.)
Example: You are facing one tight limper and you are on the button.
How do you play your hand?
QQ Raise. No need to worry about stealing blinds. Calling is a mistake.
55 Borderline fold. Unlikely to get heads-up and cannot get 4 callers.
ATs Call. Proceed with caution if you flop an ace for fear of limping AJ.
76s Fold. Similar to 55 case. Borderline fold/call versus 3 tight limpers.
AQ Raise. Same with AJ and KQ. Your hand is likely best. Get heads up.
JT Fold. Dominated. Fold QJ/KJ too. Calling here is a huge mistake.
With loose players coming in with hopeless hands like T7 and J6, then
it's true that you can play looser after limpers, with "trashy" suited
hands like T8s and K4s, and any pocket pair. You should raise
liberally to punish them, since weak offsuit hands really get hurt by
preflop raises, as they have only a tiny chance of winning the pot.
After many limpers, even Q6s and 65s can play best with a raise on the
button; suited aces, kings, and queens and suited zero gappers win
more than their fair share of pots versus many loose limpers. Offsuit
hands likely to be best will also win more than their fair share of
pots and should raise.
Example: You are facing five loose limpers and you are on the button.
How do you play your hand?
QQ Raise. You will win the pot more than your fair share, though < 50%.
55 Call. About 8-way to flop, but it will win less than 1 in 8 times.
ATs Raise. Big suited's win more than their fair share in multiway pots.
76s Raise. Even suited zero gappers win more than their fair share here.
AQ Raise. Your hand is likely best, by far.
JT Fold. If you want to play offsuit cards, you must have the best.
It is a myth that hands like AQ are in trouble here. You are in
trouble if you don't raise, but if you raise you wreck the implied
odds of the suited garbage your opponents hold. AQ frequently wins
even in family pots by making aces up with queen kicker or an AKQJT
straight. Also, your cards have a better chance of being live if no
one raised, so you will win the pot considerably more than your fair
share of the time. Similarly, if you were likely to have the highest
hand with something like KJ or even KT, you should raise here, again
partially for win share, partially to wreck the implied odds of your
opponents. This advice contradicts Sklansky and Malmuth, as
well as others. Their argument is that the fish will call correctly
with gutshots and pairs on the flop if you raise preflop, but the problem
is that the fish will be calling with pairs and gutshots no matter what,
and their loose calls usually will be correct whether you raised or not.
Would you prefer they pay 3 small bets to see the turn or would you like
to let them get off cheaply for just 2 small bets to see the turn?
However, if you make a mistake by usually laying down AQ on flop that
misses even though you believed you had the best hand preflop then
perhaps you would be better off playing incorrectly preflop by not
raising. Another exception could be made if your opponents will "check
to the raiser" if and only if the flop contains an ace, king, or queen.
It is a myth that you should raise with baby pairs like 33 after six
(or fewer) limpers, even if you know the blinds will call, because
though you will flop a set more then 1 in 9 times, you will win the
pot less than 1 in 9 times. This too contradicts Sklansky and
Malmuth. A possible exception is when the raise has a decent chance of
buying you a free card on the flop, as this now improves your chance
of winning to better than 1 in 9, but it is normally rare that all 8
opponents would check to the raiser.
FACING A RAISER
The key concept when facing a tight raiser is: "run away and live to
fight another hand." Most players raise with their best hands, limp
with their worst hands, and you can exploit this by deftly
sidestepping their raises and punishing their weak limps with raises
of your own. You need a hand a couple levels higher than the raiser's
minimums to consider playing. Offsuit aces are especially vulnerable
to being dominated by a tight raiser. The implied odds of suited zero
or one gappers are trashed by raises. Medium pairs can easily be
dominated by bigger pairs, and otherwise it's usually a crapshoot
against two overcards. Versus a tight raise, you can only three-bet
profitably with AA, KK, and AK. Therefore, to avoid giving away
information, flat call with these hands preflop and go for a raise on
Example: You are facing a raise from 77/QJs/AJ or better. What do you do?
QQ Call, for fear of AA, KK, or losing to something like AK.
55 Fold. You need about 99 to call, two levels higher than his 77.
ATs Fold. Dominated. You could call with AQs, barely.
76s Fold. Implied odds are shot to hell. JTs/QJs/KQs should fold too.
AQ Fold. Against looser raises you could call. See AQs note under ATs.
JT Fold, unless you are a fish.
Versus a loose raise, such as a steal raise from one off the button
when you are on the button or small blind, you should reraise
liberally to isolate, unless you fear your hand could be beat by the
raiser but could be called by some weaker hands behind if you flat
Example: You are on the button facing a raise from one off the button
from a good player with competent opponents in the blinds.
QQ Reraise. You do not fear AA or KK here.
55 Reraise. Your hand plays much better heads up than 3-way.
ATs Borderline call/reraise. For fear of AJ, AJs is the first safe reraise.
76s Borderline reraise/fold. Your hand plays better heads up than 3-way.
AQ Reraise. Keep it heads-up to preserve chance of winning unimproved.
JT Borderline fold. Could call versus an even looser raise.
Versus a raiser plus cold callers, you have to play a bit differently
than versus just a raiser. Tight cold callers are bad news; each one
increases your calling requirements. Loose callers relax the calling
requirements for suited cards, and for pairs if you will have many
opponents for the flop.
Given how tight you have to play versus a single raise, you can
imagine how tight you have to play if there is a raise and reraise
from tight players in early position. You can still play with TT and
JJ, unless the reraiser is extremely tight. This contradicts Sklansky
and Malmuth. This is a reraise or fold situation. Make it four bets
with TT-AA, AK, AKs, and fold everything else, normally. Now if it's
a steal raise and a resteal reraise, then that's another story, and
you could wade in with 88/QJs/AQ and up, certainly, and probably a bit
weaker hands as well.
When you are in a crazy game that is constantly having capped family
pots preflop, you can call with a minimum of 22/JTs/AQ. If the game
is crazy but tighter, only getting capped once or twice per lap three
to five way, you must play very tight, playing not much more than
JJ/QJs/AK and up.
HOW TO PLAY IN THE BIG BLIND
Raising in the big blind after limpers gives away information, but a
raise often can buy you the pot by the turn if the game is not too
loose, as your opponents will often put you on AA or KK. You can raise
fairly liberally in the big blind versus loose limpers, with 88/JTs/KQ
and up, possibly a bit weaker. Versus tight limpers, you have to be
sure your hand is best.
The rankings of hands when defending the big blind versus a raise is
quite a bit different than the rankings for opening. You are getting
over 3:1 odds to flop something good, or at least a pair. Proper big
blind defense strategy varies dramatically depending on the raiser's
minimums. Against typical raises, call liberally with hands that have
straight or flush potential, as well as pairs. Get away from big
offsuit hands that are likely dominated. 65s is usually on par with
KQ here. If flopping a pair won't do you any good, because the raiser
is so tight that he is likely to have a big pair, then fold liberally,
especially offsuit hands. More specific recommendations are in the
table below. The minimum hands are listed, and you can defend with any
hands "between" the ones listed and the column headers.
Big Blind Defense vs a Raise
Raiser... Defend with minimum...
Type Minimum | AA AKs KQs QJs JTs J9s J8s Jxs AK KQ QJ JT J9 J8
===== ========= | == === === === === === === === == == == == == ==
Tight 99/AJs/AQ | 55 AJs KQs QJs T9s ... ... ... AQ .. .. .. .. ..
Legit 66/JTs/KQ | 22 A2s K2s Q2s 43s 53s 74s 9xs A2 K9 Q9 54 42 85
Steal 22/54s/76 | (all but Q3 J4 T5 94 84 73 62 32 or worse)
Notes: Versus a tight raiser heads up, do not reraise - you are
either beaten, or you'd like to check-raise on the flop. Versus
multiple loose players, you can reraise fairly liberally, e.g., with
88, ATs, K9s, QJs, AQ, KQ or better. Versus steal raises, reraise
heads up almost any time you are likely to have the best hand, as
your opponent is sure to call one more bet before the flop, but not
necessarily on the flop. Bet into a steal-raiser liberally on the
flop. Versus one or more callers in addition to the raiser, get
away from offsuit aces below about A9 and your weakest offsuit hands
like 42, but you can call with any two suited.
Example: You are in the big blind, a sane player raises in middle
position, and there is no rake. (Assume he has 66/JTs/KQ or better.)
What is your best play?
QQ Call. Go for check-raise on the flop.
55 Call. Do not necessarily give up if you do not flop a set.
ATs Call. Bet or check-raise on most flops, but check-call when ace flops.
76s Call. Check-raise the flop if you have a draw or flop a pair.
AQ Call. Consider a check-raise on the flop even if you miss.
JT Call. Proceed with caution if you flop a pair.
Keep in mind that versus a very tight raise, like from 99/AJs/AQ or
better, the situation is much different, and you should fold even AQ
in the big blind for fear of being dominated.
A rake will severely reduce the number of hands with which you can
defend heads-up. In the above scenario, JT should be mucked when
there is a rake. If the rake is harsh, like 10% with a cap, you should
defend with very few hands indeed.
In games where you are facing a preflop raise that is bigger than the
big blind (like a $4 raise to $6 against the $2 big blind in 1-4-8-8),
obviously you are not getting much odds and must play much tighter
HOW TO PLAY IN THE SMALL BLIND
Small blind openers are similar to button openers, but you should go 2
levels looser on the suited hands, and a bit looser on zero and one
gap offsuit cards as well, while actually playing a bit tighter on
weak offsuit widely gapped hands. Do not raise with all playable hands,
as you would like to call with your weakest hands and you need to provide
them some cover, and also there is no small blind to knock out.
When the pot is not raised and you only have a fraction of a bet to
call, the situation is similar to calling a raise in the big blind, as
you are getting big odds. You still need to get away from hopelessly
dominated hands like Q5 except versus many loose limpers. Getting big
odds to see the flop is no good if you are dominated.
The small blind's size relative to the preflop call amount of course
makes a big difference. There are 3 common blind sizes:
Blind Size Example
1/3 $2 blind in $6-$12 with $2 and $6 blinds
1/2 $5 blind in $10-$20 with $5 and $10 blinds
2/3 $10 blind in $15-$30 with $10 and $15 blinds
The $1 small blind with $1 and $2 blinds, $2 to go, in a 1-4-8-8 type
game, is more like a 2/3 type blind, than a 1/2 blind, due to the
implied odds of flopping something.
It also matters how many opponents you face and how tight they are.
The more opponents, the looser you can be on the suited hands. If the
limpers are tight, you still have to be extremely conservative with a
1/3 blind, especially with your offsuit hands, as shown in the table
below. Again, you can play any hand "between" the listed hand and the
Small Blind Defense Versus 1 Tight Limper
Blind Size AA AKs KQs QJs JTs J9s J8s Jxs AK KQ QJ JT J9 J8
========== == === === === === === === === == == == == == ==
1/3 22 A9s K9s ... ... ... ... ... AQ .. .. .. .. ..
1/2 22 A2s K2s Q7s 76s T8s J8s Jxs A7 K8 Q9 T9 J9 J8
2/3 22 A2s K2s Q2s 43s 53s 74s 9xs A5 K7 Q8 98 97 J7
Note: Play tighter if the big blind is likely to raise.
Example: You are in the small blind versus one tight limper. Best play?
QQ Raise. Calling would give the big blind a free shot to beat you.
55 Call. A raise will be unlikely to get rid of the big blind.
ATs Borderline raise/call. Call when you have to put in 2/3's of a bet.
76s Borderline call/fold. Fold for 2/3 bet, since 3-way is bad.
AQ Raise. You want to be heads up so you can win unimproved.
JT Call for 1/2 or 1/3 of a bet, fold for 2/3 of a bet. Be careful.
When the players are looser, you can loosen way up when you are
getting your discount in the small blind:
Small Blind Defense versus 5 Loose Limpers
Blind Size AA AKs KQs QJs JTs J9s J8s Jxs AK KQ QJ JT J9 J8
========== == === === === === === === === == == == == == ==
1/3 22 A2s K2s Q2s 43s 42s 74s Jxs A3 K7 Q8 JT .. ..
1/2 22 A2s K2s Q2s 32s 42s 52s 62s A2 K2 Q5 54 J9 ..
2/3 22 A2s K2s Q2s 32s 42s 52s 62s A2 K2 Q2 32 42 J8
Example: You are in the small blind after 5 loose limpers. Best play?
QQ Raise, for the same reason you would normally after limpers.
55 Call. See if you flop your set before investing more.
ATs Raise, for the same reason as in late position after limpers.
76s Call. Harder to win pot out of position so may not win your fair share.
AQ Raise, as you will win more than your fair share.
JT Call. You certainly cannot raise. Enough of a discount to call.
When defending the small blind versus a raise, your minimum
requirements are about midway between your minimums for calling in the
big blind versus calling a raise cold - a bit tighter for a 1/3 blind,
and a bit looser for a 2/3 blind. Additionally, a 2/3 blind can call
a raise with any suited ace. When defending versus a raise and
reraise, defending the small blind is not significantly different from
calling 3 cold.
The above is the result of a lot of hard work on my part, trying and
discarding many approaches before arriving at my goal of an accurate
preflop strategy that novices can understand (I hope.) However, I
stood on the shoulders of giants. My thinking has been especially
influenced by David Sklansky, Roy Cooke, Mike Caro, Paul Pudaite, Jim
Geary, Annie Duke, Ed Hill, J.P. Massar, "tangram", "Randall Flagg",
"Lonestar", "Ramsey", Andy Latto, Roy Hashimoto, Lee Jones, Barry
Tannenbaum, Steve Brecher, Michael Maurer, Eric Holtman, Tad Perry,
and hundreds of rec.gambling.poker posters, though this is not to say
they would agree with my recommendations. All the charts come from
Turbo Texas Hold'em 2.0 and 3.0 simulations, as interpreted by me, and
so I have to thank Bob Wilson most of all.
I don't understand the format of your chart. For example:
TT AQs AK
99 AJs AQ
6/7 88 ATs KQs
Would you be kind enough to go into further detail. Does this mean that these
are the hands that you are recommending for this particular position? I
understand what the 6/7 means. Another example:
1 33 A2s K6s Q8s 87s 97s J8s A7 KT QT JT
22 K4s Q6s 76s 86s T7s A6 K9 T9
These are the minimum hands for one seat to the left of the button? This is not
criticism as I want to make sure that I am reading it correctly. I will
probably have a few comments once I make sure that I am reading it right.
> I don't understand the format of your chart. For example:
> JJ AKs
> TT AQs AK
> 99 AJs AQ
> 6/7 88 ATs KQs
> Would you be kind enough to go into further detail. Does this mean that
> these are the hands that you are recommending for this particular
> position? I understand what the 6/7 means.
6 or 7 off the button you should open with 88-AA, ATs-AKs, KQs, AQ-AK.
As noted these are conservative requirements, based on opponents who
are slightly too loose (or playing with a rake.) Normally, IMO, you can,
open with slightly more hands (the next line or two down in the table.)
"Missing" hands belong with the next lower hand, so AA-QQ are included
with JJ. Actually, they definitely deserve to be listed individually
above JJ, but I just stuffed them in with JJ to save space.
> Another example:
> 1 33 A2s K6s Q8s 87s 97s J8s A7 KT QT JT
> 22 K4s Q6s 76s 86s T7s A6 K9 T9
> These are the minimum hands for one seat to the left of the button?
The line with 33 A2s etc are the minimums for 1 to the *right* of the
button. You can of course play everything above it in the chart as well.
Again, you can sometimes open with hands a line or two (or more) further
down in the table, depending on the circumstance.
In the article I gave an example for the "0" case, on the button,
which was supposed to make this clear.
Pity it wasn't clear. It's not supposed to be complicated.
The rankings are listed in such detail mostly for the sake of completeness,
and expressiveness. I can say "raise with 22/76s/T9 caliber or better
hands" for some particular circumstance, and people instantly know about
what I mean, and if they want they can look in the chart and figure out
exactly what hands I mean.
> HOW TO PLAY IN THE BIG BLIND
> ... More specific recommendations are in the
> table below. The minimum hands are listed, and you can defend with any
> hands "between" the ones listed and the column headers.
> Big Blind Defense vs a Raise
> IF THEN
> Raiser... Defend with minimum...
> Is Has
> Type Minimum | AA AKs KQs QJs JTs J9s J8s Jxs AK KQ QJ JT J9 J8
> ===== ========= | == === === === === === === === == == == == == ==
> Tight 99/AJs/AQ | 55 AJs KQs QJs T9s ... ... ... AQ .. .. .. .. ..
> Legit 66/JTs/KQ | 22 A2s K2s Q2s 43s 53s 74s 9xs A2 K9 Q9 54 42 85
> Steal 22/54s/76 | (all but Q3 J4 T5 94 84 73 62 32 or worse)
> Keep in mind that versus a very tight raise, like from 99/AJs/AQ or
> better, the situation is much different, and you should fold even AQ
> in the big blind for fear of being dominated.
(99/AJs/AQ or better means 99-AA, AJs-AKs, or AQ-AK.)
Oops. Note that this paragraph and the table contradict each other.
The table came from simulations I did a while ago. The paragraph
I wrote more recently using intuition, not realizing that I was
contradicting the simulation results. Assume AQ will fold on the flop
if and only if it does not flop an ace or queen, and that one bet will
go in each round (check-and-call flop, check-and-call turn, bet river);
if my mathematical scribbles are correct then AQ should call here.
If an ace or queen flops then it's much more likely that the opponent
has a pocket pair that you beat than a hand that beats you.
It's hard to do this math correctly, however. Barbara?!
It may ultimately hinge on how well your opponent plays
his AK. If he gets away from AK when a queen flops, as he should,
you could be in trouble with AQ here.
What do you think? Call or muck AQ here in the big blind versus
a raise from 99/AJs/AQ or better?
> Thank you very much for this and all the other research you have posted.
> My only concern is the fact that you base your results on tth simulations.
> (Do you test these results in real games? What has been your real-life
> experience with these starting requirements?)
Of course I use much of what I learn from Turbo sims in live games.
I have found the recommended style fairly devastating to my opponents;
by that I mean I can see that the plays have their desired effect.
Luring your opponents into playing dominated hands while avoiding
playing dominated hands yourself gets you most of the way to victory.
However, much of the recommended style comes from my logic and intuition
and some pure mathematical balancing of the strength of hands I can have
for various preflop scenarios cross various flops, not from Turbo.
> I am a statistician and economist professionally. I have also done some
> chess programming. Perhaps this leads me to my skepticism about tth,
> somewhat. I am a fairly new poker player, so my poker knowledge does not
> compare to yours. However, my experience is that tth is simply not as
> strong as most of the players in actual games (at 3-6 and 5-10 in AC).
> I read a posting of yours in which you stated that you found tth to be
> as strong as players in 10-20/15-30 games. This may be true at first,
> when you are playing the program. But the tth opponents are simply more
> predictable than actual players and this becomes apparent when you
> play with the program for awhile (I have played against it pretty
> much every day for the past five months trying to improve my game).
> I believe this has to do with the problem of introducing heuristics
> into computer programs.
Okay, so maybe *you* can predict what the computer players will do.
But the computer players cannot. They may be worse at playing tricky
hands like KJ for this reason, but basically I think the hand rankings
are likely close to correct for tight games. If we were to play a hand
heads up and you were to try to put me on a hand, your thinking would
go something like "Okay, he has raised late preflop and bet the flop of
K75. This means he could have just about anything. Okay, I call. Now he
has checked the turn of a 9. This means he is either really strong or
really weak or somewhere in between." And so on. You can't normally
get a good read on a good player heads up. You just have to play your hand
and the board, contesting the pot the right frequency of the time. Turbo
seems to do this well. I'm not asking for it to do any brilliant plays.
In fact, the more straightforwardly it plays the better, for the
purpose of hand rankings and that sort of thing. If the Turbo players
tried to exploit weaknesses, then I would be less sure of the simulation
results. In any case, to the extent that Turbo players outplay (or
don't outplay) each other, it all sort of evens out.
> In the early days of chess programming, there was a great deal of enthusiasm
I know that sims have had troubled histories in chess and backgammon.
I believe heads-up poker is a different beast. You simply have to contest
the pot any time you have a reasonable hand, and sometimes when you don't.
Making good laydowns is a dangerous way to try to win. Raising too much
will also make you exploitable. So you're often going to the showdown,
you can't get a good read on your opponent, and you can't get too fancy.
Basically, playing a good game of heads up hold'em is not nearly so
complicated as playing a good game of chess. Good heads up players
devastate weak heads up players simply because weak heads up players don't
have a clue about how often to contest the pot.
Full table tight games play very similarly to heads-up games, since you
are usually heads up by the flop.
Now loose games are a problem. Playing KJ correctly in such games is
extremely difficult for humans, much less a program. Truly expert
play in mulitway pots might surpass the complexity of expert play for
chess. Loose Turbo games also don't play much like loose real games.
So I'm worried about the loose game simulation results. However, much
of the strategy in loose games is simply understanding pot odds, and
Turbo does understand pot odds for straight and flush draws. (How well
it understands it is not clear, but it appears good enough for government
work.) When the pots get big, again you simply have to keep contesting
the pot if you have a reasonable hand or a reasonable draw.
For example, yesterday in a Mirage 20-40 game I have AA in the small
blind and raise after 4 limpers, everyone calls. The flop comes
589 two suited. I bet, big blind raises, two players cold call,
the button reraises, I four bet, the big blind calls, the two middle
players fold, and the button caps it at 5 bets. Now I have to
suspect the button has 76 or a set. But by now the pot is monstrous.
My knowledge does me no good. He *could* have a flush draw (or two
pair.) I feel I have to call him down. Indeed he has 76. The point
is, pot odds often prevent you from being very smart in limit hold'em.
But again, I have little confidence in Turbo sims of multiway pots.
Not only are this situations very complicated, but the interaction
of the different strategies will have a huge impact on the results.
> In any case, my concern is that by using simulations, any errors of this type
> might be magnified by the program. Perhaps you can give me you opinions in
> this regard.
Most people side with you. Sklansky proposes the alternative of no
fold'em simulations, adjusted for how "scared" the hands are (implied odds
and reverse implied odds), arguing that this is much better than a true
TTH type simulation. Malmuth proposes that his "expert" intuition is much
better than a TTH type simulation. I think you're all wrong. Ultimately,
once other/better simulators have been developed, researchers can
independently confirm or disprove my hypotheses. Until then, any findings
do have to be viewed skeptically. The problem is that I don't think much
of the present alternatives of no fold'em simulations or peoples'
intuitions, and so Turbo sims, math, and logic are it for me.
Have you checked out what the group at www.cs.ualberta.ca/~games/poker
has been doing? Their combination of opponent modeling and selective
search looks to be promising. Perhaps when they release the code, you
and others could use it to check your (plural) results.
Thank you very much for this and all the other research you have posted.
My only concern is the fact that you base your results on tth simulations. (Do
you test these results in real games? What has been your real-life experience
with these starting requirements?)
I am a statistician and economist professionally. I have also done some chess
programming. Perhaps this leads me to my skepticism about tth, somewhat. I am
a fairly new poker player, so my poker knowledge does not compare to yours.
However, my experience is that tth is simply not as strong as most of the
players in actual games (at 3-6 and 5-10 in AC). I read a posting of yours in
which you stated that you found tth to be as strong as players in 10-20/15-30
games. This may be true at first, when you are playing the program. But the
tth opponents are simply more predictable than actual players and this becomes
apparent when you play with the program for awhile (I have played against it
pretty much every day for the past five months trying to improve my game). I
believe this has to do with the problem of introducing heuristics into computer
In the early days of chess programming, there was a great deal of enthusiasm
about mimicking human-type play with programs that had "human-like" judgement
(this is part of the field of artificial intelligence). However, it soon became
apparent that computers were very poor judges of chess positions because they
are unable to distinguish exceptions to general rules. Chess grandmasters play
chess so well in large part because they are able to judge positions with
heuristics of this sort and are able to quickly concentrate on only the best
moves in a given situation. They are able to ignore superflous moves that a
lesser player has to ponder over. However, grandmasters are also able to
quickly recognize all those situations when general rules don't apply. This is
the part computers have trouble with.
At the present time, the heuistic chess programs have been replaced with "brute
force" programs that pretty much look at all possible moves (millions of them)
in a given position and rely very little on any kind of "judgement" heuristics.
Deep Blue, the program that beat world champion Kasparov was just such a
program, as are the very strong commercially available chess programs.
I feel that the tth program suffers from this same problem. It appears to rely
quite heavily on the attempt of the program to judge poker situations using a
set of pre-ordained "rules". Therefore, the program seems, in my opinion, to
have some trouble understanding exceptions to the general rules in a given
situation. It is the ability to judge these exceptions that make real world
players, even at the lower limits, stronger and more resourceful than the
For example, the new version of tth check raises more than the old version.
However, the tougher opponents seem a bit too sensitive when they themselves
are check-raised, and they respect this play far too much. Once you realize
this against specific opponents it is possible to bluff them off of hands far
too easily. I have never been able to do this in an actual game with humans
even against tighter players. This is one of those situations where the
program simply is not able to learn from experience or recognize what the human
opponent is up to.
In any case, my concern is that by using simulations, any errors of this type
might be magnified by the program. Perhaps you can give me you opinions in
Muck AQ to agression by an opponent that is that tight. My rule
of thumb with AQ in a tight game is: if other people like their
hands, you don't like yours.
>Criticisms are welcome...
I'll second the poster that said "wow!" An excellent report that must
have been a huge amount of work. Much appreciated. This report is
what makes wading thru the stupid flaming
worth the time.
BTW, I keep posts that I find interesting, informative, or valuable.
Looking back, over 50% are from one poster... Abdul.
Of course, if I kept posts that made me smile, giggle, or laugh, then
Mike Caro would be the man.
>It's hard to do this math correctly, however. Barbara?!
>It may ultimately hinge on how well your opponent plays
>his AK. If he gets away from AK when a queen flops, as he should,
>you could be in trouble with AQ here.
>What do you think? Call or muck AQ here in the big blind versus
>a raise from 99/AJs/AQ or better?
Going back and reading the quoted section in context, I think the
raise was from a middle position tight raiser, which indicates a
strong hand, unless he is know for long distance steals. You say
the set of possible hands are:
AA 3 ways
KK 6 ways
QQ 3 ways
JJ 6 ways
TT 6 ways
99 6 ways
AKs 3 ways
AQs 2 ways
AJs 3 ways
AK 9 ways
AQ 7 ways
total 54 ways
You are getting 3.5:1 odds from the big blind. However, you are
getting the worst of it by far as all the other hands do better
Here are the rundowns (I eroneously assume that no suits are shared):
wins losses ties
(vs. AA )% cmp2 as qd ah ac 123893 1567203 21208
(vs. KK )% cmp2 as qd kh kc 483206 1223243 5855
(vs. QQ )% cmp2 as qd qh qc 506094 1185055 21155
(vs. JJ )% cmp2 as qd jh jc 727947 978874 5483
(vs. TT )% cmp2 as qd th tc 727707 979380 5217
(vs. 99 )% cmp2 as qd 9h 9c 754286 952801 5217
(vs. AKs)% cmp2 as qd ah kh 380742 1256059 75503
(vs. AQs)% cmp2 as qd ah qh 37211 122557 1552536
(vs. AJs)% cmp2 as qd ah jh 1139313 479013 93978
(vs. AK )% cmp2 as qd ah kc 405624 1228082 78598
(vs. AQ )% cmp2 as qd ah qc 37211 37211 1637882
% loss % against product
(vs. AA )% cmp2 as qd ah ac 0.92 0.056 0.052
(vs. KK )% cmp2 as qd kh kc 0.71 0.111 0.079
(vs. QQ )% cmp2 as qd qh qc 0.69 0.056 0.039
(vs. JJ )% cmp2 as qd jh jc 0.57 0.111 0.063
(vs. TT )% cmp2 as qd th tc 0.57 0.111 0.063
(vs. 99 )% cmp2 as qd 9h 9c 0.56 0.111 0.063
(vs. AKs)% cmp2 as qd ah kh 0.73 0.056 0.041
(vs. AQs)% cmp2 as qd ah qh 0.07 0.037 0.003
(vs. AJs)% cmp2 as qd ah jh 0.28 0.056 0.016
(vs. AK )% cmp2 as qd ah kc 0.72 0.167 0.120
(vs. AQ )% cmp2 as qd ah qc 0.02 0.130 0.003
total % loss .542
% win % against product
(vs. AA )% cmp2 as qd ah ac 0.07 0.056 0.004
(vs. KK )% cmp2 as qd kh kc 0.28 0.111 0.031
(vs. QQ )% cmp2 as qd qh qc 0.30 0.056 0.017
(vs. JJ )% cmp2 as qd jh jc 0.43 0.111 0.048
(vs. TT )% cmp2 as qd th tc 0.42 0.111 0.047
(vs. 99 )% cmp2 as qd 9h 9c 0.44 0.111 0.049
(vs. AKs)% cmp2 as qd ah kh 0.22 0.056 0.012
(vs. AQs)% cmp2 as qd ah qh 0.02 0.037 0.001
(vs. AJs)% cmp2 as qd ah jh 0.67 0.056 0.038
(vs. AK )% cmp2 as qd ah kc 0.24 0.167 0.040
(vs. AQ )% cmp2 as qd ah qc 0.02 0.130 0.003
total % win .290
Which gosh darn it, is about 2:1 odds. So I guess you should
call when given those 3.5:1 odds.
The different suit assumtion is slightly optimistic, and you'll
have to go to the river to get these odds, and you'll have a
positional disadvantage all the way through, and the raiser is
assumed to raise the above hands, and only the above hands
*all the time* in middle position when opening, and...
>BTW, I keep posts that I find interesting, informative, or valuable.
>Looking back, over 50% are from one poster... Abdul.
Same for me. I've read Abdul's preflop work twice
and it's sinking in and making sense. There is a lot
here to digest and many items that directly contradict
my own play book. As with most new ideas I'll need to
implement them in small units and get comfortable with
them slowly. I have little doubt that the underlying reasearch
is the most rigorous to date. Thanks Abdul.
Here's what I do when the flame wars get to be too much.
I mainline str8 Abdul posts at Dejanews. Here's a link:
Ed (no, not that Ed!)
Friends may come and go
but enemies accumulate.
>I mainline str8 Abdul posts at Dejanews. Here's a link:
The link didn't fly. I'll try again.
or perhaps this will work this time
According to pr...@acm.org <Andrew Prock>:
>Which gosh darn it, is about 2:1 odds. So I guess you should
>call when given those 3.5:1 odds.
>The different suit assumtion is slightly optimistic, and you'll
>have to go to the river to get these odds, and you'll have a
>positional disadvantage all the way through, and the raiser is
>assumed to raise the above hands, and only the above hands
>*all the time* in middle position when opening, and...
One slight but. Since you won't be getting other callers down the
road you'll probably be keeping the total $ in the pot to 1 bet
4.5 SB preflop
2 SB postflop
4 SB postturn
4 SB postriver
You'll have put in 6 SB to the 14.5 SB in the pot, which makes
your "true" odds you are getting more like 2.5:1.
This makes it a lot closer of a call than the seemingly good
3.5:1 odds you're getting to see the flop.
But, it still is a reasonalbe call as long as all the assumptions
This is a very thoughtful and comprehensive piece of work.
Thank you very much for posting it.
I've already printed out 3 copies and marked up 2.
Live long and prosper,
This is a very thoughtful and comprehensive piece of work.
Thank you very much for posting it.
I've already printed out 3 copies and marked up 2.
Live long and prosper,
That appears to be the main thrust of certain elements in your preflop
strategy. And those are really the elements which set it apart from
more traditional approaches. I think that line would be a good addition
to the intro to the article, as it makes clearer for the reader much of
the thinking behind the strategy.
I might quibble a bit, however, with the assertion that this preflop
accomplishment "gets you most of the way to victory". As you know, it's
generally accepted among top hold'em players that most of your profit
comes after the flop. There are games which are exceptions, but
postflop play is typically key to winning more than a little.
Could it be, though, that for a very good player a shift from a
"standard" preflop approach to yours would add so much to his profits
that it would now be safe to say that most of his edge comes before the
flop?? I guess that's possible. But I think that the answer will lie
in the hourly rates of skilled players who apply your approach over many
hours in games where the maximum hourly rate obtainable through the
traditional approach has been relatively well established.
It seems to me that your preflop opening standards are not really far
from those used by many better players influenced by Sklansky and
Malmuth and others. What stands out most saliently as a unique feature
of your approach is the liberal use of the limp-reraise and the
variations in play that it appears you incorporate in order to allow for
this play without opponents adjusting to it too easily. What do the
sims say happens to the EV of a hand like 88 when you limp-reraise with
it. I would think that you give up a good deal of EV there in order to
add more EV to your play of big pairs, no? Just wondering.
I haven't investigated doing sims such as yours with recent versions of
TTH, but I've played a bit against TTH V3. Does the quality of play of
the TTH V3 advisors for short handed games have implications for the
simulation results? I ask because I've now played a fair number of
hands in heads up and three handed games against the TTH advisors 2 and
3, and found them to be fairly easily beatable after a pretty short
period of familiarizing myself with their play. In other words I found
it rather easy to adapt quickly to the play of the TTH opps, and then
pretty much crush them, while they apparently cannot (significantly)
adapt to my play. It wasn't long before I actually had to purposely
avoid plays that I knew took advantage of typical, rigid TTH patterns in
order to obtain more realistic "practice". (But lest this be
interpreted as a slam to the product I should add that TTH does do a
surprisingly good job. It's well ahead of where the old simulators
were, and does make you work a bit to beat it. It's just not like facing
a real life lineup of thinking players.)
But maybe this doesn't matter. Maybe the important thing is really to
come up with the appropriate "baseline" strategies with variations to be
developed as one plays with adapting human opponents. I would not say
that TTH was easily beatable for the first 50 hands or so. So maybe
we're talking about a first level of play, that which occurs prior to
the adaptations thinking players will make. Then, maybe, finding the
correct hands and playing strategies for this level holds the bulk of
the value of correct preflop play. That seems at least plausible.
But maybe it *does* matter. I think that a huge portion of your profit
in a game comes from your play *after* some adjustments have been made.
It actually *results* form your adjustments. e.g., much of your edge
comes from players having adjusted to you, and you therefore having the
opportunity to exploit what they're thinking about your play.
One place where I'd wonder about the implications of the TTH opps'
failure to adjust is with regard to the limp-reraise. I see that you
aim to add enough mix to your play that it should sort of camouflage
this play, but I wonder how it will stand up to the adjustments of real
world opps. Maybe it stands up quite well; I just wonder.
I'm not clear on the implications of all these things. I'm certainly
not trying to pick apart your sims or analyses. These are just some of
the questions that come to mind. Maybe I would need to dig into doing
some similar sims in order to fully appreciate their intricacies and
therefore their strengths and limitations.
> > In any case, my concern is that by using simulations, any errors of this type
> > might be magnified by the program. Perhaps you can give me you opinions in
> > this regard.
> Most people side with you. Sklansky proposes the alternative of no
> fold'em simulations, adjusted for how "scared" the hands are (implied odds
> and reverse implied odds), arguing that this is much better than a true
> TTH type simulation.
The "brave hand/scared hand" concept is not simply implied odds and
reverse implied odds. Though implied and reverse implied odds
situations provide clear *examples* of brave and scared hands, it is not
the equivalent concept. Sklansky has clarified this over on 2+2. It
covers a broader range of situations, with the common element being the
question of whether or not the existence of future bets increases a
hand's EV. This can become relevant in ways that have little to do with
implied and reverse implied odds. See the threads on 2+2 for more
detail. Just wanted to clarify that.
I think that what hangs people up on these sims is simply this question:
Can simulation data with true value for the real world be obtained from
a simulator which does not really *play* all that well against a human
opponent? The TTH ring games are as easily beatable (or more so) than
the short handed games. That includes the "tough" lineup. While I
haven't played hundreds of thousands of hands against TTH, it only takes
so long to seen that one can soundly beat any combination of TTH players
you can put together. Though there appear to be some elements of
adaptability in the profiles' play, the general rigidity and lack of
adjustment seem to be key among TTH's weaknesses as a *player*. Those
weaknesses give it a feel different from a real game. Just how that
affects the conclusions we can draw from the sims is something I'd like
to see discussed further.
I look forward to your comments (even if you do quietly delete them
I've been getting questions on this chart in email, so apparently
it's not clear. Would the following be more clear?
[Begin proposed replacement text]
The following table ranks the opening hands. The higher in the table,
the stronger the hand is and the farther to the right of the button
it can open. The bottom part of the table is geared towards late
Opening Hand Rankings
Pairs Suited Hands Offsuit Hands
===== ============================ ============================
TT AQs AK
99 AJs AQ
88 KQs ATs
77 QJs A9s KJs AJ
66 JTs A8s KTs QTs KQ AT
55 T9s A7s K9s J9s QJ KJ
44 98s A3s K7s Q9s T8s
33 87s A2s K6s Q8s 97s J8s JT A7 KT QT
22 76s K4s Q6s 86s T7s T9 A6 K9
65s K2s Q4s 75s 98 A3 K7 Q9 J9
54s Q2s 64s 85s Txs 87 A2 K6 Q8 T8 J8
43s 42s 63s 7xs 76 K5 Q7 T7 J7
32s 62s 65 K4 Q5 86 96 J6
54 K3 Q4 75 85 J5
43 K2 Q2 53 63 J2
So, for example, TT, AQs, and AK are equally strong when opening.
When a hand is "missing", it belongs to the lower position. For example,
in the table there is AT and then there is A7; the missing hands, A8 and
A9, belong with A7. Jxs belongs with Txs; Jxs refers to baby suited jacks,
J2s through J7s, and similarly for Txs, 9xs, 8xs, and 7xs (6xs is just 62s.)
An easy way to refer to a whole group of hands is by listing the
minimum pair, suited, and offsuit hands. Use the first columns of the
suited and offsuit hands as an index. For example, a minimum
of 77/QJs/AJ caliber hands refers to the following set of hands, clipped
from the table above:
TT AQs AK
99 AJs AQ
88 KQs ATs
77 QJs A9s KJs AJ
Using this notation, here are the minimum openers by position:
Conservative Minimum Openers
6 off button 88/KQs/AQ
5 off button 77/QJs/AJ
4 off button 66/JTs/KQ
3 off button 55/T9s/QJ
2 off button 44/98s/QJ
1 off button 33/87s/JT
on the button 22/65s/98
"6 off button" means 6 players to the right of the button, which is the
case when you are "under the gun" (first to act) in a 9-handed game.
The minimum openers table assumes your opponents are a tad on the loose
side, which forces you to open with less hands than normal. You can
open with more hands if your opponents are either very loose or too
tight or even properly tight. In early position you can usually open
one caliber looser (e.g., 88/KQs/AQ becomes 77/QJs/AJ), and if your
opponents are extremely tight or loose you can go two calibers looser.
In late position, open with fewer hands against very loose opponents,
due to your inability to steal the blinds. However, against overly
tight players in the blinds, you can open with any hand listed in the
[End proposed new text.]
Is that more clear? I showed that minimum openers table to one
person before posting and he thought it was too confusing and too
cumbersome to have to flip back to the other table, so I got
rid of it and added the "# off button" column to the rankings table,
but readers found that too confusing.
> Abdul Jalib wrote:
> > pau...@aol.com (Paul950) writes:
> > > ...My only concern is the fact that you base your results on tth simulations.
> > > (Do you test these results in real games? What has been your real-life
> > > experience with these starting requirements?)
> > Of course I use much of what I learn from Turbo sims in live games.
> > I have found the recommended style fairly devastating to my opponents;
> > by that I mean I can see that the plays have their desired effect.
> > Luring your opponents into playing dominated hands while avoiding
> > playing dominated hands yourself gets you most of the way to victory.
> That appears to be the main thrust of certain elements in your preflop
> strategy. And those are really the elements which set it apart from
> more traditional approaches. I think that line would be a good addition
> to the intro to the article, as it makes clearer for the reader much of
> the thinking behind the strategy.
Yes. Something like that may have been in there before. I condensed
whole paragraphs into single sentences in an attempt to keep it short.
> I might quibble a bit, however, with the assertion that this preflop
> accomplishment "gets you most of the way to victory". As you know, it's
> generally accepted among top hold'em players that most of your profit
> comes after the flop. There are games which are exceptions, but
> postflop play is typically key to winning more than a little.
Unfortunately I cannot really argue with you here, as it would be a
silly argument. I can argue that table selection is the most important
thing, as a good player with bad game selection skills might be only
breakeven in expectation. Or I can argue that hand selection is the
most important thing, as someone who plays very well after the flop
but has horrible hand selection skills will be destroyed. Or I could
argue that any number of other things are most important, but really
everything is interdependent.
But anyway, once you enter a pot, pot odds dictate that you should
continue contesting it quite frequently, so your preflop play tends
to set you on a course for the rest of the hand. If you are calling
tight raises when you have KJ (or even KQ), no matter how creative
your postflop play is, you're not going to be able to reduce the
effect of this disaster by much.
> Could it be, though, that for a very good player a shift from a
> "standard" preflop approach to yours would add so much to his profits
> that it would now be safe to say that most of his edge comes before the
> flop?? I guess that's possible. But I think that the answer will lie
> in the hourly rates of skilled players who apply your approach over many
> hours in games where the maximum hourly rate obtainable through the
> traditional approach has been relatively well established.
It depends on your opponents anyway. Against weak-tight opponents,
exploitive postflop play would allow you to steal the pot often.
> It seems to me that your preflop opening standards are not really far
> from those used by many better players influenced by Sklansky and
> Malmuth and others. What stands out most saliently as a unique feature
> of your approach is the liberal use of the limp-reraise and the
> variations in play that it appears you incorporate in order to allow for
> this play without opponents adjusting to it too easily. What do the
> sims say happens to the EV of a hand like 88 when you limp-reraise with
> it. I would think that you give up a good deal of EV there in order to
> add more EV to your play of big pairs, no? Just wondering.
It costs just 8 cents in a tight $10-$20 TTH simulation to limp-reraise
with 88 instead of open-raising, and this 8 cents is not a statistically
significant difference. In a tight game in the real world, I would expect
limp-reraising with 88 to be more effective, if anything, because real
players do not expect to see this from 88, and they are likely to lay down
overcards against it in a bloated pot, which is a major mistake.
In Turbo, limp-reraising with 77 is fairly costly, a $1.70 difference.
Obviously, the actual EV's depend on the opponents, and sure
you should view the sim with skepticism. However, I believe I can
rationalize these results. Medium pairs do best with either many
opponents or one opponent. Either case is fairly likely if you limp.
If you are raised and everyone else folds, you probably have
the best hand and so you can reraise for value. (You should also
reraise if it would drive out a limper caught in the middle.) Once
the flop comes, it may not be so easy to put in that extra bet
profitably. Medium pairs are both stronger and more versatile than
commonly believed, I believe.
If you limp with 88, are raised and then there are two cold caller
callers, then reraising would likely be very unwise. Don't follow
my recommendations blindly. There are an infinite number of
situations. "Think in terms of your goals," as Jim Geary says.
In my strategy, I suggest limp-reraising with AA, KK, 99, 88, AJs, and
ATs, and half the time with AK, AKs, and AQs. This mixes the hands that
dominate a raiser with the hands that are dominated by a raiser. When
you limp-reraise you could terrorize your opponent into laying down his
AQ (if the flop doesn't hit him) when you hold ATs or 88, which would be
very nice indeed. This is especially true if he has seen you limp-reraise
with aces or kings. It's a bit of a gambit, I admit. And against players
who raise with only their very best hands and/or can't be made to fold
by the river, it would be suicidal.
Overall, the mix of plays I chose is both subtle and evil. :)
> > > In the early days of chess programming, there was a great deal of enthusiasm
> > I know that sims have had troubled histories in chess and backgammon.
> > I believe heads-up poker is a different beast. You simply have to contest
> > the pot any time you have a reasonable hand, and sometimes when you don't.
> > Making good laydowns is a dangerous way to try to win. Raising too much
> > will also make you exploitable. So you're often going to the showdown,
> > you can't get a good read on your opponent, and you can't get too fancy.
> > Basically, playing a good game of heads up hold'em is not nearly so
> > complicated as playing a good game of chess. Good heads up players
> > devastate weak heads up players simply because weak heads up players don't
> > have a clue about how often to contest the pot.
> > Full table tight games play very similarly to heads-up games, since you
> > are usually heads up by the flop.
> I haven't investigated doing sims such as yours with recent versions of
> TTH, but I've played a bit against TTH V3. Does the quality of play of
> the TTH V3 advisors for short handed games have implications for the
> simulation results?
Only for blind stealing/defense situations.
> I ask because I've now played a fair number of
> hands in heads up and three handed games against the TTH advisors 2 and
> 3, and found them to be fairly easily beatable after a pretty short
> period of familiarizing myself with their play. In other words I found
> it rather easy to adapt quickly to the play of the TTH opps, and then
> pretty much crush them, while they apparently cannot (significantly)
> adapt to my play. It wasn't long before I actually had to purposely
> avoid plays that I knew took advantage of typical, rigid TTH patterns in
> order to obtain more realistic "practice". (But lest this be
> interpreted as a slam to the product I should add that TTH does do a
> surprisingly good job. It's well ahead of where the old simulators
> were, and does make you work a bit to beat it. It's just not like facing
> a real life lineup of thinking players.)
Yes. There is one major exploitable "bug" in the short-handed play
with the "adjust to check-raises" flag on. The next version will likely
fix it or maybe there will be a patch.
To a certain extent the simulation results could be swayed by such
holes, but I'm hoping it all sort of evens out. The Turbo players
do not seek out such holes, and, for example, limp-reraising preflop
doesn't seem to have a significant effect on whether one of these
holes will be stumbled into.
> But maybe this doesn't matter. Maybe the important thing is really to
> come up with the appropriate "baseline" strategies with variations to be
> developed as one plays with adapting human opponents. I would not say
> that TTH was easily beatable for the first 50 hands or so. So maybe
> we're talking about a first level of play, that which occurs prior to
> the adaptations thinking players will make. Then, maybe, finding the
> correct hands and playing strategies for this level holds the bulk of
> the value of correct preflop play. That seems at least plausible.
Yes, I'm striving for sort of a defensive baseline strategy. Once you
figure out how to exploit your opponents you can deviate. If you are
much better than your opponents, your EV on all hands goes up, and so
you can play more hands than I suggested.
> But maybe it *does* matter. I think that a huge portion of your profit
> in a game comes from your play *after* some adjustments have been made.
> It actually *results* form your adjustments. e.g., much of your edge
> comes from players having adjusted to you, and you therefore having the
> opportunity to exploit what they're thinking about your play.
You always have to be thinking one level deeper than your opponents
or else you are in trouble and should instead play conservatively.
e.g., if you think they think you are speeding but they know you
think they think you are speeding, you are in trouble, unless you
ignore what you think and play your normal baseline way. I think.
> One place where I'd wonder about the implications of the TTH opps'
> failure to adjust is with regard to the limp-reraise. I see that you
> aim to add enough mix to your play that it should sort of camouflage
> this play, but I wonder how it will stand up to the adjustments of real
> world opps. Maybe it stands up quite well; I just wonder.
My experience is that real world opponents do not adapt to
limp-reraises. They walk into the trap time and time again. This
is due partially to my limp-reraising only rarely. I do it with
just a few hands, and if they don't raise I can't reraise, and sometimes
I don't carry through with the reraise when I wind up heads-up. So,
they never learn that they are in dire danger if they raise my limps.
I am not sure why people are so stupid, and I wouldn't be surprised
if eventually they learn. Even so, my strategy is designed to be
nonexploitable. But it might be best to fold 55, 66, A7s, and KQ in
early position until opponents become hesitant to raise my limps.
In a hold'em game I play with friends, they *are* terrified to raise
my limps, but this only seems to be because I have explained my
limp-reraise philosophy to them, and so they fear the possibility,
even though they haven't often seen it actually happen. (I'm reminded
of what Mike Caro wrote recently about talking like a fish for
advertising... you don't actually have to show down garbage, just talk
about garbage hands that you supposedly played.)
I expect limp-reraises in the real world to work better than in Turbo
for the weaker limp-reraising hands, worse for the strongest limp-reraising
hands. All human players put you on AA or KK if you limp-reraise. The
Turbo players quickly forget that you limp-reraised (by the turn they have
totally forgotten.) I tried to avoid even putting AA and KK into the
limp-reraising hands, and I did avoid it for loose games, but for tight
games even limp-calling is better than raising with AA, because stealing
the $15 in blinds really lowers your average win of $45-$80 for pocket
aces in a $10-$20 game. You'd rather give the big blind a free shot
to catch a pair than take down the blinds when you have AA.
If the game is a bit looser, so that stealing the blinds is not a big
concern, then switch to raising with AA and KK, and limp-reraising with
QQ, JJ, TT, and 99 (along with AK and AKs at least part of the time.)
This is not easily exploitable, and offers better deception against
players who don't know how you play at all.
> I'm not clear on the implications of all these things. I'm certainly
> not trying to pick apart your sims or analyses. These are just some of
> the questions that come to mind. Maybe I would need to dig into doing
> some similar sims in order to fully appreciate their intricacies and
> therefore their strengths and limitations.
> I think that what hangs people up on these sims is simply this question:
> Can simulation data with true value for the real world be obtained from
> a simulator which does not really *play* all that well against a human
> opponent? The TTH ring games are as easily beatable (or more so) than
> the short handed games. That includes the "tough" lineup. While I
> haven't played hundreds of thousands of hands against TTH, it only takes
> so long to seen that one can soundly beat any combination of TTH players
> you can put together. Though there appear to be some elements of
> adaptability in the profiles' play, the general rigidity and lack of
> adjustment seem to be key among TTH's weaknesses as a *player*. Those
> weaknesses give it a feel different from a real game. Just how that
> affects the conclusions we can draw from the sims is something I'd like
> to see discussed further.
As I said in another article, I think you may get better sim results
with the Turbo players being rather dumb and unable to exploit each
other. If the players exploited each other, then the results would
be a magnification of the holes, and you'd wind up with some very
strange results indeed that would not be generally applicable in
the real world.
In any case, I'm not asking Turbo for rocket science. With a chess
analogy, I'm trying to find out whether bishops are more valuable
than pawns. With a mediocre chess simulator, I probably cannot
determine whether bishops are more valuable than knights, but I can
determine that they are close to equal in value, which is useful
to know. The danger is that the simulator may use its knights very
poorly, resulting in their appearing much less valuable than bishops.
Turbo certainly misplays small pairs and KQ in certain circumstances
if you stick to the Advisor_T profiles (but you can try different
postflop strategies in an attempt to partially address
> I look forward to your comments (even if you do quietly delete them
> later ;-)).
> John Feeney
I'm gonna step in here on the side of Abdul's opinion. Most of the critisms
of TTH as "not a serious research tool" are about using TTH as a game to do
research. Using it to do simulations is very different from playing it as a
You can use the game features to take a look at how the play is going and do
some sight validation that way, Abdul has talked about this before. The game
feature is good for that, and you will find some little glitches when you do
The real strenght of TTH over all the alternatives is the amount of control
you have over the profiles. You don't make money in poker by being able to
distinguish between "scared" and "brave" hands. You make money by being able
to recognize and exploit mistakes made by your opponents. TTH helps a great
deal in learning the best way to exploit certain kinds of opponent mistakes.
Abdul's starting hands for tight tables, for example, shows you what starting
hands tend to have the most potential for exploiting the mistakes of players
who play to tight. This is of tremendous value. I don't know of a better way
to do it.
Expert intuiotion just doesn't get it. Human intuition is notoroious for
being wrong in low probability/high impact events. Gambling is full of those
kind of events.
> Have you checked out what the group at www.cs.ualberta.ca/~games/poker
> has been doing? Their combination of opponent modeling and selective
> search looks to be promising. Perhaps when they release the code, you
> and others could use it to check your (plural) results.
I check out the website every once in a while. So far it doesn't look like
they've gotten real far. But, I'm sure they will.
-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own
I think the emailer had to be overstating it to say that you'd be
destroyed. Even against very tough short handed players, if you
yourself are a very tough short handed player, then I think you should
come somewhere close to breaking even in this spot - though you may show
a loss. Your sims would not seem to disagree with this. But clearly
those are very marginal hands. In the ideal case they should show at
least a modest profit for an excellent player (who's good in blind
stealing/short handed situations) going after the blinds of a couple of
weak, passive players. But a mediocre player attempting to steal the
blinds of a couple of very good, aggressive players *will* be destroyed
with those hands. I'd think that the majority of players would probably
be giving up little if anything to pass them in all but very good
situations. I frequently pass on such hands just to save myself the
bother of boosted fluctuations with little profit to show for it.
That's not my intuition, but the output of a strange kind of organic
electrochemical computer. Oh, yeah, I guess it is my intuition. :)