Showing posts with label math. Show all posts
Showing posts with label math. Show all posts

Thursday, October 7, 2010

The Problems with a Polarized Range by Matthew Janda

http://www.cardrunners.com/article/110/the-problems-with-a-polarized-range/


The Problems with a Polarized Range
                A player polarizes his range when he raises all his very strong hands and some bluffs, and calls with all his medium strength hands. While the former range is very easy to play, the latter is extremely difficult to play against competent opponents who know their opponent has no strong hands in his range. Since good players will attempt to exploit their opponents using polarized ranges by over-betting to maximize their bluff frequency, you should attempt to avoid polarizing your ranges against opponents good enough to punish you for having no very strong hands in their range. 
                In order to best illustrate this concept, it’s best to start off with an extreme example of the difficulties you’ll face when you only have medium strength hands in your range. Suppose you are playing 6-max with 100 big blind stacks and under the gun opens for 3.5 big blinds and you are the only caller from the button. Your preflop range is JJ-22, KQs, JQs, JTs, T9s, 98s, AQ, AJs, and ATs, which you believe to be very close to a theoretically optimal flatting range and your opponent knows this. Suppose the flop comes 2h 2s 2d.
                Unfortunately, through no fault of your own, you received the worst possible flop for your range and your range contains no very strong hands. Raising this flop would make little sense, since your opponent knows he cannot be beat when he has JJ-AA, and has 28% equity against the top hand in your range with AK, AQ, and KQ. Instead you are forced to call or fold on the flop, and will frequently face turn and river bets and will often end up folding the best hand. Suppose you decide to flat JJ-77, AQ, and KQs. Once again, this is the same situation one finds himself in when he polarizes his range by only calling with medium strength hands.
                Your opponent will want to maximize his bluff frequency, and in order to do this he will bet equal fractions of the pot on all three streets. The pot will be 8.5bb on the flop, and in order to make the pot 201.5bb on the river your opponent must make the pot grow at the rate of “r” each street, where 8.5r3 = 201.5bb. Since r = 2.87, your opponent will want to be around 93% of the pot on each street (since his 93% pot sized bet and your call will make the pot grow at the desired rate). Since your opponent cannot recklessly fire all three streets, your opponent will have to give up bluffing some air hands on the turn and some air hands on the river. Let’s take a look at what frequency your opponent can be bluffing on every street before you can start calling with your medium strength hands profitably.   
Your opponent will want to make you indifferent to calling on the river, so 67.5% of his 93% pot sized bets on the river must be for value. An easy was to visualize why your opponent needs to be betting for value 67.5% of the time on the river when he makes a 93% pot sized bet is to imagine the pot is $100 on the river and your opponent makes a $93 bet. When you call your opponents bet and win, you’ll win both the original pot and his river bet for $193. When you call and lose, you’ll lose only $93 (the money you already lost in the pot is dead money). If you win $193 only 32.5% of the time and lose $93 the remaining 67.5% of the time when you call with your medium pocket pairs, you will be indifferent to calling  and folding since ($193 x 32.5%) - ($93 x 67.5%)  = 0.  Since calling is effectively the same thing as folding on the river and have an expected value of 0, in order to not be getting odds to call the turn your opponent must follow through on the river 67.5% of the time (because every time you face a river bet, calling and folding will both have an expected value of 0). The same logic applies on the flop, so in order to make you indifferent to calling the flop with JJ your opponent must follow through on the turn 67.5% of the time. In other words,
1)      On the river, 67.5% of your opponent’s bets must be for value.
2)      On the turn, 45.5% (0.675 x 0.675) of your opponent’s turn betting range must be able to value bet the river.
3)      On the flop, 30.7% of your opponent’s flop betting range (0.675 x 0.675 x 0.675) must be able to value bet the river.
Note how on each street our opponents range consists of more and more value hands and less bluffs, so the more betting rounds which remain the more your opponent can bluff. You can see how problematic this is to play against. While you will surely hit a six outer with AQ or spike a two outer with a pocket pair on some turns or rivers, your opponent will often do the same to you (especially since most of his bluffs have such high equity). It’s not particularly difficult for him to be able to value bet the river with 30.7% of the hands he continuation bets on the flop. Many players at NL$200 and lower think “Well, pocket fives have over 50% equity against an UTG’s continuation bet range on a 222 or 664 flop, and since I’m getting better than 2:1 this is an easy call.” Yet this is not the case. Even if your hand has a lot of equity against your opponents continuation betting range on the flop, too often you’ll be forced to fold it on the turn or river,  or end up making an unprofitable river call. There are some flops from certain positions you just cannot prevent your opponent from firing 100% of his range on. If you are lucky enough to hold 66 or TT on a 6c 6h Td flop in the small blind against a cutoff open, you need to check-call those hands to defend against your opponent who knows your range is weak and will likely be double and triple barreling many turns and rivers.
Now instead imagine another situation where the cutoff opens and you are the only caller from the button. The flop comes Qh 7s 2c, and once against your opponent continuation bets. Since this flop connects with your range around as well as his, your opponent will probably bet less than 90% of the pot and won’t be able to get it in by the river unless he overbets the turn or river.  In order to avoid the situation you were faced with in the previous example, you should refrain from polarizing your range and instead flat most of your sets on the flop. While AQ and KQ might be good hands to raise (along with some air) on the flop, it’s more useful to flat with your sets and raise them on the turn. This prevents your range from consisting only of hands QJ and worse and polarizing your range. In addition, sets don’t need fear falling behind when an overcard falls or your opponent drills two pair.
                Nevertheless, there are still situations when polarizing ones range is correct.  Two of the most common ones are when...
1)      Your opponent isn’t good enough to realize your range is polarized, or won’t punish you if he does.
2)      You are in position and the board has a lot of draws, so the vast majority of turns will put some nut type hands in your range as soon as the turn card hits.
For example, suppose the cutoff opens and you are the only caller from the button. The flop comes Ts 9s 5h, and the cutoff makes his standard sized continuation bet. Polarizing your range by raising all your two pairs, sets, monster draws, and air is perfectly acceptable in this case, because just about every turn card will put some nut type hands in your range. Any spade gives you a possible flush, any ten or nine brings possible trips, and any ace, king, queen, jack, or eight gives possible two pairs or straights. So 32 of the possible 49 cards which can turn, or about 2/3 the deck, will put nut type hands in your range and protect you from an opponent who wants to exploit polarized ranges by over-betting. Likewise, slow playing a set or two pair will often result in you either losing action or getting drawn out on many of the previously mentioned turn cards, making slow playing usually a bad idea.
                Recognizing when your opponents range is polarized provides a great way to increase your win rate.  If you do open from middle position and get a caller from the SB, don’t be afraid to bet large amounts on a very dry 955 flop (whether it’s rainbow or there are two of one suite), especially when your stack is a bit over 100 big blinds. You know your opponent can’t be strong unless he has 99 or 55, which only make up 4 hand combinations. Likewise, don’t be afraid to punish opponents when you know they can’t be strong after they check-call on a draw heavy board. if you know your opponent would never call with a hand better than KQ on a Qc Ts 6c 2s board on the turn (since he would likely check-raise AQ or better on the flop and the 2 on the turn doesn’t ever improve him), don’t be afraid to over-bet jam the river against an opponent as a bluff when the river comes a 3d. While you and your opponents will undoubtedly not play perfectly, realizing the theoretical problems polarized ranges have when facing over-bets will both allow you to better exploit your weaker opponents, and slowly gear your play style towards a more theoretically optimal style to use against opponents equally capable of exploiting you.
                Thanks for reading, and good luck at the tables.
By Matthew Janda

Wednesday, September 29, 2010

Mathematics of NLHE Ep 4

Covered in this episode:
  • EV calculations
  • EV calculations
  • and more EV calculations
Expected Value (EV) is the long term expected outcome of a given hand or situation, either positive (+EV), negative (-EV), or neutral (0EV).

EV Calculations 101
Basic EV calculations setup:
  • EV = [result of win] - [result of loss]
  • To expand a bit:
    • EV = [Our Equity] * [what we win] - [Villain's Equity] * [what we lose]
  • Basic Example:
    • We're in the big blind with AsAc. The UTG player shoves with KhKd and folds to you. It's $900 to call to win $1015 (stack + blinds) 5/10nl, $1000 stacks.
    • EV = (0.81 * 1015) - (0.19 * 990)
    • EV = 822.15 - 188.1 = +$634.05
  • Alternative Method:
    • EV = [our equity] * [total pot] - cost of our call
    • EV = (0.81 * 2005) - 990 = +$634.05
Example 1 - 98o on 7T24 vs AA. We need to call 650 to win 700, what is our EV for calling?
  • EV = 0.18 * [650 + 700] - 0.82 * 650
  • EV = 243 - 533 = -$290
Example 2 - Villain opens from the CO to $35, we 3-bet him again OTB to $125, Villain thinks for about 2 second and calls.
  • First let's put him on a range: Mostly pairs, AQ (he would 4-bet AK a lot), occasionally a SC or AXs type hands.
  • On the flop Kc9c8d, Vil checks, we bet $200 into $265, he check/raises all in for $875 meaning it costs us another $675 to call. Do we call or fold?
  • Let's evaluate his range now that we have information: He always has 8 outs+ if drawing (OESD, FD or better), and let's assume he 4-bets AK 100%, but he could call AA/KK planning to trap.
  • Based on our pre-flop and flop range, we now get a narrowed down range of: KK+,99,88,AQcc,AJcc,ATcc,A8cc,QJcc,QTcc,JTcc,98s,87cc,JTs,67s. Against this range, how is our hand doing?
    • Against the range above, our hand has about 35% equity.
  • EV(call) = 0.35 * (875 + 200 + 265) - 0.65 * 675
  • EV(call) = 469 - 438.75 = +$30.25

WoT's TUPAC method
  • Following these 4 steps will help guide you to mentally calculate your equity against a hand range while at the table:
    1. Tally Up the hand combinations.
    2. Pair combos to known equities.
    3. Analyze unpaired combos.
    4. Combine the analysis to estimate.

    Cont. on Example 2 using TUPAC method
    1. Tally Up the hand combinations
      • Let's break apart his range into the 3 main categories of hands: Those that crush us, those we're flipping with, and those we're decent favorite over.
        • Crush us: AA, KK, 99, 88, 98s (5 hands, 13 combos)
        • Flipping with us: AQcc, AJcc, ATcc, A8cc, QJcc, QTcc, JTcc, 87cc, 67cc (9 hands, 9 combos)
        • Decent favorites against: 67s, JTs (2 hands, 6 combos)
    2. Pair combos to known equities
      • Start matching up hands that crush us and the flips, those should even out between 25 to 35% equity depending on how badly we're crushed. A set has us drawing much thinner than overpairs.
      • There are 13 combos of hands that crush us and 9 combos of hands that flip. If we match those, we will have 9 pairs with about 25 to 30% equity (not quite because of the sets) and 4 left over unpaired combos that crush us. We can guesstimate that we have about 30% equity before the 4 unpaired combos and after adding in the left over combos, our equity is going go down and be between 25 to 30%.
    3. Analyze unpaired combos
      • There are 5 combos of OESDs and against those hands, we are a 2:1 favorite or 66%.
      • Thinking a little deeper though, we have a Q and that removes one out from the JTs hands which moves us closer to a 3:1 favorite.
      • Our JTs can average out with the left over combos of 98s from the "crush us" category which will boost our between 25 to 30% lets say 27% a percent or two to about 28% or so.
      • That leaves us with the 3 unaccounted combos of 76s where we are about a 2:1 favorite. 76s makes up about: 3/(13 + 9 + 5) or 10.7% and we have 66% equity so 0.66 * 0.107 = 7%.
    4. Combine the analysis to estimate:
      • Our base is about 27% after pairing the combos of crush and flips.
      • Pairing JTs and 98s bumps our equity up a little.
      • Our decent favorites add about 7%.
      • The total is 27 + 1 + 7 = 35%.
      • This is very close to the 34% equity that pokerstove gives us. This method is by no means 100% accurate and at times, it will be incorrect but this is a good way to start calculating equities and as you get better with pokerstove and estimating equities, this will all become easier and second nature to you.
      • Finally when our decision at the table is so close, where our estimated equity calculation is close but not 100% in line with our mathematical requirements, it is best to rely on the gut reads/intuition to tip the scales on whether to call or fold, depending on how frustrated we perceive our opponent to be.

      Alternate Method
      1. Tally up combinations
        • Crush us: 13 combos w/ about 13% equity
        • Flipping: 9 combos w/ about 50% equity
        • Decent favorites: 6 combos w/ about 66% equity
      2. Multiply range by equity
        • (13/28) * 0.13 = 0.06
        • (9/28) * 0.5 = 0.161
        • (6/28) * 0.66 = 0.141
        • You can fudge these numbers to make it easier.
        • Ex. (13/28) is close to 50% so 0.5 * 0.13 = 0.075. (9/28) is close to 30% so 0.3 * 0.5 = 0.15. (6/28) is close to 20% so 0.2 * 0.66 = 0.132
      3. Add up the equities
        • 0.06 + 0.161 + 0.141 = 36%
        • Fudge: 0.075 + 0.15 + 0.132 = 36%

        Table Estimations Thought Process
        • WoT's thought process at the table when using his TUPAC method:
        • As soon as I get checked raised all in, I think: ok, I doubt he ever does this as a pure bluff and right there I am talking about his hand range. I figure him to do this with big draws, sets, pocket aces, and the two OESD, 67s and JTs. He never has AK here because he will 4-bet it 100% pre-flop. There are about 10 or so big draws he can have depending on how many Axcc combos and suited club broadways he is calling with pre-flop to a 3-bet. I know I have been hammering on him so I know he may call with a lot of those combinations. So the more he is calling with those combinations the more draw combos there will be. As for sets, it's hard for him to have KK because I have one and the board has one. He could have AA and the smaller sets. He has aces and smaller sets about as often as he has big draws so if that's true, my equity against that part of his range is between 25-30%. he could have the 98s for two pair and the OESD as well. I am doing bad against the two pair but I am ahead of his OESD so that sort of balances out and gives me some equity. I am actually a 3:1 favorite against JTs since I have one of the Qs removing an out for him, giving me more equity. Over all I'd estimate that I am about 30 to 35% or about a 2:1 dog. I am getting about 2:1 on my money which makes it close. This leaves it down to reading my opponent. The more frustrated he is, the more apt I am to call. The less frustrated he is, the more I will fold.

        Saturday, August 28, 2010

        Suited Connectors, Implied Odds, and You (Theory/Math) by goofyballer


        maddog2030:
        I wrote a program to do these exact calcs over a year ago, and came out to the same ~25% you did, so your math is probably good (some of my numbers were also verified by BruceZ of the probability forum).

        Because I wrote a program to do this, I ran through a number of other types of hands that you may be interested in. For instance, suited one-gappers come out to 23% to flop OESD+,2pair+. So basically, most anytime you're willing to play a suited connector, you should be willing to play a similar suited one-gapper also.

        Summary of hands:
        Suited connectors: 25%
        Suited one-gappers: 23%
        Suited two-gappers: 18%
        Unsuited connectors: 17%
        Suited aces: 17%

        Note: This isn't the end-all, be-all, as it doesn't take into account draws to the nuts, etc. But it's a relatively decent gauge on the strength of those various types of hands to each other. 

        binions:
        I looked at this long ago. My posts are somewhere in the archives.

        If you factor out flopping flush draws on paired boards, and straight draws on paired and 3-flush boards against, I get 23.5% or 3.25:1 against flopping 2 pair or better made hands or at least an 8 out draw for suited max stretch 0-gap connectors.

        21.3% or 3.7:1 against for max stretch 1-gap suited connecttors.

        18.5% or 4.4:1 against for max stretch 2-gap suited connectors. 

        Jamougha:
        you will also flop some useful top pair hands with a AXs; less so with the smaller connectors.

        If you have e.g. A8s against a reasonably tight CO opening range (say 22+, 2 broadway, A8o+, Axs, 65s+, 86s+, two suited cards 9 or higher) then you are 49:51 with their range and typically getting good pot odds + position. With 65s you would be 37:63 and it would be rather harder to judge where you are. Your call is justified more by pot odds than implied odds. 

        c strong:
        "A good play may be to call with these in position only".

        This is really important. With a small PP you don't mind being OOP so much, as you'll usually be playing fit or fold on the flop. Extraction is easier in position, but you should still be able to get stacks in with a set OOP against an overpair, TPTK etc.

        With an SC, though, most of the time when you're continuing on the flop you'll have a draw. These are so much easier to play in position, where you have the option of checking behind for a free card if it's checked to you, betting or raising the flop to disguise your hand and possibly take a free card on the turn, etc. Much harder to play them OOP where none of your options are great: check/call looks like a draw, you have no FE and may not get odds to draw; leading may mean you get raised or floated so you can't take a free card; check-raising may mean you put a lot of money in to draw etc. 

        linuxrocks:
        One thing I wanted to add is the equity of playing the SCs after flopping a pair. The odds of flopping a pair that's not paired is 30%. The equity obviously is not that great, but there's certainly some value in backdoor draws, hitting your second pair, trips etc. This is especially true when we are the one raising or re-raising. This, I think is the big difference compared to small pairs. They are almost useless, if you don't hit the set.

        Calculating equity of a middle pair against a pre-flop raiser is quite a hairy calculation, but I am guessing it's much higher than the equity when small pairs don't hit set on the flop.