Bluff

With the poker

You play Poker. You come from miser your car and your house against a player having, think you, a hand better than yours, whereas you have only one pair of two.

It does not remain you any more but with bluffer.

You must make accept your adversary which you have a very good play whereas it is not the case. To make him believe, all the blows are allowed: small amused winks, false smiles…

A player who tries a “psychic” bluff must incite his adversaries to start again, not to come " for voir". He thus owes initially miser low, if he wants to encourage with the revival, then must increase his setting quickly, so that its adversaries are permanently persuaded that they have a definitely higher play (to start again) or clearly inferior (to lie down): if its setting is insufficient, it runs the risk which its adversary decides to go " for voir" and that its bluff crumbles. If a player never passes for a computer which bluffe, his revivals post only one strong play, which will dissuade its adversary to come to see. Conversely, if it frequently passes for whimsical susceptible to bluffer, its revival will not have a dissuasive effect on its adversaries, which will less hesitate to come to see. To play of the psychological bluffs, it is necessary to give an indication of very reasonable player, who starts again seldom and always with reason, which makes it possible to gain some blows of bluff without being worried.

The bluff on a weak hand is by no means an attempt at fraud, or a forced passage against the statistics, but simply an investment to be made, judiciously calculated to increase its profits on the whole of the part. The effect of the bluff is not on the weak hands (where the result is statistically indifferent), but of course the strong hands: So that a player can make profitable his strong hands, it is desirable that its adversaries follow its revivals, therefore to show that one bluffe regularly: this strategy of rational bluff will make it possible to make profitable an aggressive play.

The optimal strategy is discussed below.

Mathematics of the optimal bluff

Pair improved against improved pulling

It is supposed that Alice opened, and asked two charts (showing a pair a priori), and Bob followed and asked for a chart (showing a pulling a priori). Alice opens pot, probably showing a double pair and a brelan.

This being, when Bob carried out its pulling of fifth or color, it knows that it gains against Alice as long as she does not have Full or of Square, i.e. in 94.4% of the cases. Its play is practically with against twenty, and it can be allowed to start again, whether Alice has Brelan or a double pair. But on which level? All depends in fact on the frequency to which Bob bluffe in this case.

If Bob decides to start again, it posts a play " strongly gagnant" , which in theory cannot be a double pair. The play justifying a revival is a successful pulling, which led to a color or a fifth. This being, if it is imagined that it can bluffer, it can as well do it with a double pair of entry (it would not be whereas a semi-bluff). Thus the fact that it starts again basically does not change the possible nature of its hand, which remains: The revival of Bob posts a play which should arise only in B =14.6% of the cases, and Alice must choose between following (to sanction a possible bluff) or passing. If it has itself a gaining hand (Full or square, with a probability has =5.6%), it will certainly follow (or will start again). The following table describes the choices at the time when Bob decides bluffer:

Not neutral of the bluff

Alice can decide to sanction a possible bluff (with a frequency α), and Bob can have decided bluffer (with a frequency β).

If Bob has a weak hand, the choice is between bluffer or to pass. If it decides bluffer a little, the risk which it takes depends on the strategy of Alice: it risks an additional loss of the amount of its revival R (with a frequency of + α has), but can recover the pot P as long as Alice does not follow (the remainder of time). The point of balance for Bob is reached if R ( has + α) = P (1 ( has + α)), i.e. ( has +α) = p (P+R), and this point of balance depends only on the probability that Alice has to go to see. If Alice less frequently sanctions, the bluff of Bob can be more frequent, if Alice more frequently sanctions, the bluff of Bob must be less frequent, and if Alice exploits exactly this neutral point, the profit of Bob does not depend on its rate of bluff.
Vis-a-vis a revival of Bob, even if Alice has a potentially losing hand, it can nevertheless decide to sanction a possible bluff, in " suivant" the setting. It will lose its revival (R-r) in the B =14.6% of time where the pulling of Bob was carried out, but with a frequency of β% it recovers at the same time the pot and the revival of Bob. Balance is reached when (R-r). ( B ) =β. (P+R), which depends only on the rate of bluff of Bob (and of course, of its level of revival). If β< B . (R-r)/(P+R), to seek to sanction a bluff is a loss of money on average, therefore it is to better let bluffer without anything make. If on the contrary β is higher than this limit, the sanction of a possible bluff brings back on average money, therefore it is to better follow systematically.

The neutral point is reached when Alice comes to see the revivals with a probability has +α=P/(P+R), and that Bob bluff with a probability β= B . (R-r)/(P+R). If Bob preserves this rate of bluff exactly, it will gain on average as much, whatever the strategy of Alice: more on the bluff if Alice comes to less often see, more on successful pulling if Alice more often comes. In the same way, if Alice preserves this follow-up rate rigorously, it will gain as much, whatever the strategy of Bob.

It should be stressed that the optimal frequency of the bluff is always lower than that with the gaining hand alleged: when a rational bluffer posts a strong hand, it is present more once on two. Conversely, therefore, if a bluffer less shows his strong hand once on two when one comes to see it, its bluff is not rational but psychological, and the good strategy consists in coming to much more often see it.

Profit brought by the bluff

Which is the interest to exploit the neutral point? The profit can be calculated simply by supposing that Alice is on the neutral point, and Bob bluffe never (since it is enough that one of the two players there either so that the result is neutral):

Without bluff, when its pulling succeeds, Bob gains the pot with a frequency (1 - B ), and loses its revival with a frequency of B =14.6%. All in all, its profit without bluff is P (1 - B ) - R ( B ).
With bluff, it obliges Alice has to come to play at neutral point while coming to see with an additional frequency α, which saves to him now P (1 - B - α) - R ( B ) + (P+R) .α
the difference between the two situations is (P+R) α-P α = αR: on average, Bob gains the revivals exactly that Alice adds with her successful pullings.

It is seen that the effect of the bluff is not on the weak hands (where the result is statistically indifferent), but of course strong hands: The bluff on a weak hand is thus by no means an attempt at fraud, or a forced passage against the statistics, but simply an investment to be made, judiciously calculated to increase its profits on the whole of the part.

The exchange characteristic with a revival of Bob means then implicitly:

(Alice) Opening to the pot ( I have at least a strong pair ).
(Bob) Followed ( I have at least that ).
(Alice) Two charts ( it is a pair or a brelan ).
(Bob) a chart ( it is a pulling or a double pair ).
(Alice) Opening to the pot ( my pair improved ).
(Bob) I starts again twice pot ( I claim to have touched my pulling, but of course, I twice lie on five… with you to see ).
(Alice) Suivi ( you will laugh, but I had touched my full. ) or Followed ( I am not with a gaining hand, but I ensure my 33% of " for voir" to sanction a bluffer like you ) or Passes ( You have the right to gain this type of revival in 66% of the cases, I hope that you had play, you will not even be pleased to show it… ).

Reason to play at neutral point

For Bob, to play at neutral point presents a direct financial advantage: on average it will earn more money than without bluff, than it is on its strong hands developed by the follow-ups of Alice, or on the not held bluffs. The play on the neutral point has a psychological and statistical advantage: as the profitability of the blow does not depend any more psychological factors, it secures a regular play, without financial surprise. The only disadvantage is to compel itself with bluffer only within the rational limits, without letting itself guide by its inspiration. This being, it can continue to do it from time to time: it will be statistically undetectable.

For Alice, the play on the neutral point is not financially advantageous, because it is statistically a loss of money against a player who obviously bluffe never, or which bluffe with a frequency obviously in lower part of its neutral point. On the other hand, it is an insurance against the large bluffers or the erratic players: while exploiting the neutral point, it can play without having to guess what hiding place strategy of its adversary. The insurance has a cost, but on average, it is the same cost that it will gain when itself is in position of bluffer: on average, it is a strategy with null sum. This does not prevent it from coming to see less often those with the hands of which she thinks than the adversary cannot bluffer, of course, if its intuition is solid.

Against a player who obviously does not control the optimal rate of bluff or sanction, one should not play on the neutral point, but in a manner which makes profitable its defect of systematic play. If it bluffe too, to increase the follow-up, if it " colle" too much with the follow-up, to decrease the bluff, and so on. On average, a player who knows his neutral points earns money against a beginner who respects obviously never balance of it - it is enough to wait.
If the adversary obviously exploits the neutral point of the bluff, it does not have there reason to change its strategy: as long as it will not move of the neutral point, the average profit will be the same one. At most it is possible to try to deviate from the neutral point to see whether it follows, and to play the cat and mouse as soon as it does it.

Optimal level of revival

It is known that Alice can calculate her “neutral point” according to the rate of revival of Bob. The profit of Bob by a strategy of bluff is thus, by replacing α by its value:

Gain = R \ left (\ frac {P} {P+R} - has \ right) = P \ left (1 \ frac {P} {P+R} \ right) - a.R

It is seen that according to R (the level of revival compared to the pot), the profit follows a branch of hyperbole, and is maximum when its differential is cancelled:

\ frac {D (Profit)}{Dr.} = \ frac {P^ {2}} {(P+R) ^ {2}} - a=0 \ quad

i.e.

\ quad \ frac {P} {R+P} = \ sqrt {has}

In the case presented, the optimum would be a revival with three times the pot, because the probability for Alice of touching dangerous Full is not that of =5.6% has, therefore relatively weak. For a revival with three times the pot:

the maximum rate of acceptable bluff is about 60% that with the alleged hand: a strong bidding will be thus false in 37.5% of the cases.
the rate of checking to be ensured is not any more but of 1/4 (by counting the gaining hands).
the profit of the strategy is then 0.58 (instead of 0.54 for a revival of twice the pot).

In fact, it is not very important to exploit exactly the optimum, since around this value the average profit will not vary much. One can retain overall that if the probability that has Alice to gain are about 10%, the revivals of Bob are optimal for about twice the pot.

By taking revivals with twice pot (R=2P), Bob owes bluffer with a frequency equal to the two-thirds of its probability of having the strong hand which he claims to post ( B =14.6%), that is to say about 10%. When its pulling is losing (what arrives with relative frequency of 85.4%), long-term capacity to develop its gaining pullings, it owes all the same bluffer in 10%/85.4%= 11.7% with its losing hands, while starting again aggressively with twice the amount of the pot.

The cat and the mouse

The neutral point is stable, in the direction where if one of the two players is held to with it, its average profit does not depend on the strategy of the other player. But it is a strategy which has a cost: on average, it is necessary to come to see the revivals, typically once on three.

If Alice exploits her neutral point constantly, Bob can make provocation on a series of small blows, while not playing obviously more on his: to put itself at bluffer obviously too much, or on the contrary never. As soon as Bob deviates obviously from its neutral point, Alice can modify her rate of sanction consequently, and make pay in Bob her inconstancy: If the strategy of Bob is stable, Alice can benefit from it. But with this intention, it must itself move away from the neutral point, which exposes it to returns of claw on behalf of Bob… bob can try a psychological play: to make accept an eccentric strategy, and guess the moment when Alice will change her strategy, to reverse its behavior.

Double revivals

Another manner of seeing the bluff is that it reflects the principle: it is necessary to pay to see a strong hand. By applying this principle, vis-a-vis the revival of Bob, Alice must start again to make pay in Bob its Full - and thus, must also bluffer when it decides to denounce the bluff of Bob, so as to oblige it to follow and make profitable on average its Full.

The characteristic exchange is then:

(Alice) Opening to the pot ( I have at least a strong pair ) - the pot is worth 2.
(Bob) Suivi ( I have at least that ) - the pot is worth 3.
(Alice) Two charts ( it is a pair or a brelan ).
(Bob) a chart ( it is a pulling or a double pair ).
(Alice) Opening to the pot, therefore three ( I have at least a double strong pair ) - the pot is worth 6.
(Bob) I starts again with the double, therefore three plus nine ( I state to have made a success of my pulling, but I bluffe once on three… ) - the pot is worth 6*3=18, and Alice must add 9 to follow.
(Alice) I also starts again with the double of the pot (36), therefore 9 plus 25 ( Me it is a full, but I bluffe also once on three… ) - the pot is worth (18) *3=54, and Bob must add 25 to follow.

Arrived at this stage, Bob is always exposed to its three fundamental choices: to pass, follow, or start again?

Except particular case, a player who started again (to post a strong play) does not have objective reason to start again one second time. If it does it, it is for reasons which are attached to the psychological play.

Passes or followed (Alice plays a full rationally)

Alice having started again with the double of the pot, the neutral point is to come to see once on three, independently with the hand of Bob. In fact, if Alice plays in a rational way and makes a revival, it bluffe only once on three, and has a play gaining the remainder of time. To come to see a gaining hand with 66% is statistically expensive, but it is the price to be paid, because Alice bluffe correctly, and so that the bluff of Alice remains limited.

This being, Bob posted to have made a success of its pulling, and like it bluffe rationally, pulling is there twice on three. As long as to pass, to as much do it only on the hands which in any case were losing. Alice can bluffer with a brelan, and it would be idiotic to come to see it with a simple pair when one can do it on average with a continuation… If there is a on-revival of Alice, the strategy of Bob is thus:

If the initial revival of Bob were a bluff, to pass.
If the initial revival of Bob were justified by a color or a fifth, to come to see once on two the alleged full of Alice.

Starts again (posting of a square or a fifth flush)

A rational revival posts a play definitely more extremely than that already posted by the adversary.

Independently of the bluff, the first revival of Bob posts a pulling successful - in the most probable form, i.e. fifth or color. The revival of Alice posts that its pair improved in full or square, which beats the fifth or the color. An additional revival of Bob affirms that it even more extremely holds a play, which can only be one pulling successful with the fifth flush.

When one asks “for a chart”, the square (been useful) or the fifth flush arrives only in 0.25% of the distributions: on the hundred blows which one evening represents, the improvement of a pulling sees ten time, but a square been used or an improvement for the fifth flush is a hand which one sees only one evening out of four. And it is necessary that falls just when Alice does declare that she improves her hand with the full? Coincidence is extraordinary…

The reasoning cannot be statistical any more, but must be psychological, because such a play will certainly not arise twice to the table.

Bob can have shown pleasure to go up to the bluff, for the pleasure of the psychic tournament. It is enough to credit it with a probability of 0.5% of yielding to psychic in such a case so that its margin of bluff statistically authorized either exceeded, and that its bluff statistically justifies to be sanctioned. If Alice holds indeed a full or a square, it can come to see the alleged fifth flush of Bob on practically any dimension, with more than one hundred against one.
Bob can have shown a play without fault, worthy of a computer. Knowing that no system will enable him to try a bluff with the square only at one credible frequency of 0.12%, he prefers not to try a whole this type of bluff, well too random. If the fifth flush or the square arises, he will have fun cold sweats of his adversaries vis-a-vis his revivals, but as he will never play this scene twice in front of the same public, he gives up the idea to prepare a valorization of this type of hand by the sacrifice of preliminary bluffs. For Alice, blow, this play is sufficiently rare so that it chooses to limit breakage while never not following vis-a-vis a rational adversary: if it is the bad choice, it will never know it, because it will not on the occasion to remake similar in the evening.
the revival of Bob can simply post - for a reason or an other which he does not believe in the full of Alice. Bob perhaps felt a too strong hesitation of Alice on her revival, giving place to think that this revival is itself a bluff. If it is the case, Alice can decide to pass, which does not cost him anything (since its revival, not being justified by a strong hand, was an investment sacrificed in any case): she posts her bluff with the frequency statistically necessary to make profitable the pulling of a full another time, she shows that when she bluffe that sees herself,… mission accomplished, it is enough for him now to reproduce the same hesitations when she has indeed her full. She can also decide to start again Bob with which will crack the first: if the latter simply did not follow, it is that him also bluffe.

See too

bluff

External bond

plays of bluff on JeuxSoc

Random links: