股票外汇技术分析英文版(16)

[e-mail]minar17Feb20072ENGINEERSAREASASANYONE222)(21xe3FibanacciRatios4Patterns•Thousandsofpatternshavebeencatalogued–DoubleBottom,Head&Shoulder,Flags,Pennants,etc.–Allareanecdotalorwithintheprobabilityofchance•TuneyourTVtoanunusedchannelandstareatthescreenintently–Iguaranteeyouwillseepatternsformedoutofpurenoise•Ifseeingisbelieving,checkout–Veryinterestingopticalillusions5WaveSynthesis•Sinewavesaretheprimitivestosynthesizemorecomplexwaveswave=SIN(F*T)-SIN(2*F*T)/2+SIN(3*F*T)/3CombinedWaveform:ElliottWave?•Whynotjustdealwithmeasurableprimitives?6MomentumFunctions1.MomentumcanNEVERleadthefunction2.Momentumisalwaysmoredisjoint(noisy)T=0CONCLUSIONS:RAMPSTEPIMPULSEJERKFUNCTION1stderivative(Momentum)2ndderivative(Acceleration)3rdderivative7MovingAverages1.MovingAveragessmooththefunction2.MovingAveragesLagbythecenterofgravityoftheobservationwindow3.UsingMovingAveragesisalwaysatradeoffbetweensmoothingandlagMovingAverageLagc.g.WindowCONCLUSIONS:8RelatingLagtotheEMAConstant•AnEMAiscalculatedas:g(z)=a*f(z)+(1-a)*g(z-1)whereg()istheoutputf()istheinputzistheincrementingvariable•Assumethefollowingforatrendmode–f()incrementsby1foreachstepofz•hasavalueof“i”onthe“ith”day–kistheoutputlagi-k=a*i+(1-a)*(i-k-1)=a*i+(i-k)-1-a*i+a*(k+1)0=a*(k+1)-1Thenk=1/a-1ORa=1/(k+1)9RelationshipofLagandEMAConstantak(Lag).51.41.5.32.33.253.24.19.0519•Smallacannotbeusedforshorttermanalysisduetoexcessivelag10ConceptofPredictiveFilters•Inthetrendmodepricedifferenceisdirectlyrelatedtotimelag•Proceduretogenerateapredictiveline:–TakeanEMAofprice–Takethedifference(delta)betweenthepriceanditsEMA–Formthepredictorbyaddingdeltatotheprice•equivalenttoadding2*deltatoEMA11SimplePredictiveTradingSystem•Rules:–BuywhenPredictorcrossesEMAfrombottomtotop–SellwhenPredictorcrossesEMAfromtoptobottom•Usuallyproducestoomanywhipsawstobepractical•CrossoverALWAYShappensaftertheturningpoint12Drunkard’sWalk•Positionastherandomvariable•ResultsinDiffusionEquation•Momentumastherandomvariable•ResultsinTelegrapher’sEquation22xPDtP22221xPCtPTtP13EfficientMarket•Meanderingriverisareal-worldexampleoftheDrunkard’swalk–Randomoveralongstretch–Coherentinashortstretch•HurstExponentconvergesto0.5overseveraldifferentspans–HoweverIusedittocreateanadaptivemovingaveragebasedonfractalsoverashortspan(FRAMA)14CoherentBehaviorExampleF=maF=-kxTherefore:ma=-kxdx/dt=vdv/dt=aTherefore:a=d2x/dt2And:m*d2x/dt2=-kxAssume:x=Sin(wt)Then:dx/dt=w*Cos(wt)d2x/dt2=-w2*Sin(wt)Assumptionistrueif:w2=k/mCONCLUSION:Onecancreatealeadingfunctionbytakingaderivativewhenthemarketiscoherent(inacyclemode).i.e.Cosine(x)leadsSine(x)15ManyIndicatorsAssumeaNormalProbabilityDistribution•Example-CCI–byDonaldLambertinOct1980FuturesMagazine•CCI=(PeakDeviation)/(.015*MeanDeviation)•Why.015?–Because1/.015=66.7–66.7%is(approximately)onestandarddeviation•IFTHEPROBABILITYDENSITYFUNCTIONISNORMAL16WhatareProbabilityDensityFunctions?ASquareWaveonlyhastwovaluesASquareWaveisuntradeablewithconventionalIndicatorsbecausetheswitchtotheothervaluehasoccurredbeforeactioncanbetakenAPDFcanbecreatedbymakingthewaveformwithbeadsonparallelhorizontalwires.Then,turntheframesidewaystoseehowthebeadsstackup.ASinewavePDFisnotmuchdifferentfromaSquarewavePDF17RealProbabilitiesareNOTGaussianProbabilityDistributionofa10BarChannelOver15yearsofTreasuryBonddataProbabilityDistributionofa30BarChannelOver15yearsofTreasuryBonddata18APhasorDescribesaCycle•CycleAmplitude(PythagoreanTheorem)Amplitude2=(InPhase)2+(Quadrature)2•PhaseAngle=ArcTan(Quadrature/InPhase)•CyclePeriodwhenSPhaseAngles=3600InPhaseQuadratureqPhaseAngleCyclePhasor19SinewaveIndicatorAdvantages•Linecrossingsgiveadvancewarningofcyclicturningpoints•Advancingphasedoesnotincreasenoise•Indicatorcanbe“tweaked”usingtheoreticalwaveforms•Nofalsewhipsawswhenthemarketisinatrendmode20CycleMeasurementTechniquesConvertAmplitudetoColorsospectrumcanbeplottedinsyncwithpricesMESA8SpectralEstimate(standardagainstwhichothertechniqueswillbemeasured)21FFT•Constraints:–Dataisarepresentativesampleofaninfinitelylongwave–Datamustbestationaryoverthesampletimespan–Musthaveanintegernumberofcyclesinthetimespan•Assumea64daytimespan–Longestcycleperiodis64days–Nextlongestis64/2=32days–Nextlongestis64/3=21.3days–Nextlongestis64/4=16days•Resultispoorresolution-gapsbetweenmeasuredcycles22FFT(continued)Paradox:–Theonlywaytoincreaseresolutionistoincreasethedatalength–Increaseddatalengthmakesrealizationofthestationarityconstrainthighlyunlikely•256datapointsarerequiredtorealizea1barresolutionfora16barcycle(rightwherewewanttowork)Conclusion:FFTmeasurementsarenotsuitableformarketanalysis23SlidingDFT•RequiresspacingofspectrallinesjustlikeaFFT•ThereforetheresolutionofaSlidingDFTistoopoortobeusedfortrading24FrequencyDiscriminators•Idescribed3differentdiscriminatorsin“RocketScienceforTraders”•Measurephasedifferencesbetweensuccessivesamples–ForexampleDq=36degreesdescribesa10barcycleperiod–Dis