{
  "id": "martingale",
  "title": "Martingale",
  "lang": "en",
  "source_lang": "en",
  "summary": "A martingale is an integrable stochastic process whose conditional expected future value equals its present value, relative to a specified information flow.",
  "category": "mathematics",
  "subcategory": "probability-statistics",
  "type": "concept",
  "aliases": [
    "Martingale process"
  ],
  "generated_at": "2026-10-04",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/martingale/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/martingale/",
      "json": "https://aiwiki.page/zh/martingale.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "A martingale is a stochastic process for which the conditional expected value at a future time, given the information currently available, equals its current value. It formalizes the idea of a fair game: knowing the past provides no expected gain or loss in the process itself. The definition concerns conditional averages, not individual trajectories, which may fluctuate substantially. Martingales are defined relative to both a probability measure and a specified flow of information. (math.dartmouth.edu)"
    },
    {
      "heading": "Mathematical definition",
      "text": "Let \\((\\Omega,\\mathcal F,\\mathbb P)\\) be a probability space, and let \\((\\mathcal F_n)_{n\\geq0}\\) be a filtration: an increasing sequence of sigma-algebras representing the information available at successive times. A sequence of real-valued random variables \\((M_n)_{n\\geq0}\\) is a martingale with respect to this filtration if:\n\n- Adaptedness: \\(M_n\\) is \\(\\mathcal F_n\\)-measurable, so its current value is determined by current information.\n- Integrability: \\(\\mathbb E[|M_n|]<\\infty\\) for every \\(n\\).\n- Martingale property:\n\n\\[\\mathbb E[M_{n+1}\\mid\\mathcal F_n]=M_n \\quad\\text{almost surely}.\\]\n\nHere \\(\\mathbb E[\\cdot\\mid\\mathcal F_n]\\) denotes conditional expectation, and “almost surely” permits exceptions on an event of probability zero. (math.cmu.edu)\n\nThe tower property of conditional expectation extends the one-step condition to every pair \\(m\\geq n\\):\n\n\\[\\mathbb E[M_m\\mid\\mathcal F_n]=M_n.\\]\n\nConsequently, the expected value is constant:\n\n\\[\\mathbb E[M_n]=\\mathbb E[M_0].\\]\n\nThe converse is false: constant unconditional expectation does not ensure the conditional martingale property. (tropp.caltech.edu)\n\nIn continuous time, an adapted, integrable process \\((M_t)_{t\\geq0}\\) is a martingale when\n\n\\[\\mathbb E[M_t\\mid\\mathcal F_s]=M_s \\qquad(0\\leq s\\leq t).\\]\n\nMany continuous-time results additionally assume a right-continuous, complete filtration and paths that are right-continuous with left limits. These regularity assumptions are distinct from the conditional-expectation definition. (web.stanford.edu)"
    },
    {
      "heading": "Interpretation and examples",
      "text": "Centered sums\n\nSuppose \\(X_1,X_2,\\ldots\\) are integrable, mutually independent random variables with \\(\\mathbb E[X_k]=0\\). Then\n\n\\[M_n=\\sum_{k=1}^{n}X_k,\\qquad M_0=0,\\]\n\nis a martingale for \\(\\mathcal F_n=\\sigma(X_1,\\ldots,X_n)\\). Independence makes the conditional mean of the next increment zero. A symmetric random walk, whose increments are \\(+1\\) and \\(-1\\) with equal probabilities, is a basic example. (n.ethz.ch)\n\nMore generally, independent increments are unnecessary. It is enough that the increments \\(D_n=M_n-M_{n-1}\\) satisfy\n\n\\[\\mathbb E[D_n\\mid\\mathcal F_{n-1}]=0.\\]\n\nSuch increments form a martingale difference sequence. (math.cmu.edu)\n\nSuccessive predictions of one quantity\n\nFor an integrable random variable \\(Y\\), define\n\n\\[M_n=\\mathbb E[Y\\mid\\mathcal F_n].\\]\n\nThis is a Doob martingale, also called a Lévy–Doob martingale. As information accumulates, the prediction of \\(Y\\) changes, but its next revision has conditional mean zero. This construction is useful when a complicated random object is revealed one component at a time. (tropp.caltech.edu)\n\nBrownian examples\n\nStandard Brownian motion \\((B_t)\\) is a martingale with respect to its natural filtration. So are\n\n\\[B_t^2-t \\quad\\text{and}\\quad \\exp\\!\\left(\\theta B_t-\\frac{\\theta^2t}{2}\\right),\\]\n\nfor each fixed real \\(\\theta\\). The compensating terms remove the predictable increase in the square or exponential. These examples illustrate that nonlinear transformations generally need correction terms to preserve the martingale property. (math.uchicago.edu)"
    },
    {
      "heading": "Related classes and transformations",
      "text": "A submartingale replaces equality by\n\n\\[\\mathbb E[X_{n+1}\\mid\\mathcal F_n]\\geq X_n;\\]\n\na supermartingale uses the reverse inequality. Both retain adaptedness and integrability. Their values therefore have, respectively, nonnegative or nonpositive conditional drift; their sample paths need not be monotone. A process is a martingale precisely when it belongs to both classes. (math.cmu.edu)\n\nIf \\(\\varphi\\) is a convex function and \\(\\varphi(M_n)\\) is integrable, conditional Jensen’s inequality shows that \\(\\varphi(M_n)\\) is a submartingale. Examples include \\(|M_n|\\) and, for square-integrable martingales, \\(M_n^2\\). (tropp.caltech.edu)\n\nIn discrete time, a submartingale \\(X_n\\) has the Doob decomposition\n\n\\[X_n=X_0+N_n+A_n,\\]\n\nwhere \\(N_0=A_0=0\\), \\(N_n\\) is a martingale, and \\(A_n\\) is predictable and increasing. Its increments are\n\n\\[A_n-A_{n-1} =\\mathbb E[X_n-X_{n-1}\\mid\\mathcal F_{n-1}].\\]\n\nThis separates unpredictable fluctuations from accumulated conditional drift. (web.stanford.edu)\n\nA martingale transform takes the form\n\n\\[G_n=\\sum_{k=1}^{n}H_k(M_k-M_{k-1}),\\]\n\nwhere \\(H_k\\) is determined by \\(\\mathcal F_{k-1}\\), before the next increment is observed. If the coefficients are bounded, \\(G_n\\) is a martingale. In the gambling interpretation, adjusting bounded stakes using past observations does not create a positive expected gain from a fair game. (n.ethz.ch)"
    },
    {
      "heading": "Stopping times and optional stopping",
      "text": "A stopping time \\(T\\), taking values in the nonnegative integers or infinity, satisfies\n\n\\[\\{T\\leq n\\}\\in\\mathcal F_n.\\]\n\nThus, whether stopping has occurred by time \\(n\\) can be decided without future information. The first time a process reaches a prescribed level is a typical example. (math.dartmouth.edu)\n\nThe optional stopping theorem gives conditions under which\n\n\\[\\mathbb E[M_T]=\\mathbb E[M_0].\\]\n\nFor a discrete-time martingale, any one of the following is sufficient:\n\n- \\(T\\) is bounded by a deterministic integer.\n- \\(T\\) is almost surely finite and \\(|M_n|\\leq C\\) for all \\(n\\), for a deterministic constant \\(C\\).\n- \\(\\mathbb E[T]<\\infty\\) and \\(|M_n-M_{n-1}|\\leq C\\) for all \\(n\\).\n\nThese are alternative sufficient conditions, not interchangeable descriptions of the same assumption. (math.dartmouth.edu)\n\nAlmost-sure finiteness alone is insufficient. A symmetric random walk starting at zero reaches \\(+1\\) almost surely, so stopping at its first visit gives \\(M_T=1\\), although \\(\\mathbb E[M_0]=0\\). The stopping time has infinite expectation, preventing the bounded-increment version from applying. Similar failures explain why the gambling strategy of doubling a stake after every loss does not contradict optional stopping: it allows unbounded stakes and potentially enormous intermediate losses. (n.ethz.ch)"
    },
    {
      "heading": "Maximal inequalities and concentration",
      "text": "Martingale inequalities control an entire trajectory, rather than only its value at one fixed time. For a nonnegative submartingale \\(X_n\\), Doob’s maximal inequality states\n\n\\[\\mathbb P\\!\\left(\\max_{0\\leq k\\leq n}X_k\\geq a\\right) \\leq\\frac{\\mathbb E[X_n]}{a}, \\qquad a>0.\\]\n\nFor a martingale with \\(\\mathbb E|M_n|^p<\\infty\\), \\(p>1\\), the \\(L^p\\) version gives\n\n\\[\\mathbb E\\!\\left[\\max_{0\\leq k\\leq n}|M_k|^p\\right] \\leq \\left(\\frac{p}{p-1}\\right)^p\\mathbb E|M_n|^p.\\]\n\nThese estimates connect terminal moments with the largest earlier fluctuation. (math.cmu.edu)\n\nThe Azuma–Hoeffding inequality provides exponential concentration. If\n\n\\[|M_k-M_{k-1}|\\leq c_k\\]\n\nalmost surely for deterministic constants \\(c_k\\), then, for \\(a>0\\),\n\n\\[\\mathbb P(|M_n-M_0|\\geq a) \\leq 2\\exp\\!\\left(-\\frac{a^2}{2\\sum_{k=1}^{n}c_k^2}\\right).\\]\n\nUnlike bounds restricted to sums of independent variables, this permits dependence compatible with the martingale property. Exposure martingales apply it to random graphs and randomized algorithms. (cl.cam.ac.uk)"
    },
    {
      "heading": "Convergence and integrability",
      "text": "Martingales need not converge without further assumptions. Doob’s martingale convergence theorem states that a discrete-time martingale satisfying\n\n\\[\\sup_n\\mathbb E|M_n|<\\infty\\]\n\nconverges almost surely to a finite, integrable limit \\(M_\\infty\\). In particular, every nonnegative martingale has such a limit, since its expectations are constant. Almost-sure convergence alone does not ensure preservation of expectation. (web.stanford.edu)\n\nThe stronger condition of uniform integrability,\n\n\\[\\lim_{K\\to\\infty}\\sup_n \\mathbb E\\!\\left[|M_n|\\mathbf1_{\\{|M_n|>K\\}}\\right]=0,\\]\n\nalso gives convergence in \\(L^1\\):\n\n\\[\\mathbb E|M_n-M_\\infty|\\longrightarrow0.\\]\n\nIt follows that\n\n\\[M_n=\\mathbb E[M_\\infty\\mid\\mathcal F_n], \\qquad \\mathbb E[M_\\infty]=\\mathbb E[M_0].\\]\n\nThus uniformly integrable martingales are precisely martingales representable as successive conditional expectations of an integrable terminal variable. (tropp.caltech.edu)\n\nThe distinction is essential: a martingale bounded in \\(L^1\\) may converge almost surely while losing expectation in the limit because rare, increasingly large values continue to contribute to its finite-time means. Bounds\n\n\\[\\sup_n\\mathbb E|M_n|^p<\\infty,\\qquad p>1,\\]\n\ninstead yield both almost-sure and \\(L^p\\) convergence. (web.stanford.edu)"
    },
    {
      "heading": "Continuous-time theory and applications",
      "text": "A local martingale behaves as a martingale after stopping at each member of a sequence of stopping times increasing almost surely to infinity. A local martingale need not be a true martingale: localization does not by itself supply the global integrability needed to preserve conditional expectations. Nonnegative local martingales are supermartingales. (math.uchicago.edu)\n\nMartingales underlie stochastic calculus. For a predictable integrand \\(H\\), the Brownian stochastic integral\n\n\\[\\int_0^t H_s\\,dB_s\\]\n\nis a square-integrable martingale when\n\n\\[\\mathbb E\\int_0^t H_s^2\\,ds<\\infty\\]\n\non every finite horizon. Under weaker local assumptions it may be only a local martingale. (math.uchicago.edu)\n\nIn statistics and probability, martingale methods handle sequential information and dependent observations; maximal and concentration inequalities quantify their fluctuations. In mathematical finance, discounted asset prices are modeled as martingales or local martingales under an appropriate equivalent pricing measure, subject to the assumptions of the model. This is not a claim that actual prices have zero expected return under the real-world probability measure. (cl.cam.ac.uk)"
    },
    {
      "heading": "Historical development",
      "text": "Martingale theory grew from attempts to formalize fair games and the limitations of gambling systems. Earlier work by Sergei Bernstein, Paul Lévy, and Andrey Kolmogorov anticipated aspects of the theory. Jean Ville’s 1939 book used martingale methods and a maximal inequality in studying randomness. Joseph L. Doob subsequently formulated the measure-theoretic framework, developed stopping and convergence results, and established martingales as a central class of stochastic processes in his 1953 book Stochastic Processes. In a later interview, Doob described Ville’s work as an important stimulus while explicitly acknowledging the earlier contributions. (chance.dartmouth.edu)"
    }
  ],
  "links": [
    "stochastic-process",
    "probability-space",
    "filtration-probability",
    "sigma-algebra",
    "random-variable",
    "conditional-expectation",
    "almost-surely",
    "expected-value",
    "statistical-independence",
    "random-walk",
    "brownian-motion",
    "convex-function",
    "jensens-inequality",
    "randomized-algorithm",
    "stochastic-calculus",
    "statistics",
    "andrey-kolmogorov"
  ],
  "sources": [
    {
      "title": "The Martingale Stopping Theorem",
      "url": "https://www.math.dartmouth.edu/~pw/math100w13/lalonde.pdf"
    },
    {
      "title": "Probability Theory & Stochastic Processes / Caltech CMS 117",
      "url": "https://tropp.caltech.edu/notes/Tro24-Probability-Theory-LN.pdf"
    },
    {
      "title": "3. The Convergence of Martingales",
      "url": "https://web.stanford.edu/class/stats310c/NOTES/lnotes.pdf"
    },
    {
      "title": "Lecture 7: Martingales and Concentration",
      "url": "https://www.cl.cam.ac.uk/teaching/1819/Probablty/materials/Lecture7.pdf"
    },
    {
      "title": "Stochastic Calculus and Applications",
      "url": "https://www.math.uchicago.edu/~lawler/finbook.pdf"
    },
    {
      "title": "A Conversation with Joe Doob",
      "url": "https://chance.dartmouth.edu/Doob/conversation.html"
    }
  ],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}