ESTIMATING THE PROBABILITY THAT THE GAME OF MONOPOLY NEVER ENDS. Eric J. Friedman Shane G. Henderson Thomas Byuen Germán Gutiérrez Gallardo
|
|
- Tamsin Adele Hall
- 5 years ago
- Views:
Transcription
1 Proceedigs of the 2009 Witer Simulatio Coferece M. D. Rossetti, R. R. Hill, B. Johasso, A. Duki, ad R. G. Igalls, eds. ESTIMATING THE PROBABILITY THAT THE GAME OF MONOPOLY NEVER ENDS Eric J. Friedma Shae G. Hederso Thomas Byue Germá Gutiérrez Gallardo School of Operatios Research ad Iformatio Egieerig Corell Uiversity Ithaca NY 14853, USA ABSTRACT We estimate the probability that the game of Moopoly betwee two players playig very simple strategies ever eds. Four differet estimators, based respectively o straightforward simulatio, a Browia motio approximatio, asymptotics for Markov chais, ad importace samplig all yield a estimate of approximately twelve percet. 1 INTRODUCTION Have you ever played a game of Moopoly ad, after a few hours, started to woder if the game would ever ed? We foud that o-edig games were a distict possibility while coductig simulatios to try to determie effective strategies for the game. So we aturally became curious: What is the probability that the game of Moopoly ever eds? By ever edig we mea over a ifiite horizo, ad ot just that the game lasts loger tha reasoable players might be willig to play. It is ot easy to estimate this probability usig stadard simulatio, sice oe ca ever be certai that a game that is still goig will ot evetually ed. The mechaism by which games ca go o forever has bee uderstood at least sice Lehma ad Walker (1975). Essetially, the game reaches a stage where players cotiue to accumulate cash idefiitely. Although there is a chace, at ay poit i time, that ay player could fall o a ru of bad luck ad lose their cash, this probability is small eough that the player s wealths simply grow to ifiity. There is o shortage of cash i the game, sice the rules of Moopoly (Fritzlei 2009) state that if the bak ever rus out of moey, oe ca simply prit more o extra sheets of paper. The probability α that the game goes o forever depeds o the strategies that the players adopt. We detail our assumptios about the strategies the players adopt i Sectio 2. For ow, suffice it to say that we use very simple strategies that ivolve buyig properties wheever possible ad buildig houses or hotels wheever possible, while still maitaiig a modest threshold of cash o had to deal with cotigecies. I this paper we give four differet simulatio-based estimators of α, assumig two-player games. The estimators are built from a combiatio of a extremely detailed simulatio model coded i Java, ad a approximatig Markov chai model. Surprisigly (we thik) there have bee few serious academic studies of the game of Moopoly. Our survey revealed a few relevat books, icludig Lehma ad Walker (1975) i which two (the) Corell udergraduates studied strategy usig what were, at the time, state of the art simulatio techiques playig may moopoly games. They also performed some aalysis based o Markov chais. Somewhat more recetly, Darziskis (1987) simulated may moopoly games to derive good strategies. More recet work was spurred by Stewart s aalysis i Scietific America (Stewart 1996a, Stewart 1996b), which describes the use of Markov chais i the aalysis. This led to several aalyses o websites (although such aalyses have o doubt occurred sice the game was itroduced). For example, oe computes accurate site-visitig probabilities for Markov chais (Collis 2009) ad aother applies these to develop a fiace-ispired aalysis of good strategies (Darlig 2009). Lastly, for referece, the rules ad other facts about moopoly ca be foud o Fritzlei (2009).
2 While may have cursed the legth of moopoly games, we have ot foud ay detailed aalyses of the game s legth, or ay studies that compute the probability that the game goes o forever. But surely, millios of players have thought that the game might ever ed. Sectio 2 describes the simulatio model ad the player strategies we adopt. The, i Sectio 3 we describe a simple estimator of a upper boud o α. Sectio 4 explais i more detail how games ca go forever, describig wealth plots as a coceptual basis for uderstadig how this arises. It also explais how we ca view games as cosistig of potetially two stages. Oe ca the estimate the probability α that the game goes o forever by coditioig o first reachig Stage 2 ad the state of the board upo reachig Stage 2. Sectio 5 gives a approximatio for the coditioal probability based o a Browia approximatio of Stage-2 dyamics of player wealths. The coditioal probability is istead estimated usig asymptotics for Markov Chais usig a stadard Markov chai model of Stage 2 dyamics i Sectio 6. I Sectio 7 we implemet a importace samplig estimator usig the Markov chai asymptotics. Sectio 8 rouds out the paper with discussio of the results ad a outlie of future research efforts. 2 THE SIMULATION MODEL AND PLAYER STRATEGIES We have developed a very detailed simulatio model of the game of Moopoly i Java. The simulatio model is object based, ad ca accommodate more tha two players, although we restrict attetio to the two-player game i this paper. To verify the model we have performed a large umber of detailed game traces, ad compared the simulatio model s predictios with available results o, e.g., frequecies of ladig o particular squares, average et cash ifluxes from the bak per tur, ad so forth. We have also compared the model s results with those of a Markov chai model that is detailed i Sectio 6 below. The Markov chai model eeds to have a small state space to allow computatios to be performed quickly. Accordigly, throughout the remaider of this paper, both the simulatio ad the Markov chai operate uder the followig assumptio. I the real game of Moopoly, a player s tur ca ivolve rollig the dice up to three times, depedig o the umber of doubles rolled. I our versio of the game, each player oly rolls the dice oce, ad doubles are treated exactly the same as ay o-double roll. Players must pay ret ad so forth whe rollig doubles, ad this is i accordace with the official rules of the game. Each step of the Markov chai correspods to oe roud, i.e., oe roll of the dice for each player. We will measure time i rouds rather tha turs i the remaider of this paper. Uder this assumptio, the 3 doubles i a row leads to jail rule is igored. The impact of this assumptio is that players go to jail less ofte tha i the real game. However, the impact seems slight, sice players roll 3 doubles i a row oly oce every 216 turs, or oce every 258 rolls. Our players play a commo, simple strategy. That strategy ca be approximately characterized as follows. 1. Players maitai a reserve threshold of cash that is easily varied but we set to the maximum of $200 ad the maximum ret o a property owed by aother player. 2. Players buy every property they lad o that is for sale, uless doig so would reduce their cash below the reserve threshold. 3. Players ever bid o properties that are up for auctio. 4. Players build o streets as log as they have the cash o had to do so, buildig o the more expesive properties whe they have a choice. Buildig is always doe evely, so that for example, if 3 properties form a Moopoly, the either 3 houses are purchased or oe are purchased. 5. Whe a double is throw while i jail, the player comes out ad advaces their toke by the double that was rolled, ad this completes their tur. (They do ot roll agai.) 6. A player does ot pay to get out of jail, uless he/she has rolled o-doubles o three successive attempts. After payig to get out of jail, the last of the three rolls is take as the player s roll. 7. Whe a player receives a Get out of Jail Free card, the card is immediately sold back to the bak for $ Players do ot trade properties. 9. Whe a player receives a Advace toke to your choice of railroad card, the player moves to the railroad they are curretly o or, if ot curretly o a railroad, moves to the oe immediately precedig them. This will result i them receivig $200 for passig Go uless they are o the first Commuity Chest square after Go. 3 A SIMPLE ESTIMATOR Our first estimator of the probability α that a game goes o forever is very atural ad simple. Simulate a large umber of games, keepig track of the fractio of games that are still goig after rouds have bee completed, for various values of
3 Figure 1: Estimated fractio of games still goig after rouds have bee completed. The solid lie gives the estimate ad the dashed lies represet (poitwise) 95% cofidece itervals.. The look for a horizotal asymptote o the graph. We took exactly this approach with our simulatio model, simulatig 3100 games. Abbreviated results are give i Figure 1. The curve i Figure 1 appears to asymptote to a value somewhere ear 10%. A 95% cofidece iterval for the fractio of games still goig after = turs is 0.12 ± This represets a upper boud o α. Ufortuately, we have o way of kowig whether the curve would drop to a eve lower level were we to simulate for loger. To obtai a better sese of the probability that games go o forever we eed a differet approach. To derive such a approach, it is helpful to try to uderstad exactly how games ca go o forever. 4 HOW GAMES GO ON FOREVER To gai a coceptual uderstadig of how the game ca go o forever, let W = (W (1),W (2)) be the wealth of the two players after rouds, i.e., after each player has rolled the dice times. Here, we measure wealth as the cash o had plus the value of all other properties, houses, ad hotels if they were to be sold back to the bak. (These are worth oe half of the face value i this case.) If either player s wealth ever becomes egative the they are bakrupt ad the game fiishes. If we plot the sequece of poits W 0,W 1,W 2,... (see Figure 2) the we see that the game goes o forever as log as the plotted poits stay withi the oegative quadrat. At first, the game cofiguratio chages rapidly, with properties beig bought ad houses built. Call this period Stage 1. Some games are completed withi Stage 1, but others cotiue beyod this trasiet stage, reachig a poit where the board cofiguratio has essetially solidified, with o further property exchages or house buildig uless oe player or other approaches bakruptcy. Let us call this secod period Stage 2. For our simulatios, we defied Stage 2 as beig the
4 Figure 2: Coceptual plot of the wealth process. first time that all properties are purchased ad both players have at least $5000 i cash. For games that reach Stage 2, oe ca compute the drift of the player wealths β = (β(1),β(2)), where the term drift meas the average chage i player wealth over a large umber of turs. If either drift is egative, the oe is essetially certai that oe of the players will go bakrupt evetually ad the game will ed. But if both drifts are positive, the there is a chace that the wealth process will escape to ifiity before hittig either axis, ad the game goes o forever. Figure 3 gives some examples of such plots for simulated games. Stage 1 is idicated by a cluster of poits ear the origi ad the, after reachig Stage 2, the wealth of the two players accumulates ad we see rays headig up ad to the right. Cosider how to estimate 1 α, the probability that the game eds i fiite time. This probability is simply the probability that the game eds i Stage 1, plus the probability that the game eds i fiite time after reachig Stage 2. The first probability is easily estimated, sice there is a clear time at which to stop the simulatios the miimum of the times whe oe player goes bakrupt ad the time at which Stage 2 is reached. It is the secod probability that is challegig to estimate. Suppose we have reached Stage 2, so that the board cofiguratio is fixed, ad let us thik of this poit as time zero. Let A(i) be the evet that Player i goes bakrupt, i = 1,2, i.e., the wealth of Player i hits 0. Here we assume that each player cotiues to play over a ifiite horizo, eve if the opposig player goes bakrupt. Coceptually, play cotiues after such a poit, with paymets goig to ad from a phatom player. We are iterested i P(A(1) A(2)), the probability that oe of the players goes bakrupt ad the game eds. This probability seems complicated to estimate, sice it ivolves a hittig problem i two dimesios. Boole s iequality bouds this probability by P(A(1)) +P(A(2)), ad so we ow have two hittig problems each i oe dimesio, which seems much more tractable. We expect that these bakruptcy probabilities are very small for the cases of iterest, ad that oe will ted to domiate the other, so the overestimatio implied through the use of Boole s iequality should be slight. We have suppressed the depedece of these probabilities o the iitial wealth w = (w(1), w(2)) of the two players, ad the iitial layout of the board (player locatios, property owership, house ad hotel developmet etc.). We will later wat to vary these parameters, so P(A(1)) P(A(2)) i geeral.
5 Figure 3: Plots of the wealth process for two simulated games. I both cases we see the liear tred of the curve that correspods to the drift β. 5 BROWNIAN MOTION The ext questio we must address is how to estimate P(A(1)), the probability that Player 1 goes bakrupt, where we agai suppress the iitial coditios correspodig to the board cofiguratio ad iitial player wealths upo reachig Stage 2. (The calculatios for Player 2 are idetical.) Let w be Player 1 s iitial wealth, ad W be Player 1 s wealth after rouds, with W 0 = w. The player goes bakrupt if there exists a 0 such that W 0. Lookig at Figure 3, it is atural to approximate ( W : 0) by a Browia motio startig at w i oe dimesio. Such a process is characterized by a drift γ, ad a variace parameter η 2. If γ < 0, the the Browia motio drifts to, so is certai to hit 0, ad so P(A(1)) = 1. Similarly, if the drift γ = 0 the the Browia motio hits ay poit o the real lie i fiite time, ad so P(A(1)) = 1. Whe γ > 0, the Browia motio drifts to + so is ot certai to hit 0, but istead hits 0 with probability exp( 2γ/η 2 ). (This is a stadard calculatio that ca be foud, e.g., i Ross 1996, Sectio 8.4.) We thus estimate γ ad η 2 from the path of ( W : 0) ad the estimate P(A(1)) by mi{exp( 2γ/η 2 ),1}. The drift γ is easily estimated as ( W W 0 )/ for some large. To estimate η we first remove the drift to get Y i = W i W 0 iγ for all i, ad the use o-overlappig batch meas with batches of size b: We first compute the k batch meas M j, j = 1,...,k, where M j is the average of the observatios Y ( j 1)b+1,...,Y jb, j = 1,...,k (assumig kb = ). We ca the estimate η 2 by b multiplied by the sample variace of M 1,M 2,...,M k. I our code we took b = 60 rouds ad k = 170 for a total rulegth of = after reachig Stage 2. This estimator of secod-stage bakruptcy probabilities is compared with two others i Sectio 8, oce we defie those two other estimators. The above steps are completed for each game that reaches Stage 2 for each player. Let us ow reuse some coutig idices to describe the overall estimator that Moopoly goes o forever. Defie U k = 1 if the kth game reaches Stage 2, ad U k = 0 otherwise. Also, let V k be zero if the kth game does ot reach Stage 2, ad otherwise let it be the miimum of 1 ad the sum of the Browia estimates of the game edig as outlied above. Our overall estimator that the game goes o forever is the 1 1 U k +U k V k ] = k=1[1 1 k=1 U k (1 V k ). This estimator is a sample average of i.i.d. observatios, so ca be aalyzed usig stadard cofidece-iterval methodology. A 95% cofidece iterval for the probability that games will go o forever, based o = 3100 replicatios is 0.12 ± This is the same (to two decimal places) as the cofidece iterval obtaied usig the brute-force strategy of simulatig for a very log time horizo, thereby reiforcig that result. This estimator is appealig i the sese that it attempts to model the ifiite-time behaviour of the wealth process. However, it suffers from the fact that each replicatio that reaches Stage 2 requires further estimatio of the drift ad volatility parameters of the approximatig Browia motio, ad this is computatioally expesive. Furthermore, the estimated
6 parameters are subject to the usual oise i simulatio estimates ad are combied i a oliear fuctio, ad oe might expect that this could lead to errors i the overall estimate of the probability that the game goes o forever. We ow tur to aother approach to estimatig this probability that addresses, to some extet, each of these difficulties. 6 MARKOV CHAINS It has bee oted may times that the game of Moopoly ca be modeled, at various levels of detail, by a Markov chai. See, for example, Stewart (1996a), Stewart (1996b) where the dice rolls are modeled as a Markov chai, as is the board motio of a sigle player, resultig i the steady-state probabilities of ladig o various squares. I this sectio we preset a simple Markov model of a 2-player Moopoly game. Our purpose is to apply Markov chai theory to obtai two further estimates of the probability that the game goes o forever, preseted i this ad the followig sectio. We rely heavily o the theory preseted i Lehtoe ad Nyrhie (1992) for rui probabilities i Markov processes. We use a Markov chai X = (X : 0), where each state trasitio represets a roll of the dice by both players. Here X = (X (1),X (2)), where X (i) is the curret locatio of Player i o the board. As is stadard, X (i) ca take oe of 3 values for i jail, which distiguish how may turs the player has bee preset i jail. So there are 4 states for jail: just visitig, ad i jail for j turs so far, j = 0,1,2. Let d = 42 2 = 1764 be the umber of states i the chai. (There is o state for the Go to Jail square.) Let B(x,i) deote a geeric radom paymet paid to Player i from the bak durig oe trasitio of the Markov chai startig from the state x. This quatity could be egative if the player makes a paymet to the bak, e.g., whe assessed for street repairs. Similarly, let R(x,i) deote ay ret or other paymet received by Player i from the other player i oe trasitio of the Markov chai startig from the state x. This quatity could also be egative. The amout of moey received by Player i (i = 1,2) over m steps of the Markov chai is the where m 1 M (i) = C (i), =0 C (i) = B (X,i) + R (X,i). If Player i s iitial wealth is w(i), the the probability that Player i evetually goes bakrupt is P(A(i)) = P( 0 : w(i) + M (i) 0), i.e., Player i goes bakrupt if his et worth hits 0. Suppose we fix the layout of the board i the sese that all properties are owed ad o further buildig will occur. The coditioal distributio of ( C (1), C (2)), coditioal o (X : 0), the depeds oly o X ad X +1, ad so ((X, M ) : 0) is a Markov-additive process. We ca therefore apply the theory of rui probabilities of Markov-additive processes i our cotext. To apply that theory it is coveiet to work with C (i) = C (i) ad M (i) = M (i), ad defie the probability of bakruptcy as P(A(i)) = P( 0 : M (i) w(i)), i.e., the chace that et outgoigs will equal or exceed iitial wealth. This bakruptcy probability depeds o the log-ru drift of M (i) for each i. To compute this drift, first let c(x,i) be the coditioal mea of C (i) coditioal o X = x. Next, ote that the chai X is irreducible ad aperiodic o a fiite state space, ad is therefore positive recurret. Let π deote its statioary distributio. The strog law of large umbers for Markov chais, e.g., Mey ad Tweedie (1993), Theorem , the esures that M (i) π(x)c(x, i) = µ(i) x as with probability 1. We refer to µ(i) as the drift of Player i s outgoigs. It correspods to the egative of the previously computed drift β(i), which is the drift of the wealth process.
7 Figure 4: Scatter plots of the estimated drift for a sigle player. The solid lie is the y = x lie. The expected drift o the horizotal access is computed usig steady-state calculatios from the Markov chai, whereas the vertical axis values are simulatio results. The left (right) plot gives results before (after) debuggig, where a bug i the simulatio ad a differece i modelig assumptios betwee the simulatio ad the Markov chai model geerated a couple of outlier poits ad a skewed slope of the poits, respectively. We have used these calculatios to help verify our simulatio code. I particular, we ca compute drifts of games that reach Stage 2 usig both the Markov model as above, ad the simulatio code, ad compare those drifts. See Figure 4. If µ(i) > 0 the Player i will evetually ru out of moey with probability 1, i.e., P(A(i)) = 1, so that the game will certaily ed. (Note that this probability is, as metioed earlier, coditioal o the iitial coditios of the chai, which we specify below.) If µ(i) = 0 the agai P(A(i)) = 1, i.e., whe the drift is 0, Player i is agai certai to go bakrupt. This follows from the fuctioal cetral limit theorem Mey ad Tweedie (1993), Theorem If µ(i) < 0 the the situatio is more complicated. Here M(i) = (M (i) : 0) has egative drift, so there is a probability, depedig o the iitial wealth of the players ad the iitial state of the chai, that Player i will go bakrupt. We eed to compute this probability. We apply a result due to Lehtoe ad Nyrhie (1992) for Markov-additive processes o discrete Markov chais. Lehtoe ad Nyrhie specialized existig large-deviatios theory for Markov-additive processes to develop asymptotics ad a simulatio estimator for the rui probability for Markov chais. The simulatio estimator has provably good performace (asymptotic optimality i the laguage of rare-evet simulatio) as the probability of rui becomes smaller. We ow specialize their developmet to the Moopoly cotext. I order to do so, we eed to compute the root of a certai equatio ivolvig Laplace trasforms. Defie ˆp xy (θ,i) = E[e θc (i) I(X +1 = y) X = x] to be the Laplace trasform (LT) of the oe-step outgoigs of Player i o the evet that the chai trasitios from x to y. This quatity does ot deped o, ad here x ad y each deote a pair of locatios o the Moopoly board. This quatity is 0 if trasitios from x to y have zero probability. The trasform is defied for all θ (, ), because C is bouded. Let ˆP(θ,i) be the d d matrix of LTs defied above, which is a o-egative matrix. The theory of such matrices implies that it has a eigevalue λ(θ,i) such that 1. λ(θ,i) is simple, real ad strictly positive, ad λ(θ,i) > λ for ay other eigevalue λ, ad 2. λ(θ,i) has a strictly positive right eigevector h(θ,i). Defie r(θ,i) = logλ(θ,i) (here ad elsewhere we use the atural logarithm uless otherwise specified). It is kow that r(,i) is strictly covex i θ, that r(0,i) = 0, ad that r (0,i) = µ i < 0. So r(,i) iitially dips below 0, ad ca have at most oe positive root θ (i) (so that r(θ (i),i) = 0 ad r (θ (i),i) > 0). The it is kow that lim w(i) 1 w(i) logp(a(i)) = θ (i). (1)
8 The logarithmic asymptotic result (1) suggests the approximatio P(A(i)) e θ (i)w(i), (2) although there are may other possibilities for the right-had side of (2) that also satisfy (1). So how ca we compute this approximatio? We first have to idetify θ (i). This ivolves formig the matrix ˆP(θ,i) for various θ, computig its maximum eigevalue, ad the seekig the positive value θ (i) for which that eigevalue = 1, i.e., r(θ (i),i) = 0. This is a root-fidig problem that ca be efficietly solved usig a biary search. We implemeted this approach i Java, computig the eigevalues usig the power method with 200 iteratios, which came to withi 0.002% of the eigevalues as computed usig MATLAB. The remaiig questio is how to compute ˆP(θ,i) for a give θ. Recall that ˆP(θ,i) is a d d matrix, where d = For a give value of θ we eed to fill i the values of this matrix. This is algorithmically the most difficult part of the calculatio. Our approach higes o the fact that we ca compute the LT of the total outgoigs of Player 1 due to Players 1 ad 2 each takig a sigle tur by computig the LT of the total outgoigs of Player 1 due to Player 1 s tur, the the LT of the total outgoigs of Player 1 due to Player 2 s tur separately, ad the multiplyig these LTs. (Recall that the LT of a sum of idepedet radom variables is the product of the idividual LTs.) We do ot provide further details here. So our estimator based o the asymptotic approximatio (2) is as follows. Recall that U k is the idicator that the game reaches Stage 2. Defie Ṽ k to be the miimum of 1 ad the sum of the probabilities (2) for the two players. The estimator is the 1 1 U k +U k Ṽ k ] = k=1[1 1 k=1 U k (1 Ṽ k ). This estimator is a sample average of i.i.d. observatios, so ca be aalyzed usig stadard cofidece-iterval methodology. A 95% cofidece iterval for the probability that games will go o forever based o = 3100 replicatios is 0.12 ± This agrees with the previous estimators. This approach is appealig i that it captures the ifiite-time behaviour of the wealth process, ad i cotrast to the Browia-motio estimator is relatively efficietly computed, ad coditioal o the state of the board at the start of Stage 2 cotais o simulatio oise. It suffers from a importat disadvatage, however. The approximatio (2) is oe of may possible approximatios that is cosistet with the logarithmic asymptotic (1) ad aother cosistet approximatio might yield very differet results. We ow address that cocer usig importace samplig. 7 IMPORTANCE SAMPLING I this sectio we preset the key ideas behid a importace samplig scheme for estimatig the probability that the game goes o forever, agai relyig heavily o the theory i Lehtoe ad Nyrhie (1992). If oe is familiar with that theory the our approach is coceptually exactly as they suggest. The difficulty really lies i the algorithmic implemetatio, because the discrete distributios ivolved have may poits of support. The overall simulatio procedure is as follows. We start by simulatig a game util either the game eds or Stage 2 is reached, whichever comes first, exactly as before. This is doe uder the usual probability distributio associated with the simulatio. Upo reachig Stage 2 we compute the drift of the player wealths as described i the previous sectio, ad if the drift of either player s wealth is ot strictly positive the the coditioal probability that the game eds is estimated to be 1. Otherwise, we eed to compute the coditioal probability that the game will ed, coditioal o the iitial cofiguratio at the oset of Stage 2. At this poit we chage the trasitio probabilities of the Markov chai (ad the correspodig paymets) so as to esure that both players will go bakrupt evetually with probability 1. This biases the probability estimates, so to correct the estimates we have to multiply by the likelihood ratio. Let p xy (c,i) be the probability, uder the usual probabilities that gover Moopoly, that from state x we trasitio to state y, ad i the process player i icurs outgoigs c. The importace samplig scheme chages these probabilities upo reachig Stage 2 to q xy (c,i) say. If we let τ(i) be the time at which player i goes bakrupt, the coditioal o reachig Stage 2 ad the cofiguratio of the board ad so forth at that time, ad takig the time of reachig Stage 2 to be time zero, the the coditioal probability that Player i goes bakrupt is EI(τ(i) < ) = Ẽ [ I(τ(i) < ) τ(i) k=1 ] p Xk 1,X k (C k (i),i), (3) q Xk 1,X k (C k (i),i)
9 Figure 5: The expected drift of the wealth process uder the stadard ad importace-samplig distributios. The effect of importace samplig is approximately to reverse the sig of the drift. The poits are broke dow ito games where there are o hotels, where oe player plays agaist a player with 2 hotels, where oe player ows 2 hotels ad is playig someoe with o hotels, ad whe there are more tha 2 hotels o the board. where Ẽ deotes the fact that we simulate the chai uder the ew dyamics q rather tha p i Stage 2. The theory i Lehtoe ad Nyrhie suggests that i order to estimate P(A(i)), we should simulate the chai uder the particular choice of trasitio probabilities q xy (c,i) = ˆp xy(θ )h(y) e θ c p xy (c) λ(θ )h(x) ˆp xy (θ ), where we write θ for θ (i) ad so forth for brevity. We have used this form of the expressio to show that if oe first adds over all possible values of c, the secod fractio sums to 1, ad the addig over all possible y the first fractio sums to 1, so that this is, ideed, a set of trasitio probabilities. However, these trasitio probabilities ca also be writte q xy (c,i) = eθ c p xy (c)h(y), (4) h(x) sice λ(θ ) = 1. This secod expressio better reflects how we geerate trasitios usig the ew probabilities. With this particular choice of importace-samplig trasitio kerel, I(τ(i) < ) = 1 a.s., ad so (3) simplifies to [ ] Ẽ e θ τ(i) k=1 C k(i) h(x T ), (5) h(x τ(i) ) ad this is our estimator of the coditioal probability that Player i goes bakrupt. As with may other estimators of rui probabilities i the light-tailed settig, we have see that the effect of importace samplig is approximately to reverse the sig of the drift. Figure 5 shows this effect, where we plot the expected drift as computed by the Markov chai model uder the origial probability dyamics versus the expected drift uder importace samplig for a umber of games.
10 Figure 6: Scatter plots of the estimated coditioal probability of games that have reached Stage 2 edig i fiite time. The three plots give, from left to right, scatter plots of the Browia estimator versus the estimator based o the asymptotic approximatio (2), the Browia estimator versus the importace-samplig estimator, ad the asymptotic-approximatio estimator versus the importace-samplig estimator. The full estimator of the probability that the game goes o forever is the as follows. As before, let U k be the idicator that Stage 2 is reached o the kth simulatio replicatio. Upo reachig Stage 2, we switch from the full Moopoly simulatio uder the usual probabilities to simulatios of the Markov chai uder the importace-samplig trasitio probabilities (oe for each player), ad simulate each path util the players go bakrupt. We ca the compute the quatity iside the expectatio i (5) as a sigle-replicatio estimator of the coditioal expectatio of bakruptcy for each player i = 1,2. Next, defie ˆV k to be the miimum of 1 ad the sum of these two estimators of the coditioal expectatios i Stage 2. The importace samplig estimator of the probability that the game goes o forever is the 1 1 U k +U k ˆV k ] = k=1[1 1 k=1 U k (1 ˆV k ). This estimator is a sample average of i.i.d. observatios, so ca be aalyzed usig stadard cofidece-iterval methodology. A 95% cofidece iterval for the probability that games will go o forever based o = 3100 replicatios is 0.12 ± COMPARISON, DISCUSSION AND FUTURE RESEARCH All four of our estimators yield cofidece itervals that suggest that the probability that the game goes o forever is close to 12%. This was the coclusio we obtaied usig the simplest estimator of all, i Sectio 3. However, we could ot be certai, based o that estimator aloe, that the probability would ot actually be sigificatly lower tha this value. The other three estimators were motivated by this issue ad the fact that we had fu computig them. Usig those estimators we cofirmed that, i fact, the potetial for the game edig after Stage 2 is reached is very slight. The three estimators based o Stage 2 calculatios are all biased due to the use of Boole s iequality to boud the probability that either player goes bakrupt by the sum of those probabilities. Beyod that bias, the Browia motio estimator ad the asymptotic-approximatio estimator both possess additioal bias, but the importace-samplig estimator does ot. It is iterestig to cosider how these three estimators compare. We coducted a small experimet where we idetified approximately 100 games that reach Stage 2, ad that have coditioal probabilities that the game will evetually ed that are i the rage of 0.01 through We the computed each of the estimators, ad produced scatter plots of their values i Figure 6. There is close agreemet betwee the asymptotic-approximatio estimator ad the importace-samplig estimator, but the Browia estimator differs. This differece does ot exhibit itself i our fial estimators of the probability that the game goes o forever sice the probability of reachig a poit i Stage 2 where these differeces occur is so small that it caot sigificatly ifluece the results. I future work we might see how strategies ca ifluece the legth of the game. Our assumptio that players do ot trade severely limits the possibility of developig properties, which i tur limits the probability that the game will ed. Whe players have hotels, or may houses, the variace i the player wealth process is much higher tha whe there is oly
11 limited developmet, ad so the chace that the game will ed would likely be higher. I that settig we would expect that the three estimators that are based o Stage 2 calculatios would provide a beefit beyod the simple estimator, ad that the probability that the game goes o forever would be smaller tha twelve percet. But primarily we hope to use the tools we have developed to this poit to idetify highly effective player strategies. We caot hope to approach the complexity of the strategies outlied i Lehma ad Walker (1975), sice those strategies ivolve extremely complicated ad subtle trades that go well beyod what we evisage codig. However, we ca certaily ivestigate questios such as whether oe should build through to hotels, or to halt developmet earlier so as to soak up houses. A importat questio is how to use some kid of automated learig process to idetify the best of a family of strategies. The reaso this is importat is ot just that we are ethusiasts ad see it that way, but also because Moopoly is a microcosm of may importat systems i real life, where sequetial decisios uder ucertaity are made i the face of competitio from other etities. True ethusiasts of estimatig the probability that the game goes o forever might also implemet importace-samplig estimators that do ot rely o the use of Boole s iequality, so as to obtai completely ubiased estimators of the probability that the game goes o forever. However, our results suggest that such a effort would have to idetify the probability extremely accurately i order to provide additioal iformatio beyod the cofidece itervals produced i this paper. ACKNOWLEDGMENTS This work was partially supported by Natioal Sciece Foudatio Grat Numbers CMMI , ITR ad CDI We would like to thak curret ad former studets Alex Aidu, Raghu Chadrasekara, Sea Choi, Deis Li, ad Ady Tuchma for simulatio programmig. REFERENCES Collis, T Probabilities i the game of Moopoly. < moopoly.shtml> [accessed Aug 4, 2009]. Darlig, T How to wi at Moopoly a surefire strategy. < [accessed Aug 4, 2009]. Darziskis, K Wiig Moopoly: A complete guide to property accumulatio, cash flow strategy, ad egotiatig techiques whe playig the best-sellig board game. HarperCollis. Fritzlei Moopoly/official rules. < Rules> [accessed Aug 4, 2009]. Lehma, J., ad J. Walker ways to wi at Moopoly. Dell Books. Lehtoe, T., ad H. Nyrhie O asymptotically efficiet simulatio of rui probabilities i a Markovia eviromet. Scadiavia Actuarial Joural: Mey, S. P., ad R. L. Tweedie Markov chais ad stochastic stability. Lodo: Spriger-Verlag. Ross, S. M Stochastic processes. 2d ed. New York: Wiley. Stewart, I. 1996a. How fair is Moopoly? Scietific America 274: Stewart, I. 1996b. Moopoly revisited. Scietific America 275: AUTHOR BIOGRAPHIES ERIC J. FRIEDMAN is a associate professor i the School of Operatios Research ad Iformatio Egieerig at Corell Uiversity. His mai research iterests are i game theory, computer etworks ad other problems arisig o the Iteret. He also has a strog iterest i combiatorial games ad their coectios to dyamical systems theory ad physics. He prefers dogs to cats. His web page ca be foud via < SHANE G. HENDERSON is a professor i the School of Operatios Research ad Iformatio Egieerig at Corell Uiversity. He is the simulatio area editor at Operatios Research, ad a associate editor for the ACM Trasactios o Modelig ad Computer Simulatio ad Operatios Research Letters. He co-edited the hadbook Simulatio as part of Elsevier s series of Hadbooks i Operatios Research ad Maagemet Sciece, ad also co-edited the Proceedigs of the 2007 Witer Simulatio Coferece. He likes cats but is allergic to them. His research iterests iclude discrete-evet simulatio ad simulatio optimizatio, ad he has worked for some time with emergecy services. His web page ca be foud via <
12 THOMAS BYUEN is a udergraduate i the School of Operatios Research ad Iformatio Egieerig. He ca be reached at <tb287@corell.edu>. GERMÁN GUTIÉRREZ GALLARDO is a Master of Egieerig studet i the School of Operatios Research ad Iformatio Egieerig at Corell Uiversity. He ca be reached at <gg92@corell.edu>.
Roberto s Notes on Infinite Series Chapter 1: Series Section 2. Infinite series
Roberto s Notes o Ifiite Series Chapter : Series Sectio Ifiite series What you eed to ow already: What sequeces are. Basic termiology ad otatio for sequeces. What you ca lear here: What a ifiite series
More informationA SELECTIVE POINTER FORWARDING STRATEGY FOR LOCATION TRACKING IN PERSONAL COMMUNICATION SYSTEMS
A SELETIVE POINTE FOWADING STATEGY FO LOATION TAKING IN PESONAL OUNIATION SYSTES Seo G. hag ad hae Y. Lee Departmet of Idustrial Egieerig, KAIST 373-, Kusug-Dog, Taejo, Korea, 305-70 cylee@heuristic.kaist.ac.kr
More informationUnit 5: Estimating with Confidence
Uit 5: Estimatig with Cofidece Sectio 8.2 The Practice of Statistics, 4 th editio For AP* STARNES, YATES, MOORE Uit 5 Estimatig with Cofidece 8.1 8.2 8.3 Cofidece Itervals: The Basics Estimatig a Populatio
More informationlecture notes September 2, Sequential Choice
18.310 lecture otes September 2, 2013 Sequetial Choice Lecturer: Michel Goemas 1 A game Cosider the followig game. I have 100 blak cards. I write dow 100 differet umbers o the cards; I ca choose ay umbers
More informationLogarithms APPENDIX IV. 265 Appendix
APPENDIX IV Logarithms Sometimes, a umerical expressio may ivolve multiplicatio, divisio or ratioal powers of large umbers. For such calculatios, logarithms are very useful. They help us i makig difficult
More informationX-Bar and S-Squared Charts
STATGRAPHICS Rev. 7/4/009 X-Bar ad S-Squared Charts Summary The X-Bar ad S-Squared Charts procedure creates cotrol charts for a sigle umeric variable where the data have bee collected i subgroups. It creates
More informationCh 9 Sequences, Series, and Probability
Ch 9 Sequeces, Series, ad Probability Have you ever bee to a casio ad played blackjack? It is the oly game i the casio that you ca wi based o the Law of large umbers. I the early 1990s a group of math
More informationx y z HD(x, y) + HD(y, z) HD(x, z)
Massachusetts Istitute of Techology Departmet of Electrical Egieerig ad Computer Sciece 6.02 Solutios to Chapter 5 Updated: February 16, 2012 Please sed iformatio about errors or omissios to hari; questios
More informationDiscrete Mathematics and Probability Theory Spring 2014 Anant Sahai Note 12
EECS 70 Discrete Mathematics ad Probability Theory Sprig 204 Aat Sahai Note 2 Probability Examples Based o Coutig We will ow look at examples of radom experimets ad their correspodig sample spaces, alog
More informationAPPLICATION NOTE UNDERSTANDING EFFECTIVE BITS
APPLICATION NOTE AN95091 INTRODUCTION UNDERSTANDING EFFECTIVE BITS Toy Girard, Sigatec, Desig ad Applicatios Egieer Oe criteria ofte used to evaluate a Aalog to Digital Coverter (ADC) or data acquisitio
More informationCHAPTER 5 A NEAR-LOSSLESS RUN-LENGTH CODER
95 CHAPTER 5 A NEAR-LOSSLESS RUN-LENGTH CODER 5.1 GENERAL Ru-legth codig is a lossless image compressio techique, which produces modest compressio ratios. Oe way of icreasig the compressio ratio of a ru-legth
More informationPermutation Enumeration
RMT 2012 Power Roud Rubric February 18, 2012 Permutatio Eumeratio 1 (a List all permutatios of {1, 2, 3} (b Give a expressio for the umber of permutatios of {1, 2, 3,, } i terms of Compute the umber for
More information7. Counting Measure. Definitions and Basic Properties
Virtual Laboratories > 0. Foudatios > 1 2 3 4 5 6 7 8 9 7. Coutig Measure Defiitios ad Basic Properties Suppose that S is a fiite set. If A S the the cardiality of A is the umber of elemets i A, ad is
More information13 Legislative Bargaining
1 Legislative Bargaiig Oe of the most popular legislative models is a model due to Baro & Ferejoh (1989). The model has bee used i applicatios where the role of committees have bee studies, how the legislative
More information}, how many different strings of length n 1 exist? }, how many different strings of length n 2 exist that contain at least one a 1
1. [5] Give sets A ad B, each of cardiality 1, how may fuctios map A i a oe-tooe fashio oto B? 2. [5] a. Give the set of r symbols { a 1, a 2,..., a r }, how may differet strigs of legth 1 exist? [5]b.
More informationCS 201: Adversary arguments. This handout presents two lower bounds for selection problems using adversary arguments ëknu73,
CS 01 Schlag Jauary 6, 1999 Witer `99 CS 01: Adversary argumets This hadout presets two lower bouds for selectio problems usig adversary argumets ëku73, HS78, FG76ë. I these proofs a imagiary adversary
More informationFingerprint Classification Based on Directional Image Constructed Using Wavelet Transform Domains
7 Figerprit Classificatio Based o Directioal Image Costructed Usig Wavelet Trasform Domais Musa Mohd Mokji, Syed Abd. Rahma Syed Abu Bakar, Zuwairie Ibrahim 3 Departmet of Microelectroic ad Computer Egieerig
More informationH2 Mathematics Pure Mathematics Section A Comprehensive Checklist of Concepts and Skills by Mr Wee Wen Shih. Visit: wenshih.wordpress.
H2 Mathematics Pure Mathematics Sectio A Comprehesive Checklist of Cocepts ad Skills by Mr Wee We Shih Visit: weshih.wordpress.com Updated: Ja 2010 Syllabus topic 1: Fuctios ad graphs 1.1 Checklist o Fuctios
More informationThe Detection of Abrupt Changes in Fatigue Data by Using Cumulative Sum (CUSUM) Method
Proceedigs of the th WSEAS Iteratioal Coferece o APPLIED ad THEORETICAL MECHANICS (MECHANICS '8) The Detectio of Abrupt Chages i Fatigue Data by Usig Cumulative Sum (CUSUM) Method Z. M. NOPIAH, M.N.BAHARIN,
More informationSummary of Random Variable Concepts April 19, 2000
Summary of Radom Variable Cocepts April 9, 2000 his is a list of importat cocepts we have covered, rather tha a review that derives or explais them. he first ad primary viewpoit: A radom process is a idexed
More informationNovel pseudo random number generation using variant logic framework
Edith Cowa Uiversity Research Olie Iteratioal Cyber Resiliece coferece Cofereces, Symposia ad Campus Evets 011 Novel pseudo radom umber geeratio usig variat logic framework Jeffrey Zheg Yua Uiversity,
More information8. Combinatorial Structures
Virtual Laboratories > 0. Foudatios > 1 2 3 4 5 6 7 8 9 8. Combiatorial Structures The purpose of this sectio is to study several combiatorial structures that are of basic importace i probability. Permutatios
More informationAC : USING ELLIPTIC INTEGRALS AND FUNCTIONS TO STUDY LARGE-AMPLITUDE OSCILLATIONS OF A PENDULUM
AC 007-7: USING ELLIPTIC INTEGRALS AND FUNCTIONS TO STUDY LARGE-AMPLITUDE OSCILLATIONS OF A PENDULUM Josue Njock-Libii, Idiaa Uiversity-Purdue Uiversity-Fort Waye Josué Njock Libii is Associate Professor
More informationProblem of calculating time delay between pulse arrivals
America Joural of Egieerig Research (AJER) 5 America Joural of Egieerig Research (AJER) e-issn: 3-847 p-issn : 3-936 Volume-4, Issue-4, pp-3-4 www.ajer.org Research Paper Problem of calculatig time delay
More informationCompound Controller for DC Motor Servo System Based on Inner-Loop Extended State Observer
BULGARIAN ACADEMY OF SCIENCES CYBERNETICS AND INFORMATION TECHNOLOGIES Volume 6, No 5 Special Issue o Applicatio of Advaced Computig ad Simulatio i Iformatio Systems Sofia 06 Prit ISSN: 3-970; Olie ISSN:
More informationA New Space-Repetition Code Based on One Bit Feedback Compared to Alamouti Space-Time Code
Proceedigs of the 4th WSEAS It. Coferece o Electromagetics, Wireless ad Optical Commuicatios, Veice, Italy, November 0-, 006 107 A New Space-Repetitio Code Based o Oe Bit Feedback Compared to Alamouti
More informationJoin a Professional Association
Joi a Professioal Associatio 1. The secret job resource: professioal orgaizatios. You may ot kow this, but the career field you re i, or plaig to work i i the future, probably has at least oe professioal
More informationApplication of Improved Genetic Algorithm to Two-side Assembly Line Balancing
206 3 rd Iteratioal Coferece o Mechaical, Idustrial, ad Maufacturig Egieerig (MIME 206) ISBN: 978--60595-33-7 Applicatio of Improved Geetic Algorithm to Two-side Assembly Lie Balacig Ximi Zhag, Qia Wag,
More informationIntermediate Information Structures
Modified from Maria s lectures CPSC 335 Itermediate Iformatio Structures LECTURE 11 Compressio ad Huffma Codig Jo Roke Computer Sciece Uiversity of Calgary Caada Lecture Overview Codes ad Optimal Codes
More informationCombinatorics. Chapter Permutations. Reading questions. Counting Problems. Counting Technique: The Product Rule
Chapter 3 Combiatorics 3.1 Permutatios Readig questios 1. Defie what a permutatio is i your ow words. 2. What is a fixed poit i a permutatio? 3. What do we assume about mutual disjoitedess whe creatig
More informationDesign of FPGA- Based SPWM Single Phase Full-Bridge Inverter
Desig of FPGA- Based SPWM Sigle Phase Full-Bridge Iverter Afarulrazi Abu Bakar 1, *,Md Zarafi Ahmad 1 ad Farrah Salwai Abdullah 1 1 Faculty of Electrical ad Electroic Egieerig, UTHM *Email:afarul@uthm.edu.my
More informationRadar emitter recognition method based on AdaBoost and decision tree Tang Xiaojing1, a, Chen Weigao1 and Zhu Weigang1 1
Advaces i Egieerig Research, volume 8 d Iteratioal Coferece o Automatio, Mechaical Cotrol ad Computatioal Egieerig (AMCCE 7) Radar emitter recogitio method based o AdaBoost ad decisio tree Tag Xiaojig,
More informationMDM 4U MATH OF DATA MANAGEMENT FINAL EXAMINATION
Caadia Iteratioal Matriculatio rogramme Suway Uiversity College MDM 4U MTH OF DT MNGEMENT FINL EXMINTION Date: November 28 th, 2006 Time: 11.30a.m 1.30p.m Legth: 2 HOURS Lecturers: lease circle your teacher
More informationDepartment of Electrical and Computer Engineering, Cornell University. ECE 3150: Microelectronics. Spring Due on April 26, 2018 at 7:00 PM
Departmet of Electrical ad omputer Egieerig, orell Uiersity EE 350: Microelectroics Sprig 08 Homework 0 Due o April 6, 08 at 7:00 PM Suggested Readigs: a) Lecture otes Importat Notes: ) MAKE SURE THAT
More informationA study on the efficient compression algorithm of the voice/data integrated multiplexer
A study o the efficiet compressio algorithm of the voice/data itegrated multiplexer Gyou-Yo CHO' ad Dog-Ho CHO' * Dept. of Computer Egieerig. KyiigHee Uiv. Kiheugup Yogiku Kyuggido, KOREA 449-71 PHONE
More informationThe Institute of Chartered Accountants of Sri Lanka
The Istitute of Chartered Accoutats of Sri Laka Postgraduate Diploma i Busiess ad Fiace Quatitative Techiques for Busiess Hadout 02:Presetatio ad Aalysis of data Presetatio of Data The Stem ad Leaf Display
More informationUsing Color Histograms to Recognize People in Real Time Visual Surveillance
Usig Color Histograms to Recogize People i Real Time Visual Surveillace DANIEL WOJTASZEK, ROBERT LAGANIERE S.I.T.E. Uiversity of Ottawa, Ottawa, Otario CANADA daielw@site.uottawa.ca, lagaier@site.uottawa.ca
More informationCounting on r-fibonacci Numbers
Claremot Colleges Scholarship @ Claremot All HMC Faculty Publicatios ad Research HMC Faculty Scholarship 5-1-2015 Coutig o r-fiboacci Numbers Arthur Bejami Harvey Mudd College Curtis Heberle Harvey Mudd
More informationEfficient Feedback-Based Scheduling Policies for Chunked Network Codes over Networks with Loss and Delay
Efficiet Feedback-Based Schedulig Policies for Chuked Network Codes over Networks with Loss ad Delay Aoosheh Heidarzadeh ad Amir H. Baihashemi Departmet of Systems ad Computer Egieerig, Carleto Uiversity,
More informationWAVE-BASED TRANSIENT ANALYSIS USING BLOCK NEWTON-JACOBI
WAVE-BASED TRANSIENT ANALYSIS USING BLOCK NEWTON-JACOBI Muhammad Kabir McGill Uiversity Departmet of Electrical ad Computer Egieerig Motreal, QC H3A 2A7 Email: muhammad.kabir@mail.mcgill.ca Carlos Christofferse
More informationx 1 + x x n n = x 1 x 2 + x x n n = x 2 x 3 + x x n n = x 3 x 5 + x x n = x n
Sectio 6 7A Samplig Distributio of the Sample Meas To Create a Samplig Distributio of the Sample Meas take every possible sample of size from the distributio of x values ad the fid the mea of each sample
More informationLecture 4: Frequency Reuse Concepts
EE 499: Wireless & Mobile Commuicatios (8) Lecture 4: Frequecy euse Cocepts Distace betwee Co-Chael Cell Ceters Kowig the relatio betwee,, ad, we ca easily fid distace betwee the ceter poits of two co
More informationThe Firing Dispersion of Bullet Test Sample Analysis
Iteratioal Joural of Materials, Mechaics ad Maufacturig, Vol., No., Ma 5 The Firig Dispersio of Bullet Test Sample Aalsis Youliag Xu, Jubi Zhag, Li Ma, ad Yoghai Sha Udisputed, this approach does reduce
More information20. CONFIDENCE INTERVALS FOR THE MEAN, UNKNOWN VARIANCE
20. CONFIDENCE INTERVALS FOR THE MEAN, UNKNOWN VARIANCE If the populatio tadard deviatio σ i ukow, a it uually will be i practice, we will have to etimate it by the ample tadard deviatio. Sice σ i ukow,
More informationOn Parity based Divide and Conquer Recursive Functions
O Parity based Divide ad Coquer Recursive Fuctios Sug-Hyu Cha Abstract The parity based divide ad coquer recursio trees are itroduced where the sizes of the tree do ot grow mootoically as grows. These
More informationAn Adaptive Image Denoising Method based on Thresholding
A Adaptive Image Deoisig Method based o Thresholdig HARI OM AND MANTOSH BISWAS Departmet of Computer Sciece & Egieerig Idia School of Mies, Dhabad Jharkad-86004 INDIA {hariom4idia, matoshb}@gmail.com Abstract
More informationarxiv: v2 [math.co] 15 Oct 2018
THE 21 CARD TRICK AND IT GENERALIZATION DIBYAJYOTI DEB arxiv:1809.04072v2 [math.co] 15 Oct 2018 Abstract. The 21 card trick is well kow. It was recetly show i a episode of the popular YouTube chael Numberphile.
More informationShuffling. Shahrzad Haddadan. March 7, 2013
Shufflig Shahrzad Haddada March 7, 2013 Abstract I this paper we will talk about two well-kow shufflig methods, the Top to Radom ad the Riffle Shuffle. We are iterested i the umber of shuffles that will
More informationPROJECT #2 GENERIC ROBOT SIMULATOR
Uiversity of Missouri-Columbia Departmet of Electrical ad Computer Egieerig ECE 7330 Itroductio to Mechatroics ad Robotic Visio Fall, 2010 PROJECT #2 GENERIC ROBOT SIMULATOR Luis Alberto Rivera Estrada
More informationTHE OCCURRENCE OF TRANSIENT FIELDS AND ESD IN TYPICAL SELECTED AREAS
THE OCCURRENCE OF TRANSIENT FIELDS AND ESD IN TYPICAL SELECTED AREAS Stepha FREI Techical Uiversity Berli, Istitute of Electrical Power Egieerig Eisteiufer, 587 Berli, Germay e-mail: frei@ihs.ee.tu-berli.de
More informationObjectives. Some Basic Terms. Analog and Digital Signals. Analog-to-digital conversion. Parameters of ADC process: Related terms
Objectives. A brief review of some basic, related terms 2. Aalog to digital coversio 3. Amplitude resolutio 4. Temporal resolutio 5. Measuremet error Some Basic Terms Error differece betwee a computed
More informationGeneral Model :Algorithms in the Real World. Applications. Block Codes
Geeral Model 5-853:Algorithms i the Real World Error Correctig Codes I Overview Hammig Codes Liear Codes 5-853 Page message (m) coder codeword (c) oisy chael decoder codeword (c ) message or error Errors
More informationHistory and Advancement of the Family of Log Periodic Toothed Planer Microstrip Antenna
Joural of Electromagetic Aalysis ad Applicatios, 2011, 3, 242-247 doi:10.4236/jemaa.2011.36039 Published Olie Jue 2011 (http://www.scirp.org/joural/jemaa) History ad Advacemet of the Family of Log Periodic
More informationProcedia - Social and Behavioral Sciences 128 ( 2014 ) EPC-TKS 2013
Available olie at www.sciecedirect.com ScieceDirect Procedia - Social ad Behavioral Scieces 18 ( 014 ) 399 405 EPC-TKS 013 Iductive derivatio of formulae by a computer Sava Grozdev a *, Veseli Nekov b
More informationA study on traffic accident measures in municipal roads by using GIS
icccbe 010 Nottigham Uiversity Press Proceedigs of the Iteratioal Coferece o Computig i Civil ad Buildig Egieerig W Tizai (Editor) A study o traffic accidet measures i muicipal roads by usig GIS Satoshi
More informationELEC 350 Electronics I Fall 2014
ELEC 350 Electroics I Fall 04 Fial Exam Geeral Iformatio Rough breakdow of topic coverage: 0-5% JT fudametals ad regios of operatio 0-40% MOSFET fudametals biasig ad small-sigal modelig 0-5% iodes (p-juctio
More informationLossless image compression Using Hashing (using collision resolution) Amritpal Singh 1 and Rachna rajpoot 2
Lossless image compressio Usig Hashig (usig collisio resolutio) Amritpal Sigh 1 ad Racha rajpoot 2 1 M.Tech.* CSE Departmet, 2 Departmet of iformatio techology Guru Kashi UiversityTalwadi Sabo, Bathida
More informationA Novel Three Value Logic for Computing Purposes
Iteratioal Joural o Iormatio ad Electroics Egieerig, Vol. 3, No. 4, July 23 A Novel Three Value Logic or Computig Purposes Ali Soltai ad Saeed Mohammadi Abstract The aim o this article is to suggest a
More information4. INTERSYMBOL INTERFERENCE
DATA COMMUNICATIONS 59 4. INTERSYMBOL INTERFERENCE 4.1 OBJECT The effects of restricted badwidth i basebad data trasmissio will be studied. Measuremets relative to itersymbol iterferece, usig the eye patter
More informationTHE LUCAS TRIANGLE RECOUNTED. Arthur T. Benjamin Dept. of Mathematics, Harvey Mudd College, Claremont, CA Introduction
THE LUCAS TRIANLE RECOUNTED Arthur T Bejami Dept of Mathematics, Harvey Mudd College, Claremot, CA 91711 bejami@hmcedu 1 Itroductio I 2], Neville Robbis explores may properties of the Lucas triagle, a
More informationHOW BAD RECEIVER COORDINATES CAN AFFECT GPS TIMING
HOW BAD RECEIVER COORDINATES CAN AFFECT GPS TIMING H. Chadsey U.S. Naval Observatory Washigto, D.C. 2392 Abstract May sources of error are possible whe GPS is used for time comparisos. Some of these mo
More informationData Mining of Bayesian Networks to Select Fusion Nodes from Wireless Sensor Networks
www.ijcsi.org http://dx.doi.org/10.20943/01201604.1115 11 Data Miig of Bayesia Networks to Select Fusio Nodes from Wireless Networks Yee Mig Che 1 Chi-Shu Hsueh 2 Chu-Kai Wag 3 1,3 Departmet of Idustrial
More informationAkinwaJe, A.T., IbharaJu, F.T. and Arogundade, 0.1'. Department of Computer Sciences University of Agriculture, Abeokuta, Nigeria
COMPARATIVE ANALYSIS OF ARTIFICIAL NEURAL NETWORK'S BACK PROPAGATION ALGORITHM TO STATISTICAL LEAST SQURE METHOD IN SECURITY PREDICTION USING NIGERIAN STOCK EXCHANGE MARKET AkiwaJe, A.T., IbharaJu, F.T.
More informationTehrani N Journal of Scientific and Engineering Research, 2018, 5(7):1-7
Available olie www.jsaer.com, 2018, 5(7):1-7 Research Article ISSN: 2394-2630 CODEN(USA): JSERBR 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38
More information202 Chapter 9 n Go Bot. Hint
Chapter 9 Go Bot Now it s time to put everythig you have leared so far i this book to good use. I this chapter you will lear how to create your first robotic project, the Go Bot, a four-wheeled robot.
More informationEmpirical Estimation of Probability Distribution of Extreme Responses of Turret Moored FPSOs
Empirical Estimatio of Probability Distributio of Extreme Resposes of Turret Moored FPSOs Amir H. Izadparast Research ad Developmet Group, SOFEC. Housto, Texas, USA Aru S. Duggal Research ad Developmet
More informationOn Balancing Exploration vs. Exploitation in a Cognitive Engine for Multi-Antenna Systems
O Balacig Exploratio vs. Exploitatio i a Cogitive Egie for Multi-Atea Systems Haris I. Volos ad R. Michael Buehrer Mobile ad Portable Radio Research Group (MPRG), Wireless@Virgiia Tech Bradley Departmet
More informationADSP ADSP ADSP ADSP. Advanced Digital Signal Processing (18-792) Spring Fall Semester, Department of Electrical and Computer Engineering
ADSP ADSP ADSP ADSP Advaced Digital Sigal Processig (8-79) Sprig Fall Semester, 7 Departmet of Electrical ad Computer Egieerig OTES O RADOM PROCESSES I. Itroductio Radom processes are at the heart of most
More informationDesign of FPGA Based SPWM Single Phase Inverter
Proceedigs of MUCEET2009 Malaysia Techical Uiversities Coferece o Egieerig ad Techology Jue 20-22, 2009, MS Garde,Kuata, Pahag, Malaysia MUCEET2009 Desig of FPGA Based SPWM Sigle Phase Iverter Afarulrazi
More informationCAEN Tools for Discovery
Applicatio Note AN2506 Digital Gamma Neutro discrimiatio with Liquid Scitillators Viareggio 19 November 2012 Itroductio I recet years CAEN has developed a complete family of digitizers that cosists of
More informationINCREASE OF STRAIN GAGE OUTPUT VOLTAGE SIGNALS ACCURACY USING VIRTUAL INSTRUMENT WITH HARMONIC EXCITATION
XIX IMEKO World Cogress Fudametal ad Applied Metrology September 6, 9, Lisbo, Portugal INCREASE OF STRAIN GAGE OUTPUT VOLTAGE SIGNALS ACCURACY USING VIRTUAL INSTRUMENT WITH HARMONIC EXCITATION Dalibor
More information15 min/ Fall in New England
5 mi/ 0+ -4 Fall i New Eglad Before witer makes its appearace, a particularly warm fall bathes the forest i a golde shimmer. Durig the Idia Summer, New Eglad blossoms oe last time. Treetops are ablaze
More informationEMCdownload. Acknowledgements. Fair use
EMC_Sulight.idd 1 28/03/2013 09:06 Ackowledgemets Writte by Aa Sarchet, with Kate Oliver Edited by Kate Oliver Frot cover: Rebecca Scambler, 2013 Published by The Eglish ad Media Cetre, 2013 for EMCdowload.co.uk
More informationVIII. Shell-Voicings
VIII. Shell-Voicigs A. The Cocept The 5th (ad ofte the root as well) ca be omitted from most 7th-chords. Ratioale: Most chords have perfect 5ths. The P5th is also preset as the rd partial i the overtoe
More informationPHY-MAC dialogue with Multi-Packet Reception
PHY-AC dialogue with ulti-packet Receptio arc Realp 1 ad Aa I. Pérez-Neira 1 CTTC-Cetre Tecològic de Telecomuicacios de Cataluya Edifici Nexus C/Gra Capità, - 0803-Barceloa (Cataluya-Spai) marc.realp@cttc.es
More informationCombined Scheme for Fast PN Code Acquisition
13 th Iteratioal Coferece o AEROSPACE SCIENCES & AVIATION TECHNOLOGY, ASAT- 13, May 6 8, 009, E-Mail: asat@mtc.edu.eg Military Techical College, Kobry Elkobbah, Cairo, Egypt Tel : +(0) 4059 4036138, Fax:
More information5 Quick Steps to Social Media Marketing
5 Quick Steps to Social Media Marketig Here's a simple guide to creatig goals, choosig what to post, ad trackig progress with cofidece. May of us dive ito social media marketig with high hopes to watch
More informationSapana P. Dubey. (Department of applied mathematics,piet, Nagpur,India) I. INTRODUCTION
IOSR Joural of Mathematics (IOSR-JM) www.iosrjourals.org COMPETITION IN COMMUNICATION NETWORK: A GAME WITH PENALTY Sapaa P. Dubey (Departmet of applied mathematics,piet, Nagpur,Idia) ABSTRACT : We are
More informationThe Fundamental Capacity-Delay Tradeoff in Large Mobile Ad Hoc Networks
The Fudametal Capacity-Delay Tradeoff i Large Mobile Ad Hoc Networks Xiaoju Li ad Ness B. Shroff School of Electrical ad Computer Egieerig, Purdue Uiversity West Lafayette, IN 47907, U.S.A. {lix, shroff}@ec.purdue.edu
More informationAnalysis of SDR GNSS Using MATLAB
Iteratioal Joural of Computer Techology ad Electroics Egieerig (IJCTEE) Volume 5, Issue 3, Jue 2015 Aalysis of SDR GNSS Usig MATLAB Abstract This paper explais a software defied radio global avigatio satellite
More informationON THE FUNDAMENTAL RELATIONSHIP BETWEEN THE ACHIEVABLE CAPACITY AND DELAY IN MOBILE WIRELESS NETWORKS
Chapter ON THE FUNDAMENTAL RELATIONSHIP BETWEEN THE ACHIEVABLE CAPACITY AND DELAY IN MOBILE WIRELESS NETWORKS Xiaoju Li ad Ness B. Shroff School of Electrical ad Computer Egieerig, Purdue Uiversity West
More information(2) The MOSFET. Review of. Learning Outcome. (Metal-Oxide-Semiconductor Field Effect Transistor) 2.0) Field Effect Transistor (FET)
EEEB73 Electroics Aalysis & esig II () Review of The MOSFET (Metal-Oxide-Semicoductor Field Effect Trasistor) Referece: Neame, Chapter 3 ad Chapter 4 Learig Outcome Able to describe ad use the followig:
More informationShuffling Cards. D.J.W. Telkamp. Utrecht University Mathematics Bachelor s Thesis. Supervised by Dr. K. Dajani
Shufflig Cards Utrecht Uiversity Mathematics Bachelor s Thesis D.J.W. Telkamp Supervised by Dr. K. Dajai Jue 3, 207 Cotets Itroductio 2 2 Prerequisites 2 2. Problems with the variatio distace................
More informationData Acquisition System for Electric Vehicle s Driving Motor Test Bench Based on VC++ *
Available olie at www.sciecedirect.com Physics Procedia 33 (0 ) 75 73 0 Iteratioal Coferece o Medical Physics ad Biomedical Egieerig Data Acquisitio System for Electric Vehicle s Drivig Motor Test Bech
More informationSpread Spectrum Signal for Digital Communications
Wireless Iformatio Trasmissio System Lab. Spread Spectrum Sigal for Digital Commuicatios Istitute of Commuicatios Egieerig Natioal Su Yat-se Uiversity Spread Spectrum Commuicatios Defiitio: The trasmitted
More informationData Mining the Online Encyclopedia of Integer Sequences for New Identities Hieu Nguyen
Slide 1 of 18 Data Miig the Olie Ecyclopedia of Iteger Sequeces for New Idetities Hieu Nguye Rowa Uiversity MAA-NJ Sectio Sprig Meetig March 31, 2012 2 MAA-NJ Sprig Meetig Data Miig OEIS.b ü Ackowledgemets
More informationJoint Power Allocation and Beamforming for Cooperative Networks
It. J. Commuicatios, etwork ad System Scieces,, 4, 447-45 doi:.436/ijcs..4753 Published Olie July (http://www.scirp.org/joural/ijcs) Joit Power Allocatio ad Beamformig for Cooperative etworks Sodes Maadi,,
More informationWi-Fi or Femtocell: User Choice and Pricing Strategy of Wireless Service Provider
Wi-Fi or Femtocell: User Choice ad Pricig Strategy of Wireless Service Provider Yajiao Che, Qia Zhag Departmet of Computer Sciece ad Egieerig Hog Kog Uiversity of Sciece ad Techology Email: {cheyajiao,
More informationDensity Slicing Reference Manual
Desity Slicig Referece Maual Improvisio, Viscout Cetre II, Uiversity of Warwick Sciece Park, Millbur Hill Road, Covetry. CV4 7HS Tel: 0044 (0) 24 7669 2229 Fax: 0044 (0) 24 7669 0091 e-mail: admi@improvisio.com
More informationA generalization of Eulerian numbers via rook placements
A geeralizatio of Euleria umbers via rook placemets Esther Baaia Steve Butler Christopher Cox Jeffrey Davis Jacob Ladgraf Scarlitte Poce Abstract We cosider a geeralizatio of Euleria umbers which cout
More informationBOTTLENECK BRANCH MARKING FOR NOISE CONSOLIDATION
BOTTLENECK BRANCH MARKING FOR NOISE CONSOLIDATION IN MULTICAST NETWORKS Jordi Ros, Wei K. Tsai ad Mahadeve Iyer Departmet of Electrical ad Computer Egieerig Uiversity of Califoria, Irvie, CA 92697 {jros,
More informationMath 140 Introductory Statistics
6. Probability Distributio from Data Math Itroductory Statistics Professor Silvia Ferádez Chapter 6 Based o the book Statistics i Actio by A. Watkis, R. Scheaffer, ad G. Cobb. We have three ways of specifyig
More informationComputational Algorithm for Higher Order Legendre Polynomial and Gaussian Quadrature Method
Computatioal Algorithm for Higher Order Legre olyomial ad Gaussia Quadrature Method Asif M. Mughal, Xiu Ye ad Kamra Iqbal Dept. of Applied Sciece, Dept. of Mathematics ad Statistics, Dept. of Systems Egieerig
More informationThe Silicon Controlled Rectifier (SCR)
The Silico Cotrolled Rectifier (SCR The Silico Cotrolled Rectifier, also called Thyristor, is oe of the oldest power devices, ad it is actually employed as power switch for the largest currets (several
More informationSingle Bit DACs in a Nutshell. Part I DAC Basics
Sigle Bit DACs i a Nutshell Part I DAC Basics By Dave Va Ess, Pricipal Applicatio Egieer, Cypress Semicoductor May embedded applicatios require geeratig aalog outputs uder digital cotrol. It may be a DC
More informationOutline. Motivation. Analog Functional Testing in Mixed-Signal Systems. Motivation and Background. Built-In Self-Test Architecture
Aalog Fuctioal Testig i Mixed-Sigal s Jie Qi Dept. of Electrical & Computer Egieerig Aubur Uiversity Co-Advisors: Charles Stroud ad Foster Dai Outlie Motivatio ad Backgroud Built-I Self-Test Architecture
More informationA New Energy Efficient Data Gathering Approach in Wireless Sensor Networks
Commuicatios ad Network, 0, 4, 6-7 http://dx.doi.org/0.436/c.0.4009 Published Olie February 0 (http://www.scirp.org/joural/c) A New Eergy Efficiet Data Gatherig Approach i Wireless Sesor Networks Jafar
More informationSelection of Suitable Features of Lightning Rods for Secure Lightning Protection
Selectio of Suitable Features of Lightig Rods for Secure Lightig Protectio Hariza Che Mat Haris, Nor Farahaida Abdul Rahma, Nur Ashida Salim Faculty of Electrical Egieerig, Uiversiti Tekologi MARA 40450
More information3. Error Correcting Codes
3. Error Correctig Codes Refereces V. Bhargava, Forward Error Correctio Schemes for Digital Commuicatios, IEEE Commuicatios Magazie, Vol 21 No1 11 19, Jauary 1983 Mischa Schwartz, Iformatio Trasmissio
More informationPerformance Analysis of Channel Switching with Various Bandwidths in Cognitive Radio
Performace Aalysis of Chael Switchig with Various Badwidths i Cogitive Radio Po-Hao Chag, Keg-Fu Chag, Yu-Che Che, ad Li-Kai Ye Departmet of Electrical Egieerig, Natioal Dog Hwa Uiversity, 1,Sec.2, Da-Hsueh
More information