You are on page 1of 30

TheGamblingHabitsofOnlinePokerPlayers

September29,2011 IngoFiedler* ingo.fiedler@public.unihamburg.de Abstract Online pokeris adata goldmine. Recording actualgamblingbehavior givesriseto ahost of research opportunities. Still, investigations using such data are rare with the excep tionofninepioneeringstudiesbyHarvardMedicalSchoolwhicharereviewedhere.This paperfillspartofthevacuumbyanalyzingthegamblinghabitsofasampleof2,127,887 poker playing identities at Pokerstars over a period of six months. A couple of playing variables are operationalizedandwereanalyzedon theirownas well asconnected with eachotherinformoftheplayingvolume($rakeaplayerhaspaidinatimeframe). The main findings confirm the results of the Harvard studies: most online poker players only play a few times and for very low stakes. The median player played 7 sessions and 4.87 hours over 6 months. Multitabling was observed only rarely (median 1.05) and most players pay very low feesper hour (median US$0.87 per hour per table).Theplay ing volume is very low, too, with more than 50% of all players paying less than US$4.86 to the operators over 6 months. An analysis of the relationship between the playing ha bits shows that they reinforce each other with the exception of the playing frequency which moderates gambling involvement. The average values of the playing habits are considerably higher due to a small group of intense players: the 99% percentile player hasaplayingvolumethatis552timeshigherthanthatofthemedianplayer(US$2,685), and1%oftheplayersaccountfor60%ofplayingvolume(10%foreven91%).Thisgroup is analyzed more thoroughly, and a discussion shows that the first impulse to peg in tense players as (probable) pathological gamblers is wrong. Rather, future research is neededtodistinguishproblemgamblersfromprofessionalplayers. Keywords:online,poker,gambling,habits,behavior. * Ingo Fiedler is a Research Associate at the Institute of Law and Economics at the University of Ham burg, MaxBrauerAllee 60, 22765 Hamburg, Germany. Tel: +49 (0)40 428386454, fax: +49 (0)40 42838 6443. 1

1.Introduction
Electronic gambling opens up a new era of research on gambling behavior. So far, analyses have been limited to too small samples or to gambling behavior in laboratories where a monitoring bias cannot be accounted for. Another research method was to interview people about their gambling behavior a questionable approach since selfreports of behavior are often inconsistent (Baumeister et al. 2007). People generally tend to underreport their gambling behavior and pathological gamblers lie about theirs.1 Now, electronic gambling and online gambling in particularautomaticallyrecord actual gambling behavior.Thisallowsreliableandobjectiveanalysesofhugeandunbiaseddatasets.Suchresearchhow ever,isinitsinfancy.PioneeringworkinthisfieldcomesintheformofaseriesofninepapersfromHar vardMedicalSchool(LaBrieetal.2007,Brodaetal.2008,LaBrieetal.2008,LaPlanteetal.2008,Nelson etal.2008,LaPlanteetal.2009,Xuan&Shaffer2009,Braverman&Shaffer2010,LaBrie&Shaffer2011). Other research focusing on actual gambling behavior is still missing with the exception of Smith et al. (2009)whocomparethegamblingbehaviorofpokerplayersbeforeandafterbigwinsandbiglosses.To expand theunderstanding ofactual gambling behaviorthisstudy analyzesthegambling habits of a sam ple of2,127,887 pokerplaying identitiesatthe largest onlinepokeroperatorPokerstars over a periodof 6months. Thispaperisstructuredasfollows:thesecondsectionisareviewofthepaperseriesfromHarvardMedi cal School that focuses on the poker study by Nelson et al. (2009). The third section introduces the On line Poker Database of the University of Hamburg (OPDUHH) and operationalizes the playing habits of pokerplayerstomakethemmeasurableandcomparable.Theconnectionsbetweenthedifferentplaying habitsarealsoillustratedandthekeyfigureplayingvolumeisdefined.Section4presentstheempirical resultsandanalyzesagroupofintenseplayersindetail.Thefollowingsectiondiscussesthemainfinding
1

LyingisadefiningcriterionofpathologicalgamblingaccordingtoDSMIV

thatasmallgroupofplayersaccountforthemajorityoftheplayingvolume,byhighlightingthenecessity todistinguishbetweenpathologicaland(semi)professionalpokerplayers.

2.Literature:ThePaperSeriesfromHarvardMedicalSchool
All of the papers from Harvard Medical School on the actual behavior of online gamblers rely on a data set of approximately 47,000 betting accounts at the gambling operator bwin which were registered in February 2005.2 The studies can be divided into two separate groups. One group analyzes the gambling behavior of sport bettors, poker players and casino gamblers solely on a descriptive basis. The other group analyzes the gambling behavior of subsamples where problem gambling is indicated by account closing, selflimitation or limitation by bwin. This allows the authors to investigate the differences be tween recreational and probable pathological gamblers. The studies are unique in their approach be cause they analyze actual gambling behavior. This kind of data set overcomes the typical limitations and biases of selfreported data and allows an objective measurement of the gambling habits (see e.g. Xuan & Shaffer 2009). The advantages of a data set of actual gambling behavior are enormous, and the au thors even see a paradigm shift in gambling research (Shaffer et al. 2010). The authors distinguish be tween the heavily involved bettors (top 5% or top 1%) and the majority 95% (99%) of all participants. The main finding is that the group of the heavily involved bettors is significantly more active than the rest of the cohort. For example, the involvement of the intense poker players was roughly twice as long andtheyplayed7timesasmanysessions.Theywagered44timesandlost6timesmoremoneythanthe majorityoftheplayers. The analyses of the data set and the derived conclusions are manifold. Still, there are some limi tations. The major limitation was addressed by the authors and is inherent to these kinds of data sets:
2

Althoughthestudiesusethesamedataset,thenumberofaccountsdiffersbetween47.000and48.114.

bwin is just one online gambling operator and players may have accounts at multiple sites. Multiple ac counts seem especially likely for the most intense players. Hence, their playing behavior can only be ob served partially and the results underestimate their true gambling involvement. This problem is aggra vated in the study focusing on poker players by La Plante et al. (2009) because bwin is mainly a sports bettingoperatoranditcanbeassumedthatthesamplemostlyconsistsofpeoplewhoseprimarygameis sportsbetting,meaningthatthesubsampleofpokerplayersconsistsprimarilyofplayersforwhichpoker is their second or even third choice. Gamblers who mainly play poker games may instead sign up with other operators specializing in these games. But as these players are, by definition, more involved, the results of the bwin study underestimate the playing intensity of poker players.3 Although the authors admitthisdrawbackasabias,theydonotseeitasprobablybutratherasplausible.However,thechoice of the operator is important for the players, especially in poker. The reason is that the larger the opera tor and its network, the higher the liquidity of poker players there. This means players can choose be tween more tables to find their preferred game structure and limit. Economists call this effect a (posi tive)networkexternality(Katz &Shapiro,1985).ComparedtoPokerstarsorFullTiltPoker(atthattime), bwinisaminorplayerinthepokermarket(Fiedler&Wilcke,2011a)meaningthatonlyfewprimarypok erplayers will havesigned upwithbwin and if so,onlyas a second orthirdchoicesitetoplay at. Conse quently, the results of poker players gambling habits are not representative but underestimated. Al though the present study cannot overcome the inherent problem of people playing at multiple opera tors,thedatacomesfromthelargestpokeroperatorsand,hence,theanalyzedplayerpoolisrepresent ativeforallonlinepokerplayers. One issue particular to poker which was not addressed by Nelson et al. (2009) is the ambiguity of thetermmoneywagered.Moneywageredisakeyvariableformostgamblingopportunitiesbutnot
3

ThesamereasoningholdstrueforthecasinostudybyLaBrieetal.(2008).

for poker. First of all, it is unclear what is money wagered in poker: the money a player puts on the tablewhichisthenatrisk,orthesumwhichheactuallyputsinthepotduringahand,oreachindividual bet(meaningmultiplebetsperhand)?Thefollowingexamplesofdifferentplayerbetspointoutwhythe variabletotalwageredshouldbeconsideredcarefully:
Player A sits down with US$100 at a No Limit Holdem US$0.50/US$1 table. He plays just one hand,foldshiscardsandleavesthetablewithUS$99. Player B sits down with US$100 at a No Limit Holdem US$2/US$4 table. He plays just one hand, foldshiscardsandleavesthetablewithUS$96. Player C sits down with US$100 at a No Limit Holdem US$0.50/US$1 table. He plays just one hand,betsallUS$100duringthehandandleavesafterwards. Player D sits down with US$100 at a No Limit Holdem US$0.5/US$1 table. He plays 100 hands, folds 80 times without a betting, and during the other 20 hands his bets accumulate to a total of US$160. Player E sits down with 100% at a Limit Holdem US$0.50/US$1 table. He plays 1 hand, caps the bettingonallstreetstoatotalof$24andleavesafterwards.

Interpreting each of those players to have wagered US$100 omits analyzing the level of risk in different games and the betting strategies players adopt. In addition, money wagered loses value with growing difference between the expected values of bets and between their riskiness (standard deviation of the outcomes).Forexample,totalmoneywageredisperfectforinterpretingmoneywageredonredina roulette game. For money wagered on red and also numbers in roulette it loses some informative value as the riskiness of the bets differ. If now the expected value differs too, the variable total money wa gered loses even more of its explanatory power. In poker the expected values and the riskiness of bets differ greatly and the differences are aggravated by the path dependency of decisions during a poker hand. This study avoids this problem by operationalizing and measuring a players playing intensity and playing volume, which not only depend on the game and betting structure but also on the number of opponentsatatable.

3.DataandMethods
3.1TheOnlinePokerDatabaseoftheUniversityofHamburg(OPDUHH)
Online poker is a data goldmine. All operators display a lot of information in their lobbies about the people playing at their tables. It is easy to determine the origin of a player (city and/or country), the game type, betting structure, the limit of the table they are playing at, and of course the time and the date. Financed by the city of Hamburg, the Institute of Law & Economics at the University of Hamburg collected this data in the OPDUHH in collaboration with independent market spectator PokerScout. Softwareelectronicallygatheredplayerdataforthefollowingpokernetworks:Pokerstars,FullTiltPoker, Everest Poker, IPN (Boss Media) and Cake Poker. This software scanned each cash game table4 of the aforementionedpokersitesandcopiedthedisplayedinformationintoaSQLdatabase.5 Data collection was conducted for each poker site over a period of six months, enablingdata for 2,127,887 poker identities, including their country of origin and their playing habits, to be obtained. It took about ten minutes to scan all the operators tables and collect information about players seated at the tables. This translates to about 6 data points per hour or 25,920 over the course of six months and allows not only to determine the playing time per session of the players, but also to analyze differences intime.TheperiodofthedatacollectionranfromSeptember10,2009toMarch11,2010.6

3.2OperationalizationofPlayingHabits
Before measuring the playing habits of online poker players it is essential to operationalize the different observedvariablesandtheirinteractions,otherwisemisunderstandingsandconfusionabouttheirmean ing might occur. While total playing time, for example, is easy to understand, the playing volume of a
4 5

Playmoneytableswerenotobserved. ForamoredetaileddescriptionregardingthetechnicalapproachofthedatacollectionseeSakai&Haruyoshi2005. 6 Note that the period of data collection had to be extended due to technical problems such as downs of the server, software updatesanddisconnections.

player is not clear. Whats more, operationalization helps to overcome the typical question concerning totalmoneywageredwhichisakeyvariableformostgamblingopportunitiesbutnotforpoker. The information in the OPDUHH can be broken down into seven different variables to analyze the playing habits of an online poker player (they can also be connected with the origin of the player to allowcountryorregionspecificanalyses).Thesevariablesare:1)numberofsessions,2)playingtimeper session, 3) number of tables played simultaneously in a session (multitabling), 4) game structure (for example Texas Holdem or Omaha), 5) betting structure (for example No Limit or Fixed Limit), 6) number of players/seats at the table, 7) the size of the big blind7. Note: a session begins when a player who has notbeenactiveinthelast20minutessitsdownatanytable.ThisisdifferentfromthestudyofNelsonet al.(2009)whichdefinesasession(althoughnotexplicitly)asaplayerseatedatatableandbuyingchips regardless of whether he has played immediately beforehand at a different table. In addition, the present studys definition of a session allows to observe if a player is seated at multiple tables at the sametime,aspecificfeatureofonlinepoker. While analyzing each variable individually is interesting, they can also be combined with the in

formationoftheplayingduration(thetimebetweenthefirstandlastobservationofaplayer).Themost meaningful interpretations, however, are possible when the variables are connected with each other. Variables1),2),3),and7)arequantitativeandcanthereforeberelatedtoeachother.Forexample,mul tiplying the number of sessions with the playing time per session yields the total playing time over six months. However, variables 4), 5), and 6) are qualitative. Analyzing them separately is not meaningful. Butexcludingthemdistortsthepictureoftheplayinghabits.Forexample,playinganhourofFixedLimit
Inthegameofpokertheplayershavetopostasmallandabigblind.Regularly,thesmallblindishalfthebigblind.Bothblinds aretheminimumbettheplayershavetomakebeforetheygettheircards.Onlyoneplayerhastopostthebigblindandonethe smallblind(inexceptionalcasesthereisonlyonebigblindortwosmallblindsandonebigblind)whiletherestoftheplayersat atabledonothavetopostanyblind.Theplayerspostingtheblindsaredeterminedbythesocalledbuttonwhichmovesclock wise around the table from player to player after every hand to ensure that every player has to post the blinds equally over time.
7

Holdem with 9 other players cannot be compared with playing an hour of Pot Limit Omaha with 6 play ers. Someone sittingdownwith US$100in the Fixed Limit game is considerablyless exposedto riskthan someoneinthePotLimitOmahagame. Hence, the qualitative variables have to be operationalized and quantified. The one thing they have in common is that they all relate to the rake (the fee paid to the operator). Ceteris paribus: the moreplayersatatable,thelessrakeispaidperplayertotheoperator.InOmahamorerakeispaidthan in Holdem, in No Limit games the rake is higher than in Fixed Limit games. However, the magnitude of theseeffectsisnotstaticandalsodependsonthesizeofthebigblind.Hence,itisnecessarytocombine these three variables with the size of the big blind. This yields the average rake paid by a player per 100 hands a quantitative variable which can be related to the other variables of the playing habits. These values are important for the players as they determine together with bonuses and rebates the price theyarechargedforplayingpoker.Thus,theyalsodifferfromoperatortooperator. No Limit Texas Holdem is by far the most popular poker variant: 58.73% play this variant (see Appendix A). Figure 1 shows the average absolute and relative rake charged by the operators (industry average) for No Limit Holdem games with 6 and 10 players in relation to the size of the big blind. While theabsoluteamountofrakepaidper100handsonalimitincreasesinthesizeofthebigblind(themon ey at stake) it is evident that it decreases relatively to the size of the big blind. While a player at US$0.01/US$0.02 pays US$0.25 or 12.5 big blinds on average per 100 played hands to the operator at a table with 6 seats, he pays US$15.49 or 2.58 big blinds per 100 hands on a US$3/US$6 table. While the absolute amount increases sharply, the relative price in relation to the stakes (the big blind) is only 1/5. The reason for this lies in the rules for collecting the rake. Two are important here: 1) if the players do not see a flop, the operator does not take any rake and 2) the maximum rake paid is capped at a given amount (usually US$5). Both rules work in favor of the high stakes players: the pots grow larger but the

rakeiscappedandthehigherthelimits,thetightertheplayers(meaningtheyplayfewerstartinghands) sotheyseefewerflops,whichmeanstheypaylessrake.
Figure1:Rakepaidtotheoperatorsper100handsinNoLimitTexasHoldem(industryaverage) 40 20 35 30 US$/100h 25 20 15 10 5 0 18 16 14 12 BB/100h 10 8 6 4 2 0 US$/100h 10max BB/100h 10max

Pokeris azero sum game betweentheplayers so the averagerakepaid totheoperatorequals the play ing costs for the average player. While it is common for poker players to use rake paid per 100 handsto compare how much they have to pay, it is much more feasible for research questions to standardize the variable in time units.8 This allows a joint analysis with the playing time of a player and a comparison with the expenses for other games like slot machines. The average rake paid per hour by a player is an importantvariableandshallbedenotedwiththetermplayingintensity.Thehigherthestakes,themore hands per hour played,thelessopponents faced,the riskier the betting structure andthepoker variant, thehighertheplayingintensityinformoftheaveragerakepaidperhourtotheoperator.

The number of hands per hour differs in relation to the players at a table, the betting structure, the poker variant, and some what in relation to the size of the big blind. This means that the conversion of already large variety of values for the average rakepaidper100handsforthedifferentpokergamesyieldsabout2,400valuesfortherakeperhourintheend.Thevaluescan befoundintheappendixofFiedler&Wilcke,2011.

Figure 2shows the relationship between the different variables of the playing habits(number of sessions, average playing time per session, average number of tables played simultaneously and playing intensity).Theycanbeaggregatedtothetopfigureplayingvolume.Thisisdefinedastheproductofthe playingtimeovera6monthperiodtimesthenumberofaveragetablesplayedsimultaneouslytimesthe average$rakepaidtotheoperator.Theplayingvolumeofaplayerstateshowmuchmoneyaplayerhas paidtotheoperatorinthe6monthsoftheobservationperiod.
Figure2:Thedifferentvariablesoftheplayinghabitsandtheirrelationship

Playingvolume Numberofsessions Playingtime persession Gamestructure Bettingstructure Tablesize(seats) Bigblind Playingtime over6months Numberoftables playedsimultaneously Playingintensity= $rakeperhour

4.EmpiricalResults
4.1NumberofSessions
The total number of sessions observed over the 6 months period is 51,141,167. At 2,127,887 playing identities9 the average number of sessions played is 24.03. As the nicknames of the player identities wererecorded,thenumberofsessionseachplayerplayedwasalsodetermined.Themedianplayeronly played 7 sessions over 6 months. Presumably, there are lots of players who only played a few times,
Itisnotforallplayingidentitiesthatdatawererecorded.Thereasonisthatatthebeginningandattheendoftheobservation periods for each operator the session length could not be determined and, therefore, not be accounted for. For players who were only observed at these points, there exists no data on their number of sessions or playing time resulting in a small differ encebetweenthetotalnumberofobservedplayersandthenumberofplayerswithdataontheirnumberofsessions.
9

10

while a small number of intense players played frequently and created the large gap between the mean and median number of sessions played. This hypothesis is strengthened by the standard deviation of 49.3sessions7timesashighasthemedian.Thegapbetweenthemeanandthemedianvaluescanbe foundineveryvariableoftheplayinghabitsandisinvestigatedmoredeeplyineachcase.Itleadstotwo conclusions: (1) a small group of heavily involved poker players is responsible for the majority of the playing volume, and (2) the median values describe the gambling behavior of the typical online poker playermoreaccuratelythanthemeanvalues. The number of sessions played shows that a relatively large proportion of the players did not

play very often over the course of 6 months: 403,592, equivalent to more than 18% of all player identi ties, only played once. Nearly half a million identities were observed between two and four times and 18.2% between five and ten times. Another 17.2% played between 11 and 25 times while 10.3% of the sample was observed between 2650 times. 7.1% played between 50 and 100 sessions and 3.5% be tween 100 and 180 sessions. A group of 2.1% of the sample was seen more often than 180 times at the tablestheyplayedmorethanonesessionperday.

4.2Playingtimepersession
The average poker player stayed at the table for 50.27 minutes per session. At 42 minutes, the median player had an average session length of only slightly less. In comparison to the number of sessions the gap is relatively small and the average is not affected by a few extremely long sessions. The standard deviation of 37.76 minutes (0.9 times the median) supports this finding. Still, there is a gap between a majority playing short sessions (25% of the players played less than 25.5 minutes and 70% played less than an hour per session on average) and some people playing long sessions regularly (10% play 94.8 minutes or more per session and nearly 5% two hours or more). Analyzing the session length not by player but by sessions shows that nearly one third (32.4%) of all sessions ended in less than 30 minutes. 38.2% lasted between 31 and 60 minutes, 17.8% 61 to 90 minutes, and 6.8% 91 to 120 minutes. For 11

another3.7%ofthesessionsalengthoftwotothreehourswasrecordedand1.1%ofthesessionslasted morethan3hourswithoutabreak.

4.3Totalplayingtimeover6months
Asdiscussedbefore,thecombinationofthenumberofsessionsandtheirlengthsyieldsthetotal playingtimeofaplayerovertheobservationperiodof6months.Thisispossiblebecauseeachnickname isuniqueoneachpokerplatformandtheplayerscanberecognizedandtracked.Itisnoticeablethatthe averageplayingtimeover6monthswas25.28hoursfortheaverageplayerwhilethemedianplayeronly played 4.88 hours over the course of 6 months. Hence, the average value is again impacted by a small group of intense players, a hypothesis supported by the huge standard deviation of 65.21 hours (13.36 timesthemedianvalue).Itmeansthatthegapbetweentheaverageandthemedianvaluesofthenum berofsessionsandtheplayingtimepersessionisamplifiedbycombiningthemtothetotalplayingtime. Analyzingtherelativefrequencyoftheclassifiedtotalplayingtimeshowsthatalargeproportion of the players play poker rarely: 22.9% of all players did not play for more than an hour, 27.6% of the observed player identities played between 1 and 5 hours poker for real money over the course of 6 months,and20%haveatotalplayingtimebetween5and15hours.12.8%oftheplayerswereobserved for 15 to 35 hours and 10.6% of the sample for 35 to 100 hours (which still is not to be categorized as excessiveifpokerisahobbyforthem).Theproportionofplayerswhospentmorethan100hoursatthe virtual poker tables however, is not to be disregarded. 6.1% of all players have played more than 33 mi nuteseachdayonaverage.

4.4Multitabling
Online poker offers the possibility to play at more than one table at the same time. This is called multi tabling and onlypossiblebecause theplayers donothave tosit physically at thepokertablebutinfront ofacomputerscreenwheretheycanarrangemultipletablesnexttoeachother.Incontrasttotheanec 12

dotal evidence, online poker players do not tend to multitable frequently. In the average session the playerplayed at 1.31 tables simultaneously and in the median session at1.05 tables. Thegap is not very large and the standard deviation of 1.04 tables also suggests that the mean value is only marginally af fected by a small group playing many tables. Still, there is a gap between a majority playing just on one or partially at a second table and some people playing on more tables regularly: 10% of the players play at1.65,5%on2.36,and1%at6.03tablesonaverage. Analyzing multitabling not by player but by session shows that multitabling is most often not practiced on a regular basis (which yields a high average over all sessions of a player) but instead some times tried out by a lot of players (yielding only slightly increased averages per sessions for many play ers). Still, in 60.3% of all sessions the player was singletabling. In 15.8% of all sessions two tables were played simultaneously. In another 5.8% three tables were played at the same time and 5.1% of all ses sions were played at four tables simultaneously. Five or six tables were observed in 4.4%, seven oreight tables in 2.2%, and nine to 12 tables in 3.1% of all sessions. In 3.2% of the sessions the player played at 12ormoretablesatthesametime.Hence,massivemultitablingisnotexercisedregularlybymanyplay ersbutonlysometimesbyseveralplayers. Giventhatonaverageabout70handsareplayedpertableperhour,aplayerplayingat12tables simultaneouslycompletes840handsperhouror14perminute.Whilemostcombinationsofcardsare folded directly and on autopilot by the practiced player, it takes a lot of effort to analyze the informa tion in hands where he sees a flop. The player does not only have to evaluate the strength of his own hand but also has to consider a lot of other factors including the possible hands his opponent may hold in this particular situation as well as the perceived range of hands his opponent gives him. As the hand plays out, the information to be processed and evaluated by a player can get quite complicated and trainingvideosforpokerplayersandhandanalysisinpokerforumslike2+2showthatitiseasilypossible 13

to think about a single poker hand for a couple of hours. Hence, playing 14 hands per minute requires a veryhighamountofconcentrationandfocusorontheotherhandsuggestsrecklessness.

4.5Playingintensity
Theplayingintensityisdefinedastheaveragerakepaidbyaplayerperhourtotheoperator.Itdepends on the game structure (for example Holdem or Omaha), the betting structure (for example No Limit or FixedLimit),thenumberofplayersseatedatthetable,andthesizeofthebigblindcorrespondingtothe money at stake. The playing intensity is the cash flow from the players to the operator and equals the average loss per hour of an average skilled player. The average playing intensity was US$2.40 per hour per table. The median player paid considerably less rake: US$0.87 per hour per table. Paired with the relatively large standard deviation of US$4.46 (5.06 times the median amount) this leads to the conclu sion that there is a small group of players with a high playing intensity who drive the mean value. Com paredwiththecostofothergamblingopportunitieslikeslotmachines,onlinepokerisquiteinexpensive (for most players). Key reasons for cheap offers are operators situated in small countries (tax oases) whopaylowtaxesandaverysmallfeeornothingatallfortheirlicense.Butthemainreasonisprobably thatthemarginalcostsoftheoperatorare(nearly)zerobecausetheydonothavetopaydealersorcov er rent costs. Instead they use scalable software which costs the same, regardless of how many tables areoffered. Nearly every fifth online poker player (19.9%) pays US$0.20 or less per hour per table to the op

eratorand18.1%haveaplayingintensitybetweenUS$0.20andUS$0.50.Another15.1%oftheobserved playing identities played with an intensity of US$0.50 to US$1 per hour per table. Hence, 53.1% of all playerspaylessthanUS$1rakeperhouror,theotherwayround,theoperatorsearnlessthanUS$1per hour with more than half of their customers. The group of players paying US$12 per hour accounts for 15.7%ofthesampleandtheplayerswithanintensitybetweenUS$2andUS$5accountfor17.7%ofthe

14

sample.Another11.3%paybetweenUS$515andonly2.0%ofallplayeridentitieswereobservedtopay morethanUS$15perhourpertable.

4.6Topfigure:Playingvolumeover6months
The multiplication of the total playing time, multitabling and playing intensity yields the playing volume. Itisthetopfigureregardingplayinghabitsandstateshowmuchmoneyaplayerhaspaidtotheoperator overtheobservationperiodof6months.Theaggregatedplayingvolumeofallplayersequalstheopera torsrevenuesandtheplayerslosses. While the analysis of the individual variables of theplaying habits was already greatly influenced

by a small group of heavily involved poker players, this finding becomes even more evident through an analysis of playing volume. The total observed playing volume over 6 months for all players was US$378 million.10 This leads to an average player loss of US$177.51. The emphasis, however, is the huge gap be tweenthemeanandthemedianplayingvolume:50%ofthesamplepaidonlyUS$4.86over6monthsto the operators. The standard deviation of US$1,935 is 398 times the median amount and amplifies this difference. It can only be explained by a small group of players who have a huge playing volume and strongly impact the average value. These figures suggest that there is a small group of excessive poker players. This hypothesis is supported by further evidence in this subsection before the group of heavily involvedplayerswillbeanalyzedseparatelyinthefollowingsection. 29.8% of all player identities paid less than US$1 rake over 6 months. Their playing volume is negligible. Presumably, they deposited a small amount, lost it in a few hands to their opponents, and never logged inagain. Nearly one million player identitiesor 20.6% ofthe samplehave aplayingvolume of US$15 which can be regarded as marginal. Also the playing volume of the 15.0% of the players who
10

Thetotalplayingvolumecanbeextrapolatedtothetotalmarketforawholeyeartostatethemarketsizeandalsobebroken downbyoperatorstostatetheirrevenuesmarketshares.ThiswasdoneinFiedler&Wilcke,2011a.

15

paid between US$5 and US$15 is very small. Relative to the observation period of 6 months even ex penses of US$1550 by approximately 200,000 players (14.2%) is not much. Nearly every tenth person (9.4%) has a playing volume of US$50150 over 6 months which cannot be disregarded but is not exces sive either and is still below the average value. 6.3% of the players paid between US$150 and US$500 raketotheoperatorsandtheyarepotentiallyatrisk.4.7%ofthesamplepaidmorethan$500.Giventhe smallfeesinonlinepoker,theirplayingvolumecanbecalledexcessive. Before analyzing the group of the intense players in more detail in the next section, it is to be highlighted that the playing volume of a player equals the payment to the operator but does not equal the players losses. Players can also lose money to their opponents (or win from them). Presumably, un trainedplayerswhoplaypokerinfrequentlylosemoneyonaveragetotheiropponentswhilethetrained players usually win (for empirical evidence, see Fiedler & Rock 2009). Hence, players with a low playing volume tend tohave higher losses than the rake paid tothe operators, while players with a high playing volumehavelessexpensesorevenwinnings.Forthisreason,aninterpretationofanindividualsplaying volume as his total losses is not meaningful. The playing volume can only be interpreted as players losses when aggregated. Still, the use of playing volume to determine the involvement of an individual player is reasonable. It allows the conclusion that most players have a small playing volume and are not at risk to develop an addiction, while a small group has an excessive playing volume and may be patho logicaland/orprofessionalgamblers.

4.7Playinghabitscombinedwithplayingduration
Some of the variables of the playing habits like the number of sessions are necessarily affected by the playing duration which is the time between the first and last observation of a player. On average the players were observed 55.32 days and the median player 27 days. The standard deviation of 60.83 days is more than twice as large as the median value and points out that the distribution is affected by some players who were observed over the whole data collection period. Combining playing duration with the 16

number of sessions yields an average of .74 and a median of .60 sessions per day. This suggests sup ported by the relatively low standard deviation of .66 sessions/day that even the most intense players do not play much more often per day than recreational players. On the other hand this value might be biasedbythegroupofplayerswhowereonlyobservedononedayandstoppedplayingthereafter.They have a sessions/day ratio of at least one and account for more than 20% of the sample. This is a draw back not inherent to the ratio playing time/playing duration. On average the sample played 38.70 mi nutes per day. The median value is 20 minutes per day and the standard deviation 53.62 minutes/day. Herewefindagainthattheaverageisstronglyaffectedbyasmallgroupofplayerswithahighexposure. The most interesting combination is playing volume per playing duration. The average rake/day is US$2.48 and more than 9 times larger than the median value of US$.27/day. This suggests, again, that a small group of players account for most of the playing volume. However, although the standard devia tionof14.44US$/dayisrelativelyhuge,itisnotaslargecomparedtothemedianvalueasintheanalysis ofplayingvolumewithoutconsiderationoftheplayingduration(53.5xto398x).Thisleadstotheconclu sion that the small group of the most involved poker players dominate in every variable of the playing habits.

4.8Relationsbetweenthedifferentvariablesofthegamblinghabits
The above results suggest that the variables of the playing habits reinforce each other. This hypothesis can be tested by analyzing the relationship betweenthem which allows conclusions about whether they reinforceeachotherorthereisamoderatingvariabletobedrawn.Infact,itisobviousthattotalplaying time has to be positively related to the number of sessions as it is defined as the product of number of sessions and average session length. However, the relationship is not necessarily positive between all variables. For example, there might be a negative relationship between playing intensity and number of tables: The higher playing intensity, the more money is at stake, hence people may decide to play at fewer tables. A KolmogorovSmirnovLillefors test for normality of a random 1% sample shows that all 17

variablesarenotnormallydistributed(allsignificantatp<0.001).11Therefore,therelationshipsarecalcu lated as nonparametric rankorder correlations accordingto Spearman (). Table 1 presents the results. As is to be expected by its definitions, the relationships between total playing time and number of ses sions and session lengthis exceptionallypositive as well as the correlation ofplaying volume toallother variables of theplaying habits. This is also true for the correlations between playing duration tonumber ofsessions,playingtime,andplayingvolume.
Table1:NonparametricSpearmancorrelationsamongthevariablesofthegamblinghabits(n=2,127,887).
Session Total Sessions Length PlayingTime Sessions .360** .939** SessionLength .644** PlayingTime Tables PlayingIntensity PlayingVolume Playingduration Sessions/day Time/day Rake/day **Correlationsignificantatp<.01. Tables Playing Playing Playing Sessions/ Time/ Rake/ Intensity Volume Duration Day Day Day .534** .160** .793** .831** .171** .139** .264** .428** .031** .527** .235** .049** .612** .483** .587** .141** .836** .749** .100** .345** .399** .113** .564** .416** .037** .246** .357** .633** .187** .106** .078** .674** .667** .130** .243** .689** .607** .290** .029** .771** .448** .642**

More meaningful are the correlations between number of tables played which is strong positive

to number of sessions and session length (and hence to total playing time) and playing duration. This means that people who play longer and more often also tend to play at more tables (and vice versa). Anotherinterestingfindingcomesfromthecorrelationsofplayingintensity.Althoughplayingintensityis positively relatedto all othervariables oftheplayinghabits, thecorrelation is relatively weak(especially to the session length) compared to the other relationships between the variables. This suggests that when people play more often, for a longer time, and at more tables they only slightly increase their stakes.

11

Sessions: KS=.322, session length: KS=.117; total playing time: KS=.350; number of tables: KS=.382; playing intensity: K S=.296; playing volume: KS=467; betting days: KS=.186; sessions/day: KS=.177; playing time/day: KS=.236; rake/day: K S=.432.

18

Analyzing the relationship between the combination of playing habits with playing duration yieldsimportantresults.Thecorrelationbetweensessions/dayandplayingdurationisstronglynegative. Thismeansthatthehigherplayingfrequencyofaplayerthemorelikelyheistostopgambling.Withthe exception of the correlation to session lengths, sessions/day shows a weak negative correlation to all other playing habits. This means that playing very often in a short period of time reduces overall gam bling involvement. This finding might be counterintuitive when it comes to pathological gambling. But it is reasonable for recreational players who have a given limit for their expenses and stop when it is reached (they reach it faster when they play more frequently). However, playing frequently does not mean playing long sessions. And the correlations of the time spent playing poker per day are different fromthoseofsessions/day.Whiletime/dayisnegativelycorrelatedtoplayingintensityandplayingdura tion it is positively related to the other playing variables. Rake/day is also positively related to all va riables of the playing habits with the exception of playing duration. Overall, it can be concluded that the only moderator for gambling involvement is playing frequency while all other playing habits reinforce eachother.

4.8Thegroupofintenseplayers
The playing habits of intense players differ from those of casual players. Table 2 presents a summary of theresultsforthedifferentvariablesofplayinghabitsandcomparesthemeanandmedianwiththosein thetop10%,top5%andtop1%players.
Table2:Summaryoftheplayinghabits(n=2,127,887). Median Numberofsessions 24.03 7 Sessionlengthinmin. 50.27 42.0 Totalplayingtimeinh 25.28 4.88 Numberoftables 1.31 1.05 Playingintensity($rake/h) 2.40 .87 Playingvolumein$ 177.5 4.86 Playingduration 55.32 27 Sessions/day .74 .60 Playingtime/day 38.70 20.00 2.48 .27 Playingvolumein$/day

49.30 37.76 65.21 1.04 4.46 1,935 60.83 .66 53.62 14.45

Top10% 63 94.78 62.78 1.65 6.12 173.9 160 1.50 98.34 4.42

Top5% 108 118.6 117.6 2.36 9.90 460.1 175 2 142.03 9.15

Top1% 247 182.3 318.0 6.03 19.75 2,685 182 3 259.00 35.42

Total 51,141,167 53,785,011 377,714,269

19

The increase in the session length from the median player (42 minutes) to the intense players is moderate. The top 10% player played 94.8 minutes on average per session, the top 5% player 119 mi nutes and the top 1% player 182 minutes. The increase is considerably higher with the number of ses sions. While the median player only played 7 sessions over the course of 6 months, the top 10% player played63,thetop5%player108,andthetop1%player247sessions.Hence,itcanbededucedthatthe huge difference between the total playing time of the median player (4.88 hours) and intense players (63,118and318hours) isdueto the number of sessionsandonlyslightly affectedbytheplaying length per session. Multitabling is uncommon among recreational players and median players only play 1.05 tables at the same time, but it is common among intense players: the top 10% player played 1.65, the top 5% player 2.36 and the top 10% player 6.03 tables simultaneously. The ratio intense player to me dian player is also notable when it comes to playing intensity. While the median player pays US$0.87 rakeperhourtotheoperator,thetop10%playerpaysUS$6.12,thetop5%playerUS$9.90,andthetop 1% player even US$19.75 or nearly 21 times the median amount. Combining playing habits with playing volume widens the gap between median and intense players greatly. The median player paid US$4.86 rake to the operators over 6 months and the top 10% player already 36 times as much (US$174) which equals the average of US$178. The average is mainly driven by the most intense players. The top 5% playerwas observed tohave paid US$460 andthetop 1%playereven US$2,685 552 timesthemedian amount. The analysis by percentiles supports the evidence that most online poker players only have a very low playing volume and a small group plays intensely (see figure 9). The playing volume increases exponentially with rising percentile. The 90% percentile player nearly pays nearly twice as much as the 85%player,theincreasetothe95%percentileis164%andtothe99%percentileeven483%.
Figure3:Playingvolumein$rakepaidover6monthsbypercentiles((n=2,127,887).

20

3000 2500 $rake 2000 1500 1000 500 0


460 294 362 174 204 243 89 65 0.1 0.2 0.4 0.7 1.1 1.6 2.4 3.4 4.8 6.7 9.4 13 19 27 41 852 608

2685

1334

5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 91 92 93 94 95 96 97 98 99 Percentile

Hence, the operator needs more than 500 recreational poker players to get as much revenue as he gets from one very intense player and it can be concluded that the operators generate most of their revenue from the intense players. This finding is validated by the comparison of the aggregated playing volume of the intense players to the whole sample (see table 3). 10% of the players account for 91.06% of all rake paid, 5% for 83.1% and still more than half of each dollar (59.59%) is generated by just 1% of the players. Their share of the total expenses is more than the 80/20 Pareto principle, which states that for most consumer goods about 80% of the revenues come from 20% of the customers. Viewing such numbers in the context of gambling, the first idea that comes to mind is that the intense players are ei ther pathological gamblers or at risk of becoming pathological. But this conclusion may be premature in thelightoftheskillelementinpokerandtheprofessionalplayers.
Table3:Aggregatedplayingvolumeoftheintenseplayersandtheirshareofthetotalplayingvolume Playergroup Playingvolumein$rakepaid Shareoftotalplayingvolume 225,086,489 Top1% 59.59% Top5% 313,888,432 83.10% Top10% 343,956,948 91.06% All 377,714,269 100%

21

5.DiscussionandPerspectives
5.1Intensepokerplayersaretheypathologicalgamblers?
One major challenge arises when analyzing games with skill elements. In poker and to a somewhat lesser extent in sports betting the influence of skill is large enough that professionals can play with a positive expected value and win in the long run (see pokertableratings.com and sharkscope.com for the results of professional poker players). Skill matters a great deal in the game of poker (Cabot & Hannum, 2005).Playershaveseveralpossibilitiestoinfluencetheoutcomeofthegame.Theseare:folding,calling, betting, raising, and reraising before the flop, on the flop, on the turn, and on the river. If the game is played as No Limit, the player can also decide how much to bet, raise, or reraise. These decisions de pend on many influential factors, such as the position at the table, the size of the pot (pot odds), the range of the possible hands of the opponent(s) and, of course, on the cards of the player and the com munitycards.Theskillinpokeristointerpretandweighupthesefactorsaccordinglyandthenmakethe bestdecisions(Fiedler&Rock2009). In poker, relative skill matters (Dreef et al. 2003). There are relatively skilled players who consis tently win money from their opponents and relatively unskilled players who lose this money (although thisgroupmaybeskilledinrelationtootherplayers).Duetothefeesinformofraketheplayershaveto paytotheoperator,mostplayersloseoverall,includingthosewhoarebetterthantheiropponents.Still, there are players who are so skilled that they overcompensate this disadvantage and win money in the longrun.12Thegroupofthewinningplayerscanbebrokendownintothreesubgroups:1)thesuccessful recreational players, 2) the semiprofessional players, and 3) the professional players. The successful recreationalplayersarethelargestofthesegroups.Theymaybelongtermwinnersbuttheirskillisonly sufficienttobeatthelowestlimits.Thatmeanstheyeitherwinjustsmallamountsofmoneyandpokeris
12

Seeforexamplewebsiteswhichkeeptrackoftheresultsofpokerplayerslikepokertableratings.com.

22

not attractive for purely financial reasons or they fulfill the peterprinciple and are water boys who climb to limits where more money is at stake but where they are not good enough any longer to win. Still, they probably play more than the average recreational player. The group of the semiprofessional playersconsistsofindividualswhoseskillissufficienttohavesuccessinafinanciallymeaningfulcontext. However, the people in this group have a fulltime occupation. Hence, they only play in their free time but on higher limits than the successful recreational players and they see poker as a lucrative possibility foranadditionalsourceofincome.Thegroupoftheprofessionalplayersisverysmall.Itconsistsofplay ers who are sufficiently skilled to consistently win money by playing poker to an extent that they do not need another job. They are not necessarily more skillful than the players in the semiprofessional group buttheyspendconsiderablymoretimeplayingpokerandregarditastheirjob.Alloftheseplayershave an incentive to play often and (and for larger amounts) and a higher than average playing volume. This may be reached by playing high limits, playing many tables, many or long sessions or a combination of these.Allsemiprofessionalsandprofessionalsandalargenumberofthesuccessfulrecreationalplayers have a high involvement and can be found in the group of intense players. Hence, they affect the dis tributions of the playing variables. This is a huge problem when trying to identify excessive or even pa thologicalpokerplayers(andsportsbettors)bytheirplayingvolume. Ontheotherhand,notallintenseplayersarewinningplayerswhichindicatesthatalsopatholog

ical players are in the group of intense players. Thus, the question is how many players of the intense players are pathological and how many are professionals, and also whether these players are only good at poker and play due to the financial incentive, orwhether they are (also) addicted to poker. According to Weinstock & Petry (2009) pathological and professional gamblers differ only in the degree of their impulsivity. However, this data set does not cover the betting patterns of the players which may give insights to the degree of impulsivity. As Smith et al. (2009) analyze betting patterns of poker players but not in direct connection to impulsivity, the question has to be addressed by future research which ana 23

lyzesactualplayingbehaviorinmoredetail.Untilsuchresearchisavailable,itcanonlybesuspectedthat the group of the intense players mostly consists of (semi)professional players, pathological players and (semi)professional players who are addicted to poker but have not suffered any negative financial con sequences(yet).

5.2Limitations
Althoughthestudyyieldsmanyfindingstherearesomelimitations.Pokerplayerscaneasilyplayonmul tiple sites and, somewhat less likely, on the same site with multiple user names. This data set cannot take this fact into consideration and as a consequence every observed nickname at each site is inter preted separately. Thus, players with multiple accounts are interpreted as multiple players. This is a problem inherent to all analyses of actual playing behavior: they are always partial analyses as gambling behavior at different locations or games is not recorded. Underestimation is the result. For this study, it mainly affectstheplayingbehaviorofintenseplayersastheyaremostlikelytoplayatmultiplesites.On the other hand, it may also be possible that more than one person uses the same player identity (ac countsharing),forexamplefriendsorfamilymembers. A more important limitation is that cash flows between the players were not observed. Thus, it

cannot be determined whether a player is winning or losing. However, this is important information which would help to give a clearer insight into high volume play. It was shown for example, that players who play more often lose less (Nelson et al., 2009) and even win (Fiedler & Rock, 2009). Thus, it can be suspected, that the group of the intense players in this sample is losing less than the rest of the sample. However,thishypothesiscouldnotbeinvestigated.

5.3ConclusionsandPerspective
Electronic gambling and online gambling in particular offer the possibility of analyzing large unbiased data sets of actual playing behavior and invite new promising research. Harvard Medical School was the 24

firsttodothisinaseriesofninepapers.Onegroupofpapersdescribedgamblingbehaviorandthemain conclusion was that most players do not play very often, while a small group plays intensely. However, the conclusions for the poker players by Nelson et al. (2009) have to be considered carefully because these data sets are not representative as bwin is mainly a sports betting operator and only offers poker on the side. Furthermore, the authors did not address therole ofskill inpokerwhich canleadtoprofes sionalgamblersinfluencingthevariablesofgamblingbehavior. This paper advances research in this field forward by analyzing actual gambling habits of online poker players by means of a large and unbiased sample of 2,127,887 player identities from the Online Poker Data base of the University of Hamburg (OPDUHH) who were tracked over 6 months at five dif ferentpokeroperators. Inadditionto a players city orcountry of residence,software recorded who sits athowmanyandwhatkindoftableseverytenminutes.Thisdatawasoperationalizedintothefollowing variables:numberofsessions,timespentpersession,totalplayingtimeandtheplayingintensityinform of $ rake paid per hour and table to the operator. This way of operationalizing the variables of playing habits makes sense, not only against the background that poker is a game between players and not against the house, but also because the variables of the playing habits can be analyzed in isolation as well as in combination with each other. This allows the key figure total playing volume to be defined, indicatinghowmuchrakeaplayerhaspaidtotheoperatoroveragiventimeframe(here6months). ThemainfindingconfirmstheresultsoftheHarvardstudies:mostonlinepokerplayersonlyplay

rarely and for low stakes. The median number of sessions is 7 and the median playing time over 6 months is 4.88 hours. Regarding the playing intensity, it is notable that most players pay very low fees per hour (median is US$0.87 per hour per table). Also, a different definition of the term session al lowed to analyze multitabling, a specific feature of online poker made possible due to the virtual nature of gameplay.It was foundthat mostplayersdonotmake useof theunique opportunityofonline poker toplayatmultipletablesatthesametime.Themedianplayerwasobservedat1.05tablessimultaneous 25

ly. Hence, the total playing volume of the median player is also very low: more than 50% of all players paidlessthan US$4.86inrake to theoperators over 6 months.However, theaverage values oftheplay ing habits are considerably higher than the median values and they are highly affected by a small group of intense players. For example, the 99% percentile player has a 552 times higher playing volume than the median player (US$2,685). This is a value much higher than that found by Nelson et al. (2009). This small group of players accounts for most of the playing volume: operators earn 59.6% of their revenues fromonly1%ofthesample.5%oftheplayersaccountfor83.6%and10%for91.1%ofplayingvolume. Thegroupofhighvolumeplayersisnotonlyinterestingfortheindustrybecauseoftherevenue

they generate but also for research on gambling addiction. However, it is wrong to label every one of themasa(probable)pathological gambler,becauseinthelongrunskillplaysakeyrolefortheoutcome inpoker.Sophisticatedplayersareabletoplaywithapositiveexpectedvalue.Thus,incontrasttotypical gambling where no skill is involved, the group of intense players in poker consists of pathological gam blers as well as (semi)professional players earning a living by playing poker. When analyzing poker it is important to keep this in mind. Consequently, it is importantthat future research addresses the issueof a reliabledistinctionbetweenprofessional and pathologicalpoker players. There aretwodifferent alter natives to accomplish this goal. One approach is to dig deeper into the actual betting decisions of poker players (or other gamblers) to find tendencies of chasing, reinforcement or irrationality. The other ap proachistocombinedataonplayinghabitswithinterviewdata.Bothideasseempromisingandcapable ofpushingtheboundariesofcurrentresearchforward.

Literature
Baumeister,R.F.,Vohs,K.D.&Funder,D.C.(2007).PsychologyastheScienceofSelfReportsandFinger Movements. Whatever Happened to Actual Behavior?, Perspectives on Psychological Science, 2(4), 396 403. 26

Blaszczynski, A. (1999). Pathological gambling and obsessive compulsive spectrum disorders. Psychologi calReports,84,107113. Braverman, J., & Shaffer, H. J. (2010). How do gamblers start gambling: Identifying behavioral markers forhighriskInternetgambling.EuropeanJournalofPublicHealth,16. Broda,A.,LaPlante,D.A.,Nelson,S.E., LaBrie,R.A.,Bosworth,L.B.,&Shaffer,H.J.(2008).Virtualharm reductioneffortsforInternetgambling:EffectsofdepositlimitsonactualInternetsportsgamblingbeha vior.HarmReductionJournal,5,27. Cabot, A. & Hannum, R. (2008). Poker: Public Policy, Law, Mathematics and the Future of an American Tradition,T.M.CooleyLawReview,22:443513. EUCommission(2011).OnonlinegamblingintheInternalMarket,GreenPaper. Fiedler, I., & Rock, J.P. (2009). Quantifying Skill in Games Theory and Empirical Evidence for Poker, GamingLawReviewandEconomics,13,5057. Fiedler, I., & Wilcke, A.C. (2011). Der Markt fr Onlinepoker. Spielerherkunft und Spielerverhalten, BoD Verlag,Norderstedt. Fiedler, I., & Wilcke, A.C. (2011a). The Market for Online Poker, forthcoming, download at SSRN: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1747646. Guyatt, G.H., Sackett, D.L., & Cook, D.J. (1993). Users guide to the medical literature: II. How to use an articleabouttherapyorprevention:A.Aretheresultsofthisstudyvalid?JAMA,270(21),25982601. KatzM.L.,&ShapiroC.(1985).Networkexternalities,competition,andcompatibility.AmericanEconom icReview,75,424440.

27

LaBrie, R. A., LaPlante, D. A., Nelson, S. E., Schumann, A., & Shaffer, H. J. (2007). Assessing the playing field:AprospectivelongitudinalstudyofInternetsportsgamblingbehavior.JournalofGamblingStudies, 23,347362. LaBrie R.A., Kaplan, S.A., LaPlante, D.A., Nelson, S.E., & Shaffer, H.J. (2008). Inside the virtual casino: A prospective longitudinal study of actual Internet casino gambling. European Journal of Public Health, 18(4),410416. LaBrie, R. A., & Shaffer, H. J. (2011). Identifying behavioral markers of disordered Internet sports gam bling.AddictionResearch,&Theory,19(1),5665. Langewisch, M.W., & Frisch, R.G. (2002). Classification of Pathological Gambling as an Impulse Control Disorder.TheelectronicJournalofGamblingIssues,3. LaPlante, D.A., Schumann, A., LaBrie, R.A., & Shaffer, H.J. (2008). Population trends in Internet sports gambling.ComputersinHumanBehavior,24,23992414. LaPlante, D. A., Kleschinsky, J. H., LaBrie, R. A., Nelson, S. E., & Shaffer, H. J. (2009). Sitting at the virtual pokertable:A prospectiveepidemiologicalstudy of actualInternetpokergamblingbehavior. Computers inHumanBehavior,25(3),711717. Nelson,S.E.,LaPlante,D.A.,Peller,A.J.,Schumann,A.,LaBrie,R.A.,&Shaffer,H.J.(2008).Reallimitsin the virtual world: Selflimiting behavior of Internet gamblers. Journal of Gambling Studies, 24(4), 463 477. Partygaming(2009),AnnualReport2008. Potenza, M. N., Steinberg,M.A., Skudlarski,P., Fulbright, R. K.,Lacadie,C. M.,Wilber, M. K., Rounsaville, B. J., Gore, J. C., & Wexler, B. E. (2003). Gambling urges in pathological gambling: A functional magnetic resonanceimagingstudy.ArchivesofGeneralPsychiatry,No.160,19901994. 28

Shaffer,H.J.,Peller,A.,J.,LaPlante,D.A.,Nelson,S.E.,&LaBrie,R.A.(2010).Towardaparadigmshiftin Internet gambling research: From opinion and selfreport to actual behavior. Addiction Research, & Theory,18(3),270283. Smith, G. Levere, M., & Kurtzman, R. (2009), Poker Player Behavior After Big Wins and Big Losses, Man agementScience55(9),15471555. Weinstock, J. & Petry, N. M. (2009). Professional & Pathological Gamblers: Similarities and Differences, Paperpresentedat14thInternationalConferenceonGamblingandRiskTaking,LakeTahoe. Xuan, Z., & Shaffer, H. J. (2009). How do gamblers end gambling: Longitudinal analysis of Internet gam bling behaviors prior to account closure due to gambling related problems. Journal of Gambling Studies, 25(2),239252.

29

Appendix:Pokerplayerspervariantandlimit
Microstakes Lowstakes Midstakes Highstakes Total Absolute %Var %Tot Absolute %Var %Tot Absolute %Var %Tot Absolute %Var %Tot Absolute %Tot TexasHoldemNL 3,015,319 48.43% 28.44% 2,567,389 41.24% 24.22% 621,010 9.97% 5.86% 22,404 0.36% 0.21% 6,226,122 58.73% TexasHoldemFL 936,269 52.55% 8.83% 674,773 37.87% 6.37% 158,451 8.89% 1.49% 12,334 0.69% 0.12% 1,781,827 16.81% TexasHoldemPL 178,560 33.15% 1.68% 335,521 62.30% 3.17% 24,329 4.52% 0.23% 170 0.03% 0.00% 538,580 5.08% TexasHoldemML 8,943 68.81% 0.08% 3,913 30.11% 0.04% 93 0.72% 0.00% 47 0.36% 0.00% 12,996 0.12% OmahaNL 652 56.99% 0.01% 313 27.36% 0.00% 155 13.55% 0.00% 24 2.10% 0.00% 1,144 0.01% OmahaFL 15,657 54.97% 0.15% 7,730 27.14% 0.07% 5,073 17.81% 0.05% 21 0.07% 0.00% 28,481 0.27% OmahaPL 404,516 41.06% 3.82% 404,347 41.04% 3.81% 162,187 16.46% 1.53% 14,203 1.44% 0.13% 985,253 9.29% OmahaHi/LoNL 31,771 24.61% 0.30% 74,929 58.04% 0.71% 22,190 17.19% 0.21% 200 0.15% 0.00% 129,090 1.22% OmahaHi/LoFL 50,068 33.54% 0.47% 67,797 45.41% 0.64% 28,307 18.96% 0.27% 3,120 2.09% 0.03% 149,292 1.41% OmahaHi/LoPL 79,438 48.29% 0.75% 69,120 42.02% 0.65% 15,736 9.57% 0.15% 209 0.13% 0.00% 164,503 1.55% OmahaHi/LoML 16 44.44% 0.00% 20 55.56% 0.00% 0 0.00% 0.00% 0 0.00% 0.00% 36 0.00% 7CardStudNL 9 52.94% 0.00% 3 17.65% 0.00% 5 29.41% 0.00% 0 0.00% 0.00% 17 0.00% 7CardStudFL 70,155 44.41% 0.66% 72,992 46.21% 0.69% 13,889 8.79% 0.13% 929 0.59% 0.01% 157,965 1.49% 7CardStudPL 37 48.05% 0.00% 38 49.35% 0.00% 2 2.60% 0.00% 0 0.00% 0.00% 77 0.00% 7CardStudHi/LoFL 29,303 39.07% 0.28% 35,860 47.82% 0.34% 8,930 11.91% 0.08% 903 1.20% 0.01% 74,996 0.71% 5CardStudFL 0 0.00% 0.00% 251 100.00% 0.00% 0 0.00% 0.00% 0 0.00% 0.00% 251 0.00% 5CardDrawNL 0 100.00% 0.15% 15,382 100.00% 0.15% 0 0.00% 0.00% 0 0.00% 0.00% 15,382 0.15% 5CardDrawFL 43,448 50.27% 0.41% 38,356 44.38% 0.36% 4,569 5.29% 0.04% 49 0.06% 0.00% 86,422 0.82% 5CardDrawPL 4,513 14.76% 0.04% 17,279 56.52% 0.16% 8,727 28.55% 0.08% 52 0.17% 0.00% 30,571 0.29% 5Card7ADrawFL 5,512 26.98% 0.05% 13,085 64.05% 0.12% 1,831 8.96% 0.02% 0 0.00% 0.00% 20,428 0.19% 5Card7ADrawPL 12,348 45.89% 0.12% 14,561 54.11% 0.14% 0 0.00% 0.00% 0 0.00% 0.00% 26,909 0.25% TripleDrawLowball27NL 0 0.00% 0.00% 2,102 100.00% 0.02% 0 0.00% 0.00% 0 0.00% 0.00% 2,102 0.02% TripleDrawLowball27FL 8,738 39.34% 0.08% 9,613 43.28% 0.09% 3,059 13.77% 0.03% 799 3.60% 0.01% 22,209 0.21% TripleDrawLowball27PL 0 0.00% 0.00% 791 100.00% 0.01% 0 0.00% 0.00% 0 0.00% 0.00% 791 0.01% SingleLowball27NL 0 0.00% 0.00% 4,242 88.30% 0.04% 552 11.49% 0.01% 10 0.21% 0.00% 4,804 0.05% RazzFL 29,375 41.13% 0.28% 32,027 44.85% 0.30% 9,372 13.12% 0.09% 643 0.90% 0.01% 71,417 0.67% SokoFL 0 0.00% 0.00% 1,865 100.00% 0.02% 0 0.00% 0.00% 0 0.00% 0.00% 1,865 0.02% BadugiFL 0 0.00% 0.00% 18,395 87.19% 0.17% 2,339 11.09% 0.02% 363 1.72% 0.00% 21,097 0.20% HORSE/HEROSFL 8,589 42.04% 0.08% 10,676 52.25% 0.10% 1,127 5.52% 0.01% 40 0.20% 0.00% 20,432 0.19% HOSEFL 360 72.43% 0.00% 111 22.33% 0.00% 24 4.83% 0.00% 2 0.40% 0.00% 497 0.00% 8GameFL 0 0.00% 0.00% 19,189 82.39% 0.18% 2,881 12.37% 0.03% 1,221 5.24% 0.01% 23,291 0.22% OtherMixedGamesFL 1,480 71.64% 0.01% 574 27.78% 0.01% 10 0.48% 0.00% 2 0.10% 0.00% 2,066 0.02% SUMME 4,935,076 46.55% 4,513,244 42.57% 1,094,848 10.33% 57,745 0.54% 10,600,913 100.00% *NL=NoLimit,FL=FixedLimit,PL=PotLimit,ML=MixedLimit. NotethatthistablereferstothetotalsampleofplayersintheOPDUHH. Pokervariant*

30

You might also like