The Spam Experiment

Friday, January 07, 2005

Results from yesterday

Here is a little update on spam from yesterday:

My new domain just for testing spam: tuffwebhosting.com

Legitimate email: 0
Total Spams Received: 5
Total Viruses Received: 7
Total Suspected Spam: 0 (meaning messages that are probably spam but our software can't determine for sure.)

My control domain:
Legitimate email: 20
Total Spams Received: 104
Total Viruses Received: 0
Total Suspected Spam: 0

I still find it amazing that by only posting to a few newsgroups I'm getting up to 7 email viruses a day! I'm a bit disappointed that I am so far only receiving 5 spams per day from those newsgroups. I'm going to begin some other strategies for getting spam next week.

I'll keep you posted of my results!
- Ben
ben@benfitts.com

5 Comments:

  • This blog is awesome! If you get a chance you may want to visit this palm pilot software site, it's pretty awesome too!

    By Blogger tweedledeetweedledum, at October 1, 2005 at 11:43 AM  

  • One good resource for digital photo software and many more free software alternatives is Software4YourSuccess.com
    Yes it is my site and I would love for you to drop by for a second. From there you will have free access to several of my products such as The Marketing Toolbar (which is goldmine of information on how to do things quickly and on the cheap, thus saving you time and money).
    Also for webmasters I have created Webmaster Wizards, which will help you with almost every aspect of putting code on your site from things such as legal and privacy disclaimers to popunders. I have so much content and free software there I could fill up this whole blog, so check it out Thanks, Sincerely Rob Rudd

    By Anonymous Anonymous, at October 10, 2005 at 11:21 PM  

  • Incredible blog. I admired your site and I will be
    back once again to view it! I use much of my spare
    time searching for blogs like yours.
    Please go over my start a home based business on the internet blog.

    By Blogger Start Home Based Business, at February 15, 2006 at 9:07 AM  

  • 1. Introduction Two of the most common non-price mechanisms that allocate
    objects to individuals are auctions and lotteries. In auctions the probability
    that player i wins depends on the other bids, as well as the size of payments.
    In a lottery all agents have the same probability of wining the object, and the
    actions of the other players might affect the winning prize (for example, when
    there is more than one winner the winning prize will be divided equally) but do
    not affect the probability of winning. In this paper we conduct - both
    theoretical and empirical - analysis of a selling mechanism that combines
    elements of an auction and a lottery. The mechanism studied is used by the
    internet portal http://www.BestBidsAuction.com, which also provided data of its
    auctions. Before each auction, the auctioneer determines three parameters of the
    auction: the highest bidallowed (which is less than 10% of the retail value of
    the object), the maximum number of bids allowed before the auction closes, and
    the entry fee each bidder needs to pay when submitting his bid. All of these
    values are made public before the bidding starts. After the bidders pay the
    participation fee, they submit sealed bids, less than or equal to the highest
    bid allowed. The winning bid is the highest unique bid (in the sense that no one
    else bid exactly the same amount) among all bids received. The winner then pays
    his bid price and obtains the object. We call the selling mechanism adopted by
    the portal a Gambling Auction, because it has features that make it a
    combination of an auction and a lottery. First, the bid and the probability of
    winning are not monotonically related, because a lower bid might well win the
    auction if many bidders are placing high bids. Consequently there is no obvious
    bid that maximizes the probability of winning and, as we show, in equilibriumall
    bids provide the same probability of winning. Second, this mechanism is not a
    pure lottery either because the winning probability is determined by the action
    of the biddersand not by an exogenous randomizing device: the winner is the one
    that submits the highest unique bid. Note that, under the symmetric Nash
    equilibrium of the game, the equal winning probabilities this auction creates
    and the expected payments can be 2


    implemented using a lottery and thus the two types of mechanisms are outcome
    equivalent if the bidders are risk neutral and follow the symmetric
    equilibrium.2The theoretical analysis finds that in a symmetric equilibrium each
    bidder chooses his bid using a distribution function over a support that has no
    gap. This equilibrium strategy is increasing; namely the probability of placing
    a higher bid is not less than that of a lower bid. The intuition is that
    otherwise a higher bid would make winning more likely and thus be more
    profitable than a lower bid, which would makeeveryone prefer it, destroying the
    alleged equilibrium bidding pattern. We test this prediction with a novel data
    set collected from the portal http://www.BestBidsAuction.com, which implements auctions
    described above. The data confirms that the probability of a higher bid is not
    less than a lower bid. We also find that an increase in the number of bidders
    increases the number of bids for a given slot, although reduces the probability
    that each bidder places his bid at this given slot. This leads to an increase of
    the distance between the maximum bid allowed and the actual winning bid. We also
    tested the theoretical prediction that each bid has the same probability of
    winning by constructed a frequency table (Table 4). This table measures the
    frequencywith which the highest bid wins by calculating the number of auctions
    in which the highest bid won divided by the number of instances in which a
    highest bid was placed. We repeat this exercise at lower bid levels and ask
    whether the empirical frequencies are2This Gambling Auction is also interesting,
    because it can be used in countries or U.S. states that forbid gambling, because
    the rules of the mechanism do not meet the traditional definitions of lottery.
    The mechanism might attract people who like participating in gambling
    activities, since at a relatively low cost one have the opportunity to win a
    sizable prize. The auctioneer will make more money using this mechanism than by
    regular auction mechanisms if participants are risk lovers. Empirically, this is
    the case since these auctions have a negative expected profit for a bidder. This
    mechanism is similar to a rotating saving and credit associations (roscas) in
    which group of people save for indivisible good. Each period allthe people
    contribute to the rosca and it is given to someone randomly that is able to get
    the good. In thenext period it is given to somebody else and so on (see Besley,
    Coate and loury (1993)). In our mechanismthe good is also distributed eventually
    randomly and each individual pays the participation fees, but theexpected payoff
    is negative, since the auctioneer obtains a positive profit and the winner pays
    extra amount of money (the winning bid) in order to get the good. 3

    Page 4
    indeed equal as suggested by the theory. Some formal chi-square tests and
    informalanalysis suggest that the theoretical bid distribution is not consistent
    with the data. In addition, unlike other studies that estimated the demand for
    lottery games and found that consumers respond to the expected returns, we found
    that consumer demand for this lottery is not sensitive to the expected payoff
    but it is sensitive to the size of theprize. The paper is organized as follows.
    In the next section we characterize the equilibrium strategies of the auction
    game and provide some comparative static results. Section 3 describes the data,
    while Section 4 performs empirical analysis. A final section offers some
    concluding remarks. 2. Theoretical considerations We will first describe the
    model we consider and then show that in a symmetric equilibrium a higher bid is
    chosen with higher probability. There are kbidders3=3who all value the object at
    the retail price, v. After paying an entry fee of c each bidder submits a sealed
    bid that is less than a maximum value b << v. We assume that each bidder places
    only one bid. There is a minimum bid increment, which we normalize to 1. The
    winner is the one who placed the highest bid that was not bid by anyone else. If
    there is no such bid, then we assume that the seller runs the auctionagain with
    the same set of bidders. The internet portal reports that, in the rare event of
    no unique bid, the bidders will be notified about the situation and asked to
    submit a new bid without additional charge. The winner has to pay an amount
    equal to his bid, while the losers only pay the entry fee. In addition we assume
    that k, v and b are such that inequilibrium the winning bid is close to b; in
    other words, we assume that the bid increment is low compared to the value of
    the object, and thus the winning bid is close to 3In the auction at the above
    website only the maximum number of bidders is specified, but the number of
    actual bidders is usually close to the allowed maximum number of bidders, so one
    may assume that the number of bidders is a known constant, k.4

    Page 5
    the maximum allowed bid b.4Under such conditions we make the simplifying
    assumption that each bidder is interested in maximizing his probability of
    winning the object, ignoring the payment consequences of his bid.5The entry fee
    is already sunk at the bidding stage, so it does not affect bidding strategies.
    First, note that the above game has an equilibrium, since after imposing
    aminimum bid requirement of 0, the auction becomes a finite game. Moreover,
    using Kakutani�s fixed point theorem we may also show that a symmetric (mixed
    strategy) equilibrium exists. Claim 1: In any symmetric equilibrium there is no
    gap in the support of the equilibrium strategy. Proof: Suppose there was a gap
    at b�. Then bidding b� would strictly dominate bidding the next available bid
    b�-1, which yields a contradiction in that b�-1 is in the support of the
    equilibrium strategy. Note, that the above claim also implies that the high end
    of the support is the maximum allowed bid, b. Then a symmetric equilibrium is
    characterized by the number of bidsemployed, n, and the probabilities of each of
    those bids,)1Pr(+-=ibpiwhere i = 1,�,n. Theorem 1: In a symmetric equilibrium
    the probability of a higher bid is not less than a lower bid: i >j implies that
    pi= pj. Moreover, pi = pjcan hold only when there are four bidders. In that
    case, the unique equilibrium has p1= p2 = 1/2. Proof: See the appendix A. 4On
    average, the distance between the winning bids and the maximum allowed bid in
    our data is less than14 cents on average, and the maximum distance is less than
    $1.5. 5The bidder�s problem is to choose bithat will maximize: P(bi)(V- bi)-C=
    P(bi)(V-b+b-bi)-C= P(bi)(V-b)+P(bi)(b-bi)-C, where P(bi) is bidder i probability
    of winning the object when placing a bid of bi, V is the object valuation, b is
    the highest bid allowed and C is the participation cost. If all bidders follow a
    symmetric equilibrium, then the probability of receiving the object is the same
    for each bidder. Asmentioned before, the distance between the winning bids and
    the maximum allowed bid in our data is lessthan 14 cents on average, and the
    maximum distance is less than $1.5. So on average, when one maximizesthe
    probability of winning the object and ignores the second part of the objective
    function; one ignores a monetary incentive of only a few cents. If we drop this
    simplifying assumption then our results do not hold as stated. It is no longer
    necessarily true that the equilibrium does not have a gap, since the equilibrium
    weidentify in the simplified game is not robust to large deviations, when a
    bidder places a bid close to zero.However, since the largest admissible bid is
    less than 10% of the value of the object, the incentive for this deviation might
    be neglected in a first approach to model this game. This approach is also well
    supportedby the data, since winning with a very low bid is very unlikely, as it
    will be noted in the next section. 5

    Page 6
    The intuition behind these results is clear. Suppose, that the other bidders
    randomize equally among the bids B = {b1., b2, �, bn}, where b1> b2> � > bn.
    Then it is easy to see that if bidder i places the bid b1, then he has a higher
    probability of winning then with any other bid that belongs to B. But this
    yields a contradiction, because in a symmetric equilibrium bidder i use a mixed
    strategy with support on B, and thus he isindifferent between any of the bids
    belonging to B. The incentive to bid high iseliminated only if a bidder expects
    that there are more bidders who placed a high bid than who placed a lower one.
    Thus, in equilibrium each bidder must place a higher bid with higher
    probability. Let us consider some examples with a small number of bidders.
    First, if there are three bidders, then, in the unique equilibrium all the bids
    down to zero are used. With Tpossible bids including 0 it holds that for all 1 <
    i < T, pi= 1/ 2T-iand p0=1/ 2T-1is theunique symmetric equilibrium of the game.
    If k = 4, it is easy to show that the unique symmetric equilibrium is such that
    p1= p2= �. In the case when k = 5 an equilibrium is such that
    0.010}.p0.083,p0.197,p0.337,p0.372,{p54321=====We can confirm that it is indeed
    equilibrium. A bidder�s utility is his probability of winning plus the
    probability of a complete tie divided by five. Suppose that a bidder places the
    maximum allowed bid. A bidder wins in this case if no one else placed thisbid,
    i.e. with probability .)1(w411p-=A complete tie occurs, if one or two other
    bidders placed the highest bid and the other two or three placed the same bid,
    or if all others placed the highest bid. This probability is
    .)(p})(p)(p)(p){(p)(p6})(p)(p)(p){(p4pt4125242322213534333211++++++++=Since in
    equilibrium each bidder obtains a utility of 1/5 we obtain the following
    condition:.515w11=+t6

    One can compute the corresponding probabilities, wi, tifor i = 2,3,4,5 and write
    up thecondition that for all i:6.515wi=+itThen one obtains 5 equations in 5
    unknowns (the �s) and this system has a unique real valued solution, the vector
    stated above. Finally, one needs to check that by placing a lower bid than bid
    5, the achieved utility is not higher than 1/5. By placing such a bid the
    deviating bidder wins if and only if the other four bidders tied. Then the
    incentive constraint can be written as: ip.51622514=+???=jijiiipppThe proposed
    strategy profile satisfies these conditions and thus it is equilibrium.For k = 5
    the distribution of the winning bid is
    .}011.0,098.0,211.0,325.0,357.0{54321=====pppppFor k = 6, an equilibrium is
    0.109},p0.248,p0.309,p0.334,{p4321====and the distribution of the winning bid is
    }.122.0,247.0,303.0,329.0{4321====ppppFor k = 7 an equilibrium is
    0.078},45p0.296,{54321=0.137=,0.22=,0.26==pppp7and the distribution of the
    winning bid is .}084.0,137.0,219.0,272.0,287.0{54321=====pppppIt is apparent
    that the size of the support of the equilibrium strategy is not monotonic.
    Excluding the case of 3 bidders, which seems non-generic, one conjecture 6The
    corresponding probabilities for wiand tiare different for every i. In order to
    save space the complete set of equations is not reported here but it is
    available upon request from the authors. 7We did not show that the above
    equilibria are unique for a given k. For this one would need to show thatif one
    considers a different number of bids for a given k than the one considered
    above, then no solutionexists to the resulting system of incentive constraints.
    We only showed at this point that there are no other equilibria for k=4, 5, 6, 7
    when we consider up to 7 possible bids. Our conjecture is that these equilibria
    are unique in these cases and moreover, for any k there is a unique equilibrium
    of the game.7


    that emerges is that the more bidders there are the less concentrated become the
    equilibrium strategies. Although there is no monotonicity in the length of the
    support with respect the number of bidders, our conjecture is that the expected
    distance betweenthe maximum allowed bid and the winning bid (the Gap) increases
    with the number ofbidders. Namely, it is more likely that a bid further from the
    maximum becomes the winning bid when the number of bidders increases.
    Theoretically this is the case when the number of bidders is 4, 5, 6 or 7.83.
    The DataThis section describes the data. The data source is
    BestBidsAuction.com 9which is the Internet website of Best Bids Auction, a
    Arizona company that manages and implements private auctions designed to raise
    money for selected charities and member non-profit fundraising organizations.
    The internet auction process is a combination of a lottery and an auction.
    Before each auction, the auctioneer determines, among other things, the highest
    bid allowed and the maximum number of bids that will be accepted for the
    auction, and makes this information available for the bidders. In order to
    participate in an auction, bidders submit sealed bids, less than or equal to the
    highest bid allowed in US dollars and cents and agree to pay a bidding fee for
    each submitted bid. The auction is a sealed bid auction in the sense that when a
    bidder submits a bid he does not know what the other bids are until the auction
    is over. Each auction is closed when it receives the maximum number of bids or
    meets the other closing requirements.10After the auction closes, the participant
    that submitted the successful (winning) bid is determined. The successful bid
    8The expected Gap when k = 4 is 5.04=g, and for the other cases it is,
    078.1*4...*055515=++=ppg162.16=gand 458.17=gwhen the number of bidders are 5, 6,
    and 7 respectively. 9All the information has been taken from
    http://www.BestBidsAuction.com. 10An auction will remain open until either the maximum
    number of bids allocated for the auction is reached or the auction reaches
    maturation (63 days for auctions requiring less than 200 bids, and 183 days for
    auctions requiring 200 or more bids) and has received the minimum number of bids
    required to close. If the minimum number of bids has not been reached, the
    auction will be extended until the minimum numberof bids is met. At that time, a
    closing date of three days will be set and posted on the auction. 8

    Page 9
    is the highest unique bid out of all bids received in the auction.11Duplicate
    bids are used to calculate the number of bids required to close an auction but
    are disqualified from being selected as the successful bid. For example, if a
    single auction includes the following four bids: $69.42, $69.42, $48.69 and
    $65.44, the winner will be the one who submitted $65.44. In the very unlikely
    event that an auction closes and there is not a unique bid, all participants
    receive an e-mail describing the situation and are asked to submit a new bid
    without additional fees. Table 1 gives summary statistics from the different
    auctions that took place during 2003 and 2004. The information provided on the
    website includes all auctions that havebeen conducted in this period. The
    products auctioned were electronic appliances (computers, TV�s, video games etc)
    and gift cards (provided by Target, Shell, Wall-Mart, Starbucks etc). The mean
    retail value of the items auctioned was $414.169. The most expensive item
    auctioned was a Panasonic 42�� Plasma TV with a retail price of $4999, while the
    cheapest item was a Nintendo Game Boy with a retail price of $79.99. The Maximum
    Allowed Bid was almost always identical to the Maximum Submitted Bid, which
    means that in almost all the auctions the highest submitted bid was the highest
    allowed bid.12On average, the maximum allowed bid was 7.2% of the retail
    price,13and it had a mean of $30.83. The highest Maximum Allowed Bid, $624.38,
    occurred in the case of the Panasonic 42�� Plasma TV, while smallest Maximum
    Allowed Bid, $2.94, was in the case of a $100 Starbucks gift card. The average
    winning bid was $30.70, and it was, on average, 13.69 cents below the Maximum
    Allowed Bid (and the maximum submitted bid). We define Gap as thedifference
    between the maximum allowed bid and the winning bid. The minimum of this
    variable is 0, which mean that the maximum allowed bid was the winner. The
    maximum 11A unique bid is a bid that is not a duplicate bid. A "duplicate bid"
    is a bid submitted by a participant in anauction where another participant(s)
    has submitted a bid(s) for the identical amount.12There are 15 cases out of 310
    in which the highest submitted bid is less than the maximum allowed bid. In 10
    cases the difference is 1 cent. 13It seems that the auctioneer choose the
    Maximum Allowed Bid such that it will be, on average, less than 10% of the
    retail price. An OLS regression of the Maximum Allowed Bid on the retail price
    yield a coefficient of 0.072 with standard error of 0.0024 (t-value of 29.59)
    and R squared of 0.7398. It seems that the Maximum Allowed Bid is also
    positively correlated with the Number of Bids per auction. An OLSregression of
    the Maximum Allowed Bid on the Number of Bids yield a coefficient of 0.375 with
    standarderror of 0.020 (t-value of 18.37) and R squared of 0.5228. An OLS
    regression of the Maximum Allowed Bid on both the Retail Price and the Number of
    Bids per auction yields coefficients of 0.1765 on the retailprice and -0.6648 on
    the Number of Bids, both significant at 1% level. 9

    By Anonymous screensaver, at February 4, 2007 at 10:10 AM  

  • As a top-rated company in the world of ecommerce, Infyecommercesolution has carved out a niche for itself and with the ecommerce solution provided by the company receiving accolades from clients all over the world, it has, in the true sense of the word, grown up to be a top-notch outsourcing software development company. For details on all the services provided by the company, visit http://www.infyecommercesolution.com.

    By Anonymous ecommercewebmaster12, at November 5, 2008 at 11:16 PM  

Post a Comment

<< Home