Game Theory (was: RE: Sprint peering policy)
Scott A Crosby
crosby at qwes.math.cmu.edu
Mon Jul 1 19:27:40 UTC 2002
On Mon, 1 Jul 2002, David Schwartz wrote:
> On Mon, 1 Jul 2002 12:11:46 -0500, Paul A Flores wrote:
>
> >Since it seems we are speaking of 'zero cost' interconnects, if Either X OR
> >Y feel like they are getting ripped, they won't (and shouldn't) do it. If
> >party X feels that party Y is gaining more from the interconnect that they
> >are, X might feel the need to lay some surcharges of some time on the
> >connection, which is only fair, if they feel they aren't receiving value for
> >value.
>
> Suppose that X is presented with a take it or leave it deal. Suppose further
> that X, on net, benefits from this deal. Why should they care how much or how
> little Y, Z, or T benefit from the deal? What kind of business sense does
> that make?
>
> >Otherwise, esp. now that enough people have gotten their hands caught in the
> >cookie jars, why would they GIVE away 'free' services for nothing in return?
>
> We're not talking about "nothing in return". We're talking about an
> arrangement between two parties that both benefit from. Why should one party
> care how much the other benefits? (Except, of course, as possible leverage to
> negotiate a better deal.)
>
What you're describing is a game theory problem.. The situation you
describe, where neither party charges for as long as both parties benefit,
this only occurs when the parties have no information on each other.
Lets say you have two parties, A&B, and zero transaction costs.
Furthermore, both parties wish to obtain the MAXIMUM benefit from the
deal.
A benefits $ 50
B benefits $100
Now, there are 4 situations:
1. A knows nothing about how much B will benefit and vice versa. In this
case, if either party tries to charge the other, they run a risk of having
the deal fall through, and getting no benefit. The optimal situation is
for neither party to request any payment from the other. (This situation
is theoretical, in real business, you could make guesses as to the benefit
obtained by another party.) Total benefit is $50 & $100
2. B knows that A will benefit by $50. A knows nothing. Then, B could
demand a charge of $49. A would choose to pay it, for they would obtain a
net benefit of $1. Note that in this case, B who's already benefitting the
most would get paid by A! Total benefit is $1 & $149
3. A knows that B will benefit by $100. B knows nothing. Then A could
demand a charge of $99. B would choose to pay it, for they would obtain a
net benefit of $1. Total benefit is $149 and $1.
4. A and B both know how much the other benefits. Then.. I'm not exactly
sure how game theory reasons this out. In reality, there is more than one
party, and imperfect information on respective benefits. I believe one
would use an economic and statistical argument, and the expected result
would be that B paid A $25.
> Actually, I think you can make peering fair in a much more simple way.
> Simply explain to people a sensible and rational way to evaluate their
> peering decisions. "Does it benefit me as much as or more than anyone else"
> is neither sensible not rational. "Does it benefit me more than it costs me?"
> and "Is it the best deal I can negotiate for this amount of benefit?", on the
> other hand, are sensible, rational, and fair.
Knowing how much the other benefits from a particular transaction can lead
to a remarkable difference in how valuable the transaction is to the
respective parties. And maximizing revenue is the purpose of business.
Scott
More information about the NANOG
mailing list