Arbitrage must be the most abused word on this forum. I think that on average, I read at least a couple of postings containing the "a-word" every day and I cannot remember the last time when I found it was used appropriately.The word itself is absolutely fine, but since 99% of times those who use it have forgotten what they learnt from their 101 course on pricing theory, I would like to make a suggestion: let's ban it from Wilmott! Let's lobby the administrator to add an automatic script that replaces the word with "a*******e", so that when it is used incorrectly it won't confuse newcomers (arbitrage is very rare and not as ubiquitous in the market as the forums would suggest). Also seeing it replaced by "a*****e" may instill them some curiosity and some may go back on their books and discover for once what it really means.gc

i second the motion

The word is fine. And people do usually know what it means. So we won't be banning its use! But I would say that obsession with arbitrage is the problem, that all frameworks have to be no-arbitrage frameworks. I have lectured about this a lot lately. One problem is that it leads on to the abomination that is calibration! When people (young people, or naive people, or most authors of quant textbooks) make some statement on these forums about a model based on arbitrage I would like to see the more sensible members (and I include gc and daveangel in that category, for the moment!) politely knock some sense into them. I do it whenever I can, but I need back up!You do realise that an attack on arbitrage will equate to an attack on a certain type of quant, don't you?! This could lead to the First Quant War! But this might ultimately be beneficial to society!P

I don't have a problem with arbitrage per se .. I have a problem with the overuse of the term. as gc points out - the term is much abused.

QuoteAnd people do usually know what it meansMn... I think you are being generous with us readers of Wilmott.com. Later on I'll try to paste the link to a few threads, but it seems to me that the mainstream interpretation of a*******e in this forum is a synonim of "relatively cheap compared to something else that is even only remotely similar to what I'm looking at".QuoteWhen people (young people, or naive people, or most authors of quant textbooks) make some statement on these forums about a model based on arbitrage I would like to see the more sensible members (and I include gc and daveangel in that category, for the moment!) politely knock some sense into them. I'll start doing that now... I admire Martinghoul for having done that in the past, but the result wasn't very encouraging (in more than a situation I had the impression that the abuser felt he was unjustly reprimanded, without even considering that there may have been a point in the criticism)... Oh well, let's see...gc

It's easier to sell "arbitrage" to somebody than it is to sell "aggressive speculation".

QuoteOriginally posted by: PaulOne problem is that it leads on to the abomination that is calibration!Can you clarify what you have against calibration?If a model has meaningful parameters whose value is unknown by other means, it seems to me that calibration is a good way to get values for those parameters -- as long as you understand what the parameter is and how and why it may vary. I use calibration in this sense every day -- and a damn useful tool it is too.If you're simply against "curve-fitting" where people calibrate parameters whose meaning they don't understand, then I'm with you.

QuoteOriginally posted by: gcbut it seems to me that the mainstream interpretation of a*******e in this forum is a synonim of "relatively cheap compared to something else that is even only remotely similar to what I'm looking at".Hmmm... If the similarity is a probability relationship based on some model or other, do you object to the term "statistical arbitrage"?

Fermion, I am against the curve fitting. The parameter they are fitting may have a mathematical meaning but does not necessarily match anything that exists.P

QuoteOriginally posted by: PaulFermion, I am against the curve fitting. The parameter they are fitting may have a mathematical meaning but does not necessarily match anything that exists.PA problem, we always need to face when using data driven methods.They cannot generalize and you might have hidden influences.In principle, get data from models is structurally reliable. Take care with data-to-model creations.

In finance estimates of parameters does not usually relate to hypothetical models of distribution. That is the models are implied and parameters estimates actually do not in any degree justify models. And implied distribution is an element of the span. The left edge of the span could be identify as a someone's intuition and the right end is arbitrary imagination.