Game theory in arbitrage

I'm not an expert on this, but since artibrage is based mainly on exploiting mispricings, I would say that it's likely that they don't very much. Of course they need an active market of sufficient size and liquidity to make it worth doing.
 
my idea is this: hedge funds and institutions all have computer programs that kick in as soon as an arbitrage opportunity occurs. With the world "globalizing" the investor "universe" is expanding too. Years ago I am sure inv. banks and institutions probably knew all the players in the market, but now noone knows how many players there are. With the futures market for example, if arbitrage opportunities present themselves no one knows how many people have systems that will kick in to exploit that opportunity. so what if too many entities are following a stock and its respective future and an opportunity presents itself? seems to me that if too many people start shorting something the effect they desired might be overdone and create an arbitrage opportunity the opposite way. how do the players account for the "n" number of players in the game?
 
... seems to me that if too many people start shorting something the effect they desired might be overdone and create an arbitrage opportunity the opposite way. how do the players account for the "n" number of players in the game?

Apparently they don't based on all the algo funds hitting the skids last summer because they were all in heavily correlated trades that weren't supposed to have been.
 
adding to my statement above here is my point (and this is a real example I saw a few nights ago while observing a stock and its respective future afterhours):

Note: my biggest assumption is that the futures listed on onechicago do not trade after hours.

to keep things simple let's say stock X closes at $20.03 futures for stock X closes at $20.00
a few mins before the close individual Z along with all the institutions and hedge funds could see this opportunity. Institution programs kick in. They short the stock and long the future.

Individual Z participates hoping that with a very liquid stock there are going to be too many computers thinking the same thing without much regard for anyone else. Individual Z shorts the future (but does not short enough to move the market) right before close. After hours, there are so many institutions that shorted the stock that the market is flooded. Afterhours stock X goes down to $19.85. Individual Z buys the stock and pockets the spread.

The only way I see this opportunity appear is if Institutions don't regulate their arbitrage computers and don't take into account the number of players out there.... even if they try to, how can they be very accurate with their guess?
 
CAL,

First off, I don’t think the number of players is strictly relevant – as long as your order is entered prior to supply/demand being exhausted.

“stock X closes at $20.03 futures for stock X closes at $20.00.” Is this the right way round? Futures should (but not always trade at a premium. In this example, it may be the futures are discounting a forthcoming dividend, or the stock is ex-dividend prior to the futures expiry.

“ the effect they desired might be overdone and create an arbitrage opportunity the opposite way.” I suppose this is possible but I would suspect both sides of the trades would be entered as contingent orders, ie if one leg can’t be filled, don’t fill the other. Legging in/out is very risky.

“After hours stock X goes down to $19.85”. This is not strictly arbitrage, ie simultaneous and risk-free. In totality, and in the final analysis, this would be legging in/out of two bets (two positions) on two directions, with two possible outcomes each – not simultaneous nor risk free.

Grant.
 
your example needs to consider financing costs for the trade example. also remember that these are not PURE arbitrage as there are inherent risks involved in the trades. most stat arb/systematic funds pursue mean reversion strategys that are not really arbitrage and focus on techincals and historical data.

other participants can make a difference and actually drive derivative/cash markets, making the arb opportunity larger. however, the size of their trades matters more than the number of traders itself. for e.g. there is an increasing disparity between the cash bond and CDS market at the moment and many, many hedge funds have piled into what is called a 'basis trade'. they saw a very healthy spread that historically has never really been seen and put lots of money at work without really considering all the risks. well guess what, the spread actually got even healthier. lots of funds got stopped out, some might have even got wiped out due to effects of mark-to-market even though the trade remains potentially very profitable.

i think you make an interesting and valid point that too much money and participants are piling into what are essentially the same strategies, making the trades very crowded. Not really sure how game theory can really help in this to be honest because its the 'size' of the participants rather than the number 'n' that counts in the hedge fund game. Just my 2c.
 
Think you may be talking a touch at cross purposes here John. If I understand correctly, the original post pertains to very short term risk free arbitrage, basically (allowing for the fact that there is some basis risk in doing the cash / futures equity trade of course) whereas what you are referring to, if I'm not mistaken, is larger scale, longer term statistical arb? (i.e. what did for LTCM in 98).

Forgive me if I'm wrong, but I reckon your post was a tad off topic.

You're right in hindsight, of course, but at the time I posted that it wasn't clear (to me anyway) that the specific arbitrage in question was the short-term type.
 
It occurs to me that a fairly basic answer to the general question of whether you could fade the arbs playing a price misalignment is no on the basis of the idea that the arbs would never go that far. They are trying to grab quick profits. As the spread narrowed there would be less profit to the point of not making it worthwhile - presumably at some point ahead of that spread moving back to zero, and certainly before it inverted.
 
a couple of notes from previous postings. someone mentioned that it is size rather then the number. i believe there is some regulation as to how large your position can be so i would assume that with an extremely liquid stock there would be a few players who would push that limit (of course they would never show it).

the way i originally pictured the relationship in arb scenario i first described would be like having 2 lines on a graph with one being the sine and the other being the cosine. they meet at certain points but would be driven apart only to be brought back together. ideally the smaller the disparity between the peaks and the troughs the better, but with an overabundance of players the result would push the troughs and peaks further apart.

in theory this "arb" strategy would cancel the forces that drive the lines back together... essentially changing the cosine and sine into parrallel lines starting at the point the strategy was enacted.
 
Just a few point touching on what’s been said above.

The situation as portrayed boils down to “spread trading” or “correlation trading”, which is also how I see “statistical arbitrage” , ie the arbitrage is based on expectation of (ft338’s) “historical data” repeating and/or enduring (LTCM’s downfall).

Further, if this was an opportunity, it would be short-lived (maybe a one-off?) - arbitraged away by efficiency, or taking John’s “less profit to the point of not making it worthwhile” to it's ultimate conclusion.

Grant.
 
Cal,

Isn't the "sine" and "cosine" more common in a calculus context (beyond me) relating to derivatives? If so, why are these derived values the basis for the arbitrage rather than the prices themselves which would imply the derivatives.

If this is totally wrong, perhaps you could clarify the significance of sine and cosine.

Grant.
 
yeah spread trading can be seen as stat arb or relative value. basically pretty much what ltcm or jwm were doing. some of it can be thought of as 'pure arb' like on the run vs. off the run etc. but it still has financing and mark-to-market risk.

calisig your example of the sine/cosine graphs are a very good way to look at it graphically.

what i think you are basically talking about cash vs. derivatives trades or basically basis trades which are very very common in the market on the equities and particularly on the govt. bond side as the latter is an exceptionally liquid market. the relationship between cash and futures does move around between inception and maturity and you get it at wide points and basically wait until the spread collapses or maturity (when it definately collapses).

problem lies is the spread not being worth it as you gota leverage it up significantly to make it worthwhile. and this leverage brings with it mark-to-market risk of you being closed out of your position because the spread has widened further. and also costs are important as they have to be lower than the potential profit.
 
"problem lies is the spread not being worth it as you gota leverage it up significantly to make it worthwhile" which is what LTCM did - gearing of 30+.

Grant.
 
indeed and thats what made them money and then b*tchslapped them in the end. what hurt them the most was their reluctance to close out the trade when it went against them. normally you would be forced to close it out by your PBs when margin runs out but they didnt.
 
the street had the opposite trades on though once they got wind of what was happening at ltcm right?

will be interesting to see what happens with jwm after their much publicised losses in earlier this year...
 
GJ,

'didn't scale back, same positions with less margin, effective increase in gearing.' This is beginners' first week trading, not PhD's and Laureates. How the mighty fall.

Slightly off-topic but some parallels (don’t ask me what they are).

This could also be an illustration of why star traders eventually blow-up. A simple example, I trade 5 lots on £5000 margin. My risk (ignoring the theoretically unlimited) could be £1000 per contract. Assume I can make £1000 per contract , therefore my total return is £5000

Now I‘ve doubled my capital to £10,000 and my size doubles to 10 lots. But my risk is still the same - £1000 per contract, and my potential return is still £1000 per contract (for a total return of £10,000). Surely it would be more sensible to reduce risk but still deriving a greater potential return, eg 7 lots on £10,000 or 0.7 contracts per £1000. If potential profit is still £1000 per contract then my total return would be greater at £7000. Percentage wise it’s worse, but I don’t think this is the point here – it is the greater return with less risk. Can I have a PhD now?

FT,

Is all this covered in the book When Genius Failed (or something like that)? I remember the Metalgesellschaft debacle – that was a good story (Try Google if not familiar with it).

Grant.
 
grantx, that made me laugh though unfortunately phds are not given for common sense.

yeah i first read about it in the book but learnt more about it through work. also was able to compare that debacle with mini blowups this time around (not peloton) which was interesting. was telling my boss that we should write the next book and make some money back!

il have a look into 'Metalgesellschaft debacle' - sounds interesting.
 
Last edited:
FT,

"telling my boss that we should write the next book and make some money back!". Excellent. I bet his sense of humour suddenly disappeared: "Good idea, FT. Just jot a few ideas on the back this P45".

Grant.
 
Top