Arbitrage

[barjon]

Hi Joseph, thanks for your suggestion. I'm up to speed on what could constitute an arbitrage trade - any number of instruments as per Barjon's post above. What I'm really asking for is an actual example of an arbitrage trade. Not live of course, just one that has been done or even a hypothetical arb trade.

Just like to get my head around the mechanics of it and being a bit on the dim side find a worked example fills in the gaps that technical definitions often leave in my comprehension any process.

ok, in the very unlikely event that you see HSBC trading at 668 in London and 672 in New York (all in £ and forget the exchange rate element and the spread for the purposes of the example). The price should be the same so you expect that 4 point difference to disappear. Therefore you go:

long HSBC London @ 668
short HSBC New York @ 672

When they iron out the "inefficiency" and both trade the same - you pick a price but let's say HSBC is trading at 675 when that happens - the result will be:

long HSBC London = +7
short HSBC New York = - 3
= +4 net

 

[Purple Brain]

It's as simple as that? That's a big relief.

I'm surprised it happens at all - let alone infrequently.

I assume with the more exotic and complex combinations of the thousands of asset classes and instruments available within all of these classes it will occasionally provide miniscule profits for fractions of a second before being rectified which is where the algos come in on massive size I guess. But even with low latency and super-computing power, they must still face the risk of being out-gunned by other algos and ending up with losses.

 

[barjon]

It's as simple as that? That's a big relief.

I'm surprised it happens at all - let alone infrequently.

I assume with the more exotic and complex combinations of the thousands of asset classes and instruments available within all of these classes it will occasionally provide miniscule profits for fractions of a second before being rectified which is where the algos come in on massive size I guess. But even with low latency and super-computing power, they must still face the risk of being out-gunned by other algos and ending up with losses.

aye, don't see anything as obvious as that bit more complex with similar and related instruments, but doing pretty much the same thing (as scholfield was talking about between SBs earlier, for example)

 

[teflon142]

There are some ideas being thrown about here that aren't quite accurate. I suggested that someone googled "Arbitrage", and someone else quoted the investopedia definition, which isn't a very good one.

Arbitrage == Risk Free profit.

By trading financial instruments, arbitrage can occur when two EQUAL AND OPPOSITE positions can be made, simultaneously, resulting in a positive cash flow. They do not have to be the same instrument, but they do have to be FUNGIBLE.

In the SB example being given here, the instruments being traded ARE NOT FUNGIBLE, so there is NO ARBITRAGE. Going long cable with spreadbetter A and short with spreadbetter B is not an arbitrage - you have to finance both positions, and you can't settle one position with the other (i.e. get shop A to settle the trade with shop B). You also have counterparty risk, and all sorts of other risks.

An example of Arbitrage:

A bond is offered at $140.
The Yield on this bond is 4%*
90 day rates are at 3%*
A Future of which this bond is deliverable is bid at $140

Simultaneously, you:

Buy the bond, sell the future:

-$140 + $140 = $0

Pay 3% on the $140 you borrowed = -$4.2
Recieve 4% on the bond you own = +$5.6

Then at expiry, you deliver the bond at $140

Leaving you with a risk-free profit of $1.4

This is an overly-simplified example of arb, but it is a TRUE arb. The example of HSBC stock is along the right lines, except that for the trade to be RISK FREE, the two stocks would have to be fungible (some are), and instead of "waiting for the prices to come back in line", you would just send the stock you bought to the person you sold it to, and settle the trade that way. Waiting for the prices to come back into line is NOT arbitrage, because there is no obligation for them to do so.

* overly simplified to illustrate the principal.

 

[Purple Brain]

Many thanks for your clear and simple explanation teflon142. Fungibility and Risk Free appear to be the other key ingredients which constitute a genuine arbitrage.

 

[teflon142]

Many thanks for your clear and simple explanation teflon142. Fungibility and Risk Free appear to be the other key ingredients which constitute a genuine arbitrage.

No sweat bud. For a quick and dirty litmus test, to make a true arb, you need to

1) Do all the trades simultaneously, and
2) Not rely on anything else happening, however likely it may be.

There are probably exceptions that prove the rule, mind.

 

[Purple Brain]

1) Do all the trades simultaneously, and
2) Not rely on anything else happening, however likely it may be.

So in an earlier post where I said:

But even with low latency and super-computing power, they must still face the risk of being out-gunned by other algos and ending up with losses.

was a complete nonsense as algos or not - they get all their ducks lined up and hit it in one go.

Or is there still a possibility that between them buying the bond and selling the future (getting both legs on simultaneously, but not quite) that one or both of those markets move such as to nullify the differential and they do indeed get caught with their pants down - holding one or both of those positions at a potential loss and they have to be physically managed out?

 

[teflon142]

...
Or is there still a possibility that between them buying the bond and selling the future (getting both legs on simultaneously, but not quite) that one or both of those markets move such as to nullify the differential and they do indeed get caught with their pants down - holding one or both of those positions at a potential loss and they have to be physically managed out?

Yup.