[Trader's Handbook] Draft articles, "out of pocket" discussions - page 14

 
sanyooooook:
of course the swap is tomorrow and there will be no swap delivery after tomorrow

What prevents you from doing a swap the day after tomorrow instead of tomorrow?

Why make a swap both tomorrow and the day after tomorrow?

For example, the deal is made on the 2nd... so it's settled on the 4th. Why swap the night of 2nd/3rd and the 3rd-4th if the deal is settled only on the 4th?

 
hrenfx:

The point is that in practice (not among quantum theories) arbitrage is not free-risk. The first three points in describe insider trading rather than arbitrage. For example, front-running.

In most cases it is just execution risk.

Those definitions are given under the conditions of infinite divisibility of each unit of securities, infinite instant liquidity and no transaction costs. There are actually more conditions, but these are basic.

I also disagree about front-running. There are quite working variants of HFT which have these properties.

Practice shows that absolutely any arbitrage is statistical: EURUSD1 vs EURUSD2, EURUSD vs 6E (CME), EURUSD vs EURGBP * GBPUSD etc.

Are you saying that there is a risk of a divergence process between the two EURUSD variants? I couldn't agree more. This is true deterministic arbitrage, there is only execution risk.

Arbitrage is only possible in a cointegrated (with certain assumptions) portfolio. These assumptions allow for stronger terming of arbitrage as statistical, increasing risk and liquidity ceiling.

Co-integration is a linear relationship. By defining arbitrage in this way, you severely limit the class of strategies traded, because in real life arbitrage is quite possible between assets with non-linear relationships.

All in all, this is a kind of terminological dispute between algotraders and quantum theorists.

Imho, quite practical definitions, although I am not a theorist at all. Again, these are generally accepted definitions.

 
A100:

What prevents you from doing a swap the day after tomorrow instead of tomorrow?

Why make a swap both tomorrow and the day after tomorrow?

For example, the deal is made on the 2nd... so it's settled on the 4th. Why do we swap on the night of 2nd/3rd and the 3rd-4th if we only settle on the 4th?

so there is no settlement )

ZS: read from now till now

 
anonymous:

In most cases it is only an execution risk.

This phrase alone shows a theoretical approach to the case. In execution risk is sometimes the determining factor. Sometimes so much so that a simple portfolio EURUSD1 vs EURUSD2 can become knowingly loss-making.

Those definitions are given under conditions of infinite separability of each security unit, infinite instant liquidity and absence of transaction costs. Actually there are more conditions, but these are the main ones.

The accuracy and number of formulations are irrelevant. It is more of a quantum formalisation - the result of homing in on theories, building their clear definitions and axiomatics. Mathematics for the sake of mathematics. A kind of aesthetic pleasure derived from the orderliness and clarity of theory. In this case, it's all rubbish, regardless of regalia. The essence of arbitrage is simple.

I don't agree about front-running either. There are quite good variants of HFT which have such properties.

I don't understand what this is about at all, but it doesn't matter.

Are you saying that there is a risk of divergence of two variants of EURUSD? I couldn't agree more. This is a true deterministic arbitrage, there is only execution risk.

Deterministic, quasi-stationary and other heretical arbitrage have the same nature - cointegration. I can tell you that as a practitioner of the kind of deterministic arbitration and other perversions you call it.

Co-integration is a linear relationship. By defining arbitrage in this way, you severely limit the class of strategies traded, because in real life arbitrage is quite possible between assets with non-linear relationships.

Does a linear relationship apply to the same-quantized price BPs of different symbols? Or is it possible to apply linear methods to differently quantized TsVRs to find non-linear relationships using fast linear algorithms?

Imho, quite practical definitions, even though I'm not a theorist at all. Again, these are generally accepted definitions.

My misfortune (or happiness?) is that I am not at all familiar with mathematics, economic theories, trading books and the like. Deprived of education and an ex-soldier. So common definitions are nothing to me. I have a lot of practice and a developed practical approach in algotrading. An educated person talking to me is like talking to a monkey.
 
anonymous: ... Co-integration is a linear relationship. By defining arbitrage in this way you severely limit the class of strategies traded, because in real life arbitrage is quite possible between assets with non-linear relationships. ...
It looks like they are talking about something like
... Practice shows that absolutely any arbitrage is statistical: EURUSD1 vs EURUSD2, EURUSD vs 6E (CME), EURUSD vs EURGBP * GBPUSD etc. ...
I.e. arbitrage between derivatives and their underlying assets.
 
anonymous:

Co-integration is a linear relationship. By defining arbitrage in this way, you are severely limiting the class of strategies traded, because in real life arbitrage is quite possible between assets with non-linear relationships.

In fact it is not. What are the limitations? What prevents you from looking for cointegration in non-linear instrument transformations?
 
TheXpert:
In fact, there isn't. What are the limitations? What prevents us from looking for cointegration in non-linear instrument transformations?
That's a great idea. We'll find one. And then how do we trade them?)
 
TheXpert:
Essentially not. What are the limitations? What prevents looking for cointegration in non-linear instrument transformations?

There are no limitations, the concept of nonlinear cointegration is also quite well known and well developed.

The problem with non-linear methods (including nonlinear cointegration and neural networks) is that they fit the data too well, i.e. you will find correlations that are not actually there.

MetaDriver:
Great idea. We'll find them. Then how can we trade them?)

You can trade them (take and approximate the contractual function by a portfolio of options), but why complicate things so much?

 
Calculate the current derivative of the function (let it be equity of the options portfolio), compare it with the open position for linear instruments and make adjustments if there is a difference. Do not correct too frequently as transaction costs (spread) will eat up a lot. If the function is sufficiently wavy, then such manipulation will be reminiscent of the amusement of a young Martini/Antimartini player - which is despised by the whole progressive community.
 
hrenfx:

Sometimes you hear statements about a huge audience. I want to understand whether this is true or not. And who at least is not intimidated by the thoughtful reading of many letters on the logically obligatory topic of market literacy.

Depressing. Worse than the most pessimistic sentiment. Still, no matter how hard the developers tried to develop algotrading in the MQL community, it is not going anywhere. The community is almost entirely composed of clicker statists.

Those who liked it, are you yourself pleased that the level is so low?