I'm gonna throw my 2 cents in here...
Firstly, timeframes are product specific. On contracts where there is a large proportion of the participants looking at each transaction, in combination with the DOM, I don't think the term "noise" is really appropriate as it is in other products - I mean, on some contracts, the market is examined so closely by the participants that each transaction (or non transaction, equally important) has some level of significance - this is the letter of the words analogy NT describes...
However, in other markets, it simply isn't possible to monitor each and every transaction (in OTC markets for example); in other markets, it IS possible, but the significance of individual transactions isn't as great... say, equities (as an example... no absolutely 0 about the mechanics of equity markets so can;t say whether this is actually the case)...
.. It's quite simple. In markets where discrete transactions happen, and can be disemminated in real time, the significance of each transaction is determined by the participants of the market - in the Bund, say, spoofs, trades etc... are significant in the markets because there are enough participants paying attention to them - ergo, there's no such thing as "noise". In FX, say, no one gives a sh!t, because the participants attribute little value to granular information ("noise").
The timeframe point is rather a moot one... in a market where each new transaction is important, a specific level on information can be collected in a much shorter time than one where it is only the aggregated information that is important. You can't say 5 min on the bund isn't important, if the participants are all watching it (because, with a constant stream of information that has some value, 5 mins is alot; in contrast to FX, where little value is attributed to each new transaction, it is only when transactions are aggregated that useful information can be determined).
________
As for algo trading.... It is unlikely that computers will take over the world. at the end of the day, a computer can only do what a human has told it to. Human's are a critical part of the process...
As Goose says, most of these algo's will be to take advantage of very small price discrepancies; the algo will not say to itself... "right, crudes about to break north, and the index futures are all approaching resistance... I'm gonna start bidding for the 10yrs and try to catch some of the flight to quality" (or whatever). They cant think, they can only excecute (consider all the other applications that Tomorrows World said computers would do for us - 20yrs later, and I still have to pay the cleaner to do the hoovering properly
).
That is not to say, though, that computers won;t change te way the markets are played... even now, with autospreaders and such, individual traders are getting computers to excecute increasingly complicated tasks on their behalf - but the comp still needs to be told what to do, it cant figure it out itself.
in short: If there is any discretion whatsoever in the approach to trading, humans are critical to doing it.
Firstly, timeframes are product specific. On contracts where there is a large proportion of the participants looking at each transaction, in combination with the DOM, I don't think the term "noise" is really appropriate as it is in other products - I mean, on some contracts, the market is examined so closely by the participants that each transaction (or non transaction, equally important) has some level of significance - this is the letter of the words analogy NT describes...
However, in other markets, it simply isn't possible to monitor each and every transaction (in OTC markets for example); in other markets, it IS possible, but the significance of individual transactions isn't as great... say, equities (as an example... no absolutely 0 about the mechanics of equity markets so can;t say whether this is actually the case)...
.. It's quite simple. In markets where discrete transactions happen, and can be disemminated in real time, the significance of each transaction is determined by the participants of the market - in the Bund, say, spoofs, trades etc... are significant in the markets because there are enough participants paying attention to them - ergo, there's no such thing as "noise". In FX, say, no one gives a sh!t, because the participants attribute little value to granular information ("noise").
The timeframe point is rather a moot one... in a market where each new transaction is important, a specific level on information can be collected in a much shorter time than one where it is only the aggregated information that is important. You can't say 5 min on the bund isn't important, if the participants are all watching it (because, with a constant stream of information that has some value, 5 mins is alot; in contrast to FX, where little value is attributed to each new transaction, it is only when transactions are aggregated that useful information can be determined).
________
As for algo trading.... It is unlikely that computers will take over the world. at the end of the day, a computer can only do what a human has told it to. Human's are a critical part of the process...
As Goose says, most of these algo's will be to take advantage of very small price discrepancies; the algo will not say to itself... "right, crudes about to break north, and the index futures are all approaching resistance... I'm gonna start bidding for the 10yrs and try to catch some of the flight to quality" (or whatever). They cant think, they can only excecute (consider all the other applications that Tomorrows World said computers would do for us - 20yrs later, and I still have to pay the cleaner to do the hoovering properly
That is not to say, though, that computers won;t change te way the markets are played... even now, with autospreaders and such, individual traders are getting computers to excecute increasingly complicated tasks on their behalf - but the comp still needs to be told what to do, it cant figure it out itself.
in short: If there is any discretion whatsoever in the approach to trading, humans are critical to doing it.