TOP STORIES Algorithmic Traders Are In The Money 29 Mar 2006 U.S. algorithmic traders have seen a massive increase in pay over the past year, driven by talent shortages and increased demand for their expertise â particularly from buyside firms, suggest news reports. According to this weekâs Wall Street Letter, pay for U.S. algorithmic executives has risen from $300,000 to between $400,000 and $800,000 over the past year, and is set to continue rising as brokerages and buyside firms pile into the area. For some, bonus guarantees could push total compensation to over $1 million. In response, sell-side firms are raising compensation guarantees. U.S. recruiters estimate algorithmic trading desks could see staff numbers rise 20 percent by the end of 2006, the article says.
The figures looks about right, but I think those are still figures from '05. The new hires in algorithmic trading in '06, at least the numbers I have heard, are on the higher end, between 650-900 or so (for a SVP or director level person). For a VP level algorithmic trading person, the numbers are around 300-500. The big push, will come from the buy-side, as sell side are only increasing staff, not creating new desks (outside of a few firms setting up automated MM desks, etc).
what exactly happens on a buy side algo desk, how is this different from the classic index arb program trading ... do they mean correlation driven cross-mkt arb strategies ... mathy high freq intraday directional stuff ... what precisely IS algorithmic trading the timing is kind of funny as i finally gave up my intraday systematic trading. the competition is really really good. too good to fight for ticks any more with any size (personally .. or at least given my current knowledge). ahh, just read rufus. mm stuff. gotcha if anyone is bored some time, just experiment with joining the bid on EC, ZG etc in the middle of the night when it's just you and the bot. compete with it, without actually touching the ask. you can get a pretty clear sense of how some of these bots are working, or at least the superficial aspects. they'll repeat the same negotiation 10 times in a row
Hmm, looks like good timing for me, I'm just now to the point where I'm very very happy with my intraday high-frequency stuff. The competition is definately getting more stiff. I've been reading a lot lately about topology (the mathematical kind...), fractals, dynamical systems etc and suddenly everything kind of 'snapped' into mind and it seems so clear now and also gives me some theoretcal justification for why my shit works at all. But of course, there are others who know this stuff, and the knowledge will eventually spread. So, with all this advancement in high frequency algos, high-dimensional math, sophisicated data mining techniques etc, what is the end result? Will it be a never-ending progression towards more efficient/sophiscated markets? Will the bots eventually get so damn good that all of the human traders eventually give up, or will it simply shift traders to positions where they make more high-level decisions and bots carry out them? Seeing all the things that are possible/explained by very sophiscated models shows me how absolutely hopeless it is for most humans to truly be successful and might also show why the "99%" (who knows the real number) of traders fail. Basically I guess, is there some perfect market that could ever be reached?
Wont all these algo bots just cannibalize eachother? I dont know, my simple mind thinks the cost associated with building and maintaining these algo's will have ever increasing diminishing returns as they take up most of the volume. You'll then have teams that try to reverse engineer the programs and game them in effect you'll just have completely random market behavior. At the peak of the use of algorithmic trading you'll have extremely low volatility at worst but it should eventually pick back up as a lot of these programs are washed out due to costs not being worth it. Thats my theory. Just another market cycle.
That's an interesting point. algos can come in an infinite variety, so two algos trade against each other and they can both be winners according to their own exposure to the market, it is not a zero sum game. Modelling and predicting reality inevitably creates a feedback loop and thus in some very complicated way you change the thing you are predicting so that you are always chasing your tail the whole time. Not to say you aren't making money while chasing it. At least that is my hope, and belief that it will remain so... I have no way to prove it, but I believe that algorithms/models trading against each other will actually create higher level patterns that bots/algos will trade on, which will create more pattern.. for an indefinite amount of time. Also, your comment about costs etc, I think that is true, I dunno if huge investment banks care about the costs of the investment they'd have to make to just chase pennies all day. That also gives me a hope that I am in a niche area. http://en.wikipedia.org/wiki/Turtles_all_the_way_down
it's funny you mention feedback, because that's the exact metaphor i've been pondering wrt the credit creation. no army of quants can program a fancy high freq mm bot to realize that it's hastening the collapse of the money it's predicated on. seems like best they can do is slow it to a brittle crawl could be a relatively short honeymoon for bots that absorb liquidity too well
Well, that's kind of the whole philosophical/metaphysical thing I was thinking of. Are they really hastening the collapse, or does the very act of modelling/prediction help to create structure or take it away? I think the answer is really we are creating value, structure and complexity, and adding order to the universe, and financial markets being an abstraction that we've created within this universe. Many physicists wonder how order can arise from thermal equilibrium that everything "should" be moving towards, but I think it's pretty clear, given chaos theory,etc that some forms of complexity/intelligence do arise from pure noise/primordial soup/planet formation/evolution. I'm not so sure that the bots will kill themselves, I'm pretty certain they'll be pretty advanced, and maybe even some guys working on financial predictions will be the first to create truly 'intelligent' machines, maybe by accidently create some form of universal learning/reasoning machine. But, this is getting way off track, back to the task at hand.
My systems are getting more and more sophisticated, but at the same time, [unintuitively] I think it is my discretionary trading that will show the biggest gains as more and more programs enter the market. I love paradoxes nitro