Program Trading, Arbitrage, Ect...

[mister ed]

	Quote:	HAL9000
	BTW, for a long time I worked on algos

Thanks HAL - good to have someone here who has worked on algos - can you offer any insights, information etc on them? What sort of inputs did you use, how many inputs, how are they structured, what sort of execution technology was used ... any info really would be great?

[darthtrader]

	Quote:	mister ed
	I think 'program' trading, in a narrow sense, refers to those arbitrage activities where stock bought/futures sold or stocks sold/futures bought, whereas there is also a broader sense in which the word is used to refer to other types of automated activity. I think the best word to encompass all the types of automated activity executed by computers/robots is algorithmic trading. Wikipedia, while not a definitive source, has a good selection of articles and links about algorithmic trading.

Hey Mr Ed, sorry about the terminolgy. I was pretty much refering to algorithmic trading when I mentioned program trading, the broader sense of the word. Maybe the arbitrage side though what we might be actually able to see with PREM. I'm not sure if you caught these statistics, it was in some link I found from what you posted. This was sopposed to be an average day the way the trading breaks down:
NYSE Closing Summary
Trade Date 8/16/2007
NYSE Trades ....8, 908, 020
NYSE Volume....2, 989, 836, 240
NYSE Dollar Volume ....115, 188, 118, 914
Retail Buy Trades ....152, 848
Retail Sell Trades ....204, 524
Total Retail Trades ....357, 372
Retail Buy Volume....53, 008, 213
Retail Sell Volume.... 76, 575, 492
Total Retail Volume....129, 583, 705
Program Trading
Index Arbitrage Program Trades ....576, 924
Non-Index Arbitrage Program Trades ....6, 231, 010
Total Program Trades....6, 807, 934
Index Arbitrage Program Volume ....111, 238, 364
Non-Index Arbitrage Program Volume ....1, 534, 370, 009
Total Program Volume....1, 645, 608, 37
.
Program trading accounted for 55% of total NYSE volume.
Retail trading accounted for 4% of total NYSE volume.
Program trading accounted for 76% of NYSE trades.
Retail trading accounted for 4% of NYSE trades.

The problem came though from the index arb sites levels. I plotted lines for what they gave as the threshold on the PREM for when there was a probability greater than zero of the arbitrage programs coming in to get cash back in line with the futures. The levels were not violated all day so indexarb was basically saying there should have been zero probability of arbitrage all day when those stats say that its normally half the volume. I'll try this again all next week, maybe I just messed something up yesterday.
What I mentioned about the breakout was certainly a big leap/speculation. The thing is though I have to question that if the programs are doing that much volume, that when you see price jump ultra fast that its the programs and not humans all getting in at the same exact time. If some of programs are so directly piped into the system they can actually execute orders in 10 miliseconds they would seem to me to be the obvious culprit. I'm not sure how useful that would be if true but it could be an interesting filter to stay out of bad trades if that indexarb site levels worked. If you knew there was a 50% chance of the arbitrage programs going off it wouldn't make much sense to step in front of that on the other side.
Check out that janowski.pdf if you have not, that presentation sounds incredible. Here is his version of the terminator script
"
How is it Possible to Compete with
the Speed of Automated Systems?
• The answer to this question is to NOT
compete with the black-box systems
(at least at their own game)
• The key is to eliminate the “lay-ups”
that traders give to automated systems
• Traders should conduct trades at
prices that are somewhere between
the desired buy & sell prices of such
systems (i.e. “fair value” trades)"

[darthtrader]

I've pretty much added PREM to my trading at this point. I've pretty much given up on getting any other information on programs outside this.
my observations so far are:
-IndexArb's fair value levels are great, its basically the zero point where cash and the futures are in line.

-IndexArbs program probability levels are totally useless. The programs come in alot more than at the levels they say on the site. I've still never seen PREM hit the levels they give as the lowest level let alone be even close to the upper/lower end of their probability level.

-If you plot the fair value level it makes alot of sense to get in on the futures in the direction of PREM and not so much sense to be on the otherside of PREM(at least on YM). If you get in on the other side your basically overpaying for the futures vs cash.

-I have yet to see the futures be over/undervalued vs cash, then the cash market move so fast in the opposite direction that it triggers a program.

-If you do the above, your still never going to know when cash/futures get out of wack to trigger a program. If you get in on the futures when they are undervalued though, maybe you get lucky and a program goes off in your direction. Its really interesting on a breakout at a key level because you can kind of see in price action when the programs move YM for 10-15 points, then you get the chasers who then give the futures a bit of follow through. If your in the trade and get a program on your side, its a good time to take some profit and sell to the chasers if it fits your time frame.

-Its especially good to watch this for trade management once your in a trade if your time horizon is low enough. If PREM drifts to the other side of your position and then spikes down huge it might be a good time to scratch if the program hasn't gone off yet. If it has your stop will probly already get hit before you can click the mouse anyway.

[MidKnight]

Darth, where are you getting your PREM data from. You are aware that not all PREMs are equal?

http://www.tradingmarkets.com/.site/...2005-46122.cfm

Here is a quote from the interview:
"
Dave: I know different data feeds use different symbols for the PREM. What are the 3 main symbols and their corresponding service provider?

Hank: All PREMs are calculated by the individual data vendors. Unlike stock quotes, the PREM is not calculated or fed to the data vendor by any of the exchanges. Most data vendors either do not do PREM calculations at all,or do a horrible job with the calculations. Only a handful of data vendors do the PREMs correctly and accurately.

With DTN, all of the PREM's are good. The PREM is SP-PREM.Z. The E-mini PREM is EM-PREM.Z. They also have the Nasdaq PREM. The symbol is ND-PREM.Z and NQ-PREM.Z for the mini contract. They recently began calculation for the Russell 2000 PREM. The symbol is RL-PREM.Z and RM-PREM.Z for the mini contract.

eSignal does a good job too, especially with the E-mini PREM. Their ticker symbol for PREM is PREM A0 (zero not O) and the E-mini PREM is EPREM A0. Most traders that have eSignal do not know that. If they look at the PREM at all, they look at $PREM. That symbol is not correct. It is a snapshot every 6 seconds and therefore is usually wrong. The same is true for $EPREM. It is almost never correct, since the ES contract is so active. eSignal does not have the Nasdaq or Russell PREM.

Comstock does a good job. Their symbol is PREM.X and EPREM.X. They do not have the other PREMs. Quote.com buys their data from Comstock. TradeStation used to buy their data from Comstock too, but now calculates all of their own PREMs. Their symbols are $SPINX and $ESINX, $NDIQX and $NQIQX, and $RLIUX and $ERSIUX.
"