THE PENN-LEHMAN AUTOMATED TRADING PROJECT

PENN-LEHMAN LIVE DATA COMPETITION PLOTS
UPDATED LIVE
DECEMBER 18, 2002
MARKET CLOSED; PLOTS ARE FINAL

COMMENTARY: It was a wild ride at IRCS today... The stock took a beating early in the day, possibly on news of W's steely-eyed remarks towards Iraq. (Don't you wish you'd coded up that strategy that scraped the NY Times web site for breaking world events news?) The MSFT fall took many a strategy deep under water, and by mid-day things were looking depressing indeed. But volatility returned in the afternoon, and a late rally to close the market brought several of you back into the black after a long visit to the basement.

There were a number of dramatic reversals of fortune compared to the historical competition, but also some consistency worth noting as well. In particular, of the 5 strategies that use order book data in some way (recall 4 of these were the top finishers in the historical), 4 were profitable for the day, one was the highest earner overall, and the losses of the one ending in the red were modest. There seems to be something to investigate here, and we'll be doing it next term.

Overall, 7 of the 11 strategies fielded finished the day in the black, and 4 lost money. There were four strategies that managed to end in the black on all three historical days as well as the live day, all of them in the red pool. Also, this time no strategy violated the 100,000 share position limit. Share position limit saturations were better than the historical competition as well, with all but 2 teams hitting at least half the limit at some point during the day.

At some point we will provide additional analysis and data of interest regarding the live competition, including an analysis of how much client-client versus client-Island trading took place.

The two tables below summarize the final P&L and maximum share position (in absolute value), separated by pool. Please note that the values in the tables below are official, and may show slight discrepancies with the final values on the plots that follow. This is due to some plots not incorporating the final couple of ticks of the day.

Following the tables are the P&L, share and MSFT price plots for each team, along with some commentary on each strategy's behavior over the day.

RED POOL, FINAL RESULTS:

RED POOL RANK	TEAM MEMBERS	P&L (\|MAX SHARES\|), in thousands	DESCRIPTION	HISTORICAL RANK
1	Lao, Leung Wong	29.6 (72)	Case-based learning on top of basic SOBI	3
2	Angel, Hausman	15.0 (55)	Geometric trend lines	9
3	Mahesh	6.1 (72)	Standard deviation SOBI	1
4	Kadakia, Vora	1.9 (25)	Volatility-based MM	7
5	Morgovsky, Sharma	-32.5 (50)	Moving average trends	5

BLUE POOL, FINAL RESULTS:

BLUE POOL RANK	TEAM MEMBERS	P&L (\|MAX SHARES\|), in thousands	DESCRIPTION	HISTORICAL RANK
1	Yu (UT Austin)	24.8 (70)	Simple trend prediction	8
2	Yu, Zhang	5.8 (62)	Long-short with SOBI	4
3	Chi, Koltunov, Luss, Trinh	4.6 (85)	Volatility-based SOBI	10
4	Lowenstein, Rowin, Schwartz, Skolnick	-5.5 (65)	SOBI-moderated MM	2
5	Chavez	-13.7 (80)	Crossover moving average	11
6	Jawaid, Shah	-15.9 (33)	Breakouts & moving avg	6

PLOTS AND COMMENTARY FOR EACH TEAM:

RED POOL:

Mahesh (red pool) [PDF]: (Final P&L 6.1K, Red rank 3) The winner of the historical competition gets hammered early by the MSFT fall, but holds a long position through the day and hangs on to catch the late rally back into the black. One of four strategies that were profitable on all three historical days and the live day. A strategy meriting further investigation and refinement.

Lao, Leung, Wong (red pool) [PDF]: (Final P&L 29.6K, Red rank 1) Also took a long position and held it for the day --- but cleverly only got really long after the AM tumble. Another basic SOBI enhancement. Another of the four strategies to have 4 profitable days. Highest overall P&L for the two pools combined.

Morgovsky, Sharma (red pool) [PDF]: (Final P&L -32.4K, Red rank 5) A highly volative and active strategy, one of the few to take a consistently short position... but alas, mainly in time for the afternoon rally.

Kadakia, Vora (red) pool [PDF] : (Final P&L 1.9K, Red rank 4) For much of the early afternoon, this market-maker was the only strategy in the black, but got exposed on the sell side when the late rally hit. Still made money, though. One of the four to be profitable all 4 competition days.

Angel, Hausman (red pool) [PDF] : (Final P&L 15.0K, Red rank 2) One of the four to have 4 profitable days... but looks like they fixed their saturation issues since the historical competition, and it paid off. One of the four to have 4 profitable days.

BLUE POOL:

Lowenstein, Rowin, Schwartz, Skolnick (blue pool) [PDF] : (Final P&L -5.5K, Blue rank 4) Got very long early and never seemed to wake up again. Late rally helped, but not enough.

Yu, Zhang (blue pool) [PDF] : (Final P&L 5.8K, Blue rank 2) Saved by the late rally after spending the day long.

Jawaid, Shah (blue pool) [PDF] : (Final P&L -15.9K, Blue rank 6) One of the few to take a short position, but unfortunately in time for the late rally.

Yu, UT Austin (blue pool) [PDF] : (Final P&L 24.8K, Blue rank 1) They like to keep things simple down in the Lone Star state, and on a day like this, who can argue? This simple contrarian predictive trend strategy (if the stock has gone up recently, sell; if down, buy) started going long at the bottom of the market and benefited more than any from the late rally.

Chi, Koltunov, Luss, Trinh (blue pool) [PDF] : (Final P&L 4.6K, Blue rank 3) With greatly improved saturation, this strategy went long enough to ride the rally into the black.

Chavez (blue pool) [PDF] : (Final P&L -13.7K, Blue rank 5) For hours this strategy executed no trades at all, and seems to have made only 7 rather large trades in total. Thought-provoking, as all other strategies placed much more frequent orders. Timing was not good enough to take the rally into the black, though.