Trade based performance measurement

[gutenberg]

I have some data on individual trades of 4 traders from a HF. The products are mostly long/short equities and some equity options.

We are familiar with return based performance measures. Do you know of any interesting trade based performance measures? Are there any papers on this? Is there anything interesting one can do with individual trade data, that they would not be able to do with return data?



Thanks,



G

[Patrik]

Since the causality is that trades generate returns, trade data is a richer set of information than returns data. Looking at the trades themeselves it should be easier to understand what sort of trading you're dealing with. Before thinking about measures of performance I'd try to form a picture if I fundamentally think that what these traders are doing makes sense. Once you've made up your mind on that point you should have a reasonable view of what sort of strategies you're dealing with and this should help guide your decision around what perfmance measures you feel are the most important.

[Nonius]

G,



what do you mean by "trade"?  most likely it is a natural grouping of contemporaneous buys and sells of assets and derivatives.  most people I've talked to do have this natural grouping which is a) given a theme/strategy ID and b) the risk capital of the trade is then computed using either VaR, a trader/PM/RM constructed "worst case" loss, a stress test.  the risk capital in some, but not all shops, is computed on a marginal basis.  Then you could look at the return on this risk capital; sort of a Sharpe but accounting for makeup of current portfolio and how the trade affects the risk profile of the portfolio.

[kapital]

how fined grained is your trade data? is it intraday (slice by slice as they trade) or is it wrapped up at the daily level line by line or is it grouped by whole trade irrespective of time or with very little time detail (here is our cost, here is the PnL it generated)?



if you have the intraday level of detail and you can id start and end of putting the position on, rebalancing, taking it off etc, you may get some interesting details on transaction costs/slippage. mite be interesting especially if you were going to try to increase the amt of money already in the strategies. maybe the slippage estimates are already maxed out at current size.



depending on how sensitive the strategies are to actual trading it mite also give you info on how much priority the traders put on the implementation. maybe they are sloppy as hell (crossing the spread even when its 90bps on a low vola name just to get the trade all wrapped up, etc) or incurring unnecessary opportunity cost (lifting the name several spreads while they still have 80% to go cuz its 'fun', maintaining 30% ADV on the best offer while they wait for upticks, etc) . if your data was fine grained enough you could back out/check for some things like this, maybe.

[gutenberg]

Thanks guys. Kapital, Nonius by trade data, I mean literally all individual trades. This is how it looks:



Symbol Price Trader Action Date Quantity Proceeds(incl.fees)

XXX 45.90 1 BUY 5/5/5000 18172 -835,065

YYY 16.02 1 SHT 5/5/5000 -13800 220,310

ZZZ 16.10 2 BUY 5/5/5000 69000 -1,114,350



On this day trader 1 made about 45 trades. Kapital, interesting on transaction costs. Any other ideas?



Thanks,



G