1,166 reputation
726
bio website twitter.com/shabbychef
location San Francisco, CA
age 42
visits member for 3 years, 10 months
seen Nov 26 at 6:20

matlab/stats/linux nerd. proud, sleepless, new parent. HCSSiM alum. faking it as a quantitative analyst at a small quant fund in San Francisco.


Feb
23
revised Price of Brent versus West Texas Intermediate
latex escaping dollar signs; still renders poorly, tho
Feb
23
suggested approved edit on Price of Brent versus West Texas Intermediate
Feb
20
comment How would you test the hypothesis “There are no idiosyncratic returns available in the market”?
It seems for either of the two proposed tests I will have to correct for the correlation of stocks returns to each other (beyond just via the market) and possibly make some correction for multiple hypothesis tests. I was hoping for a more consolidated approach that looked at all stocks returns simultaneously. The graph presented with the commentary is supposed to be illustrative, I guess, but I am not sure what it tells me.
Feb
20
revised How would you test the hypothesis “There are no idiosyncratic returns available in the market”?
i can unredundant my english
Feb
18
asked How would you test the hypothesis “There are no idiosyncratic returns available in the market”?
Feb
8
answered Expected Growth
Feb
7
awarded  Beta
Feb
7
answered How to combine various equity measures into a single measure (vector magnitude)
Feb
6
comment How to combine various equity measures into a single measure (vector magnitude)
What is the purpose of combining them?
Feb
6
accepted How do you evaluate a covariance forecast?
Feb
4
answered What are the popular methodologies to minimize data snooping?
Feb
4
awarded  Editor
Feb
4
revised How do you evaluate a covariance forecast?
added 746 characters in body
Feb
4
awarded  Scholar
Feb
4
accepted Why does the VIX index have *any* correlation to the market?
Feb
3
comment How are risk management practices applied to ML/AI-based automated trading systems
@Lirik: no, I am thinking about what happens after 6 months of paper trading with mediocre results. Does one give up on finance and become a plumber? Or does one fiddle around with the algorithm, the data, etc? You always have only the data you have today, when you are deciding what to trade tomorrow. If you have any choice and the historical data guides that choice, you have datamining bias.
Feb
2
awarded  Commentator
Feb
2
comment How are risk management practices applied to ML/AI-based automated trading systems
@Lirik: actually what you really do is paper trade a whole cadre of such agents, then on some date you pick the best one and 'switch it on' with real money. At which point you have datamining bias. Unless, as I have said previously, your system is completely without knobs, has the right data, does not require 'featurization' of said data, and works the first time, you spawn a whole stable of these things, or sequentially fiddle with them until they 'look good'. It doesn't matter what your testing looks like, whenever you use the same data to select and evaluate, you have this bias. period.
Feb
2
comment How are risk management practices applied to ML/AI-based automated trading systems
@Lirik, I doubt anyone is going to trade a magical black box without some estimate of its performance going forward. You do this via backtesting. As I said earlier, either 1) all your code works great the first time you try it, and the backtest looks acceptable (gold star for you) or 2) you keep sequentially refining it and eventually trade on the one that 'performed the best'. You now have datamining bias. Yes, one can train the models in-sample, then trade them out of sample, with rolling retrain, etc, but my statement stands when looking at the system as a whole: get it right 1st time or...
Feb
2
comment How are risk management practices applied to ML/AI-based automated trading systems
@Lirik I doubt that could be the case. Backtest arb can only be mitigated by higher fidelity simulation and good coding. Grue and Bleen can perhaps be tackled by choice of algorithm. The datamining bias, however, remains. Whenever you use the same data to select your strategy and evaluate its performance, you are subject to this bias. If your online ML algo is entirely without knobs and it works the first time you run it, more power to you; otherwise, there will be a sequential process of fiddling with it until it 'looks good' at which point you have your bias.