Timeline for What methods to use for statistical prediction/forecast of trading data?
Current License: CC BY-SA 3.0
10 events
when toggle format | what | by | license | comment | |
---|---|---|---|---|---|
Aug 5, 2011 at 11:32 | history | edited | 410 gone | CC BY-SA 3.0 |
added 4 characters in body
|
Aug 5, 2011 at 11:18 | comment | added | 410 gone | @RockScience Yes: and it's also a resource for the whole web public. The audience is much much wider than quants. If this were a private forum for a quant-only audience, then I'd have given a very different answer. | |
Aug 5, 2011 at 9:37 | comment | added | RockScience | @EnergyNumbers This forum is for professionals | |
Aug 5, 2011 at 6:17 | history | edited | 410 gone | CC BY-SA 3.0 |
added 901 characters in body
|
Aug 5, 2011 at 6:07 | comment | added | 410 gone | @Michael Bishop - for almost all readers of the answer, those in profit will be there due to luck rather than skill. For the tiny minority of readers, such as Matthijs, who are professionals, then it is different: that's what I tried to cover in my paragraph on quants, above. | |
Aug 4, 2011 at 21:04 | comment | added | Michael Bishop | +1 for a good answer but it seems misleading to me to suggest that we should hope that our algorithm's previous success was luck rather than skill. | |
Aug 4, 2011 at 20:26 | comment | added | user369122 | hi, I have not optimized the parameters of the algorithm and already made quit some successful runs on shorter and / or different periods and other stocks. The systems seems to be OK. But I want to put into stats, so others can have a look at the figures as well. I can now show them that over any given long term (a couple of months) and with most liquid stocks, the P&L will move upward. But I need more than just that. They want re-liabilities etc for risk and some other stuff. | |
Aug 4, 2011 at 19:50 | comment | added | 410 gone | OK: of the 1250 points, how much was used for designing the algo? Are older data available to backtest further? If so, you could randomly sample 125-day and 250-day sequences within the data you didn't use for calibration, and look at the distribution of P&Ls across those intervals. (if you're happy to assume some persistence in variance over time ... (paging Nassim Taleb ...) | |
Aug 4, 2011 at 19:35 | comment | added | user369122 | well..thanks for your thoughts..but I'm already active as a trader for about 10 year and work for a successful company. I can understand your thoughts, but it doesn't really account for the strategies I currently work on or have been working on in the past. But uh, just looking at the numbers and my question, can you help me out? | |
Aug 4, 2011 at 13:54 | history | answered | 410 gone | CC BY-SA 3.0 |