tombrown1
61 posts
msg #105682
- Ignore tombrown1 |
3/26/2012 6:32:34 PM
Just an update on some of my backtesting experiences:
I'm only disturbed by one thing with this filter and it is from the period of March 23, 2008-March 23, 2009. I've been trying different sorting methods, and most of them lose a good chunk of money during this period - the worst was an 87% loser. There were only two out of the 12 sorting methods I tried that actually made money in this period.
One was sorting by volume descending which returned 106%. The other is sorting by close ascending which returned 41%.
Ideally you want a robust system to be relatively consistent and make money when you tweak individual settings a bit like this. This one year and ONLY this one year - which is important because it was during the crash (I'll backtest the dates of the exact crash at some point soon) - seem to be robustly losing money. Maybe higher volume or lower priced stocks did ok for a reason that year - or maybe it was a lucky random way of sorting those trades that made money.
I'm just posting this as a warning that if we see a crash like we saw then, things could get a bit dicey for this method.
For me, I'm going to continue trading by volume descending and keep my fingers crossed that Mr. Market doesn't get too irrational (that never happens, right?).
Best,
TB
|
blumberg
27 posts
msg #105686
- Ignore blumberg |
3/26/2012 9:31:51 PM
Interesting that the filter on pg 7 came up with GAZ today. GAZ is selling at a 58% premium to its NAV so going long here would be a terrible idea.
|
jimvin
173 posts
msg #105687
- Ignore jimvin |
3/26/2012 9:37:54 PM
StockFetcher is my investing analysis tool of choice in almost every way...the almost being the backtesting.
While it comes within striking distance of my investment style, running paper-trade studies in real time over the course of the last few years (and real-trading where warranted by the study results) I've found that where StockFetcher data indicates I've lost a quite a lot of money, the opposite is true.
For whatever it's worth I'd suggest running paper trades for the coming month, then doing a back-test in StockFetcher to compare the results.
Note that paper-trading is (I believe) preferable to simply taking the open-high-close and extrapolating probable profits since the simulated trading genrally returns results of 1/3 to 1/2 of the simple data draw of open-high-close.
|
Nickster8074
53 posts
msg #105688
- Ignore Nickster8074 |
3/26/2012 9:42:37 PM
Does anyone else feel that trading TVIX tomorrow is a scary proposition? It's at the lowest price it's ever been, but it's price action has been ridiculous.
|
blumberg
27 posts
msg #105689
- Ignore blumberg |
3/26/2012 9:59:14 PM
TVIX price action has been ridiculous bc this ETN has been selling at a premium to its NAV.
The premium to NAV was a result of Credit Suisse's decision to halt issuance of new units of TVIX. Following the decision to halt issuance, TVIX started to trade at a premium to NAV because demand for the product outweighed supply.
Even when TVIX reaches its NAV, traders should avoid buying TVIX for other reasons. Most importantly, the contango in the VIX will lead to continued "monthly roll" loses in the future.
This is also the case for other ETFs and ETNs (like USO, UNG, GAZ) that try to mimic the underlying commodity's price behavior through the trading of futures.
|
Nickster8074
53 posts
msg #105690
- Ignore Nickster8074 |
3/26/2012 10:31:20 PM
So, in other words, wait for this filter to spit out a stock, not an ETF, leveraged or otherwise. Got it.
|
blumberg
27 posts
msg #105693
- Ignore blumberg |
3/27/2012 1:48:06 AM
Most ETFs such as SPY or IWM won't have this problem. The trouble is with commodity ETFs and ETNs that try to mimic the underlying commodity's price behavior through the trading of futures. .
Commodity ETFs that hold their positions directly in the physical commodity itself (such as SLV or GLD) will not have this problem.
|
levamit
101 posts
msg #105694
- Ignore levamit
modified |
3/27/2012 3:18:54 AM
Hello Guys,
I personally prefer to scan without the ETF's.
They may decrease performance of backtesting.
And they differ in their volatility than stocks.
Just add to code :
stocks are not ETF.
Regards.
levamit
|
tombrown1
61 posts
msg #105701
- Ignore tombrown1 |
3/27/2012 4:51:37 PM
Just a quick update - I finally got around to testing this method with the crash of 10-11-07 to 3-6-9 and it came up with a total 72% equity gain. That's using a 7-day hold or rsi(2) > 55 for the exit.
Thanks again levamit.
Best,
TB
|
tombrown1
61 posts
msg #105702
- Ignore tombrown1 |
3/27/2012 4:56:28 PM
"StockFetcher is my investing analysis tool of choice in almost every way...the almost being the backtesting.
While it comes within striking distance of my investment style, running paper-trade studies in real time over the course of the last few years (and real-trading where warranted by the study results) I've found that where StockFetcher data indicates I've lost a quite a lot of money, the opposite is true.
For whatever it's worth I'd suggest running paper trades for the coming month, then doing a back-test in StockFetcher to compare the results.
Note that paper-trading is (I believe) preferable to simply taking the open-high-close and extrapolating probable profits since the simulated trading genrally returns results of 1/3 to 1/2 of the simple data draw of open-high-close."
Jim,
Could you be more specific as to what the problems were? Could you post a couple trades that SF picked that didn't line up with reality? I'm curious to see exactly what problems you were running into.
There are ways of "gaming" the backtester to achieve outlandish results - for example by using the profit stop feature and not holding for at least one day you can achieve amazing unrealistic results.
Best,
Tom
|