No. of Recommendations: 186
A long project is complete (or at least as near completion as such things get). I've completely redone the backtester's database to use daily price data, instead of the old database, which just had per-month gains. Remeber when having prices for every month was the big new thing?

So what does this change mean to you? The looks is practically the same (though not quite), as are the backtest results. It is, essentially, the same backtester as before. Among the visible changes:

The CAGRs for the screens have changed slightly. There's no perfection in back stock data, but I believe what I have now is better than what I had before. For the most part, the difference is on the order of a percent CAGR, i.e. negligible. There are a few standouts: one is RSWEPS which increased from 44% to 48% (5-stock monthly 86-01) which mostly comes from gains that were mysteriously missing from Dec 1998.

The other screens with siginificant CAGR change are the IBD-based ones. Formula 90 and RS-IBD both got worse (39%->35% and 37%->33% respectively) and UG and UG-90 got better (15%->19% and 21%->26%). UG-90, by the way, no longer appears in the standard menus though its code ("un") still works - this is because no one provides rankings for it, and its backtest data ends in 2000. Formula 90 and RS-IBD are now in line with other RS screens, and UG is still no good.

I got the old IBD-screen returns from Sux's price data, but for some reason I was never able to match his listed gains. I don't have them handy now so I don't know if I'm closer, but with the data I have, I'm confident that I'm on the right side of any remaining difference. There are still some problems with the IBD screens, because of their different trade dates. Mostly, it's that you can't validly mix them with other screens. The backtester continues to allow this, but the results won't be entirely real. For all the trouble of requiring both IBD and Value Line to put these screens together, IMHO they're just not worth it.

The comparitive S&P data has changed a bit too. Before, I used the total-return S&P numbers, which include didivends, but are only available on a monthly basis. Now I've interpolated them with the regular S&P's daily changes, to make estimated daily dividend-included numbers. The upshot is that the monthly S&P comparison is more accurate.

The T-Bill data (and thus the Sharpe Ratios) is also more accurate, considering that I was mistakenly using the 90-day CD rate instead of the 3-month T-Bill. Oops.

One visible difference is in the "list picked stocks" output: instead of the trade month, I'm now showing the actual Monday trade date. Thus ends possibly the longest-standing backtester FAQ.

Temporarily, 1969 is unavailable. As soon as I upgrade my server's perl to a version that understands life before Ritchie, it will be back.

There are some new screens I've been promising for a while: GAR4 GAR4PEG PlowPE RSEG RSPEG1 RSPEG2. Take a look at these and see what stocks they were picking all this time.

Lag. You can now enter a number of days to lag into the trading simulator. By default, the lag is one day: that's the typical Monday-after-Friday that's been the case all along. Now you can go from zero days (i.e. closing on Friday) up to whatever. Note that Value Line's rankings come out on Friday but used Wednesday's close to figure relative strength. So what I call zero lag is really two-day lag.

The lag stuff is certainly worth a look. Spoiler: you're not losing anything by waiting until Monday's close. I'll forgo other results I've seen in favor of encouraging experimentation.

Some Screen Builder changes: I've removed the "CRSP" fields, i.e. "4-Week Return (CRSP)". Instead of having to do something like "(tr4w | cr04) top 10", you can now just say "tr4w top 5". If you use a fields that doesn't have values for the entire backtester period, the Screen Builder will pull these values out of the daily price data. This applies to the various RS fields, as well as price and market cap. With this new method, I'm not getting the same results as before, but as is the case with the IBD-based screens, I trust the new numbers, which I can trace to actual data, more than the old numbers which I had to take for granted.


There's more that can be done, now that I have daily price data for (almost) all the stocks. I've already alluded to RRS. Another long-requested feature is stop-losses (DHYB). I'd like to improve the spreads in the trading simulator by making some way to give larger spreads to lower-priced stocks. I'm sure other ideas I haven't thought of will be requested.

- Jamie
Print the post  

Announcements

What was Your Dumbest Investment?
Share it with us -- and learn from others' stories of flubs.
When Life Gives You Lemons
We all have had hardships and made poor decisions. The important thing is how we respond and grow. Read the story of a Fool who started from nothing, and looks to gain everything.
Contact Us
Contact Customer Service and other Fool departments here.
Work for Fools?
Winner of the Washingtonian great places to work, and Glassdoor #1 Company to Work For 2015! Have access to all of TMF's online and email products for FREE, and be paid for your contributions to TMF! Click the link and start your Fool career.