Jamie Gritton's Backtest.org - Missing Models?

Investing Strategies Mechanical Investing
1

Greetings everyone. I’ve been gone a LONG time and recently pulled up backtest.org (Jamie Gritton’s site). I noticed a lot of the models are no longer options to select. Ones like, TREPPE, YLDEARNYEAR, RS26WK, etc. Does anyone know why those models have been removed from his backtest site?

1 Like
2

Jamie thought that discretion was the better part of valor and discontinued working with screens based on Lord Voldemort. His current backtester used SIPro based screens.

DB2

2 Likes
3

I’m not following - so Value Line based screens fell out of favor?

4

IIRC, Jamie got a cease and desist letter from their lawyers.

DB2

1 Like
5

Yep, I ceased and I desisted. It was not only a loss for my backtester and the MI community, but an own-goal for Value Line. The MI screens can’t have been a huge lead generator, but it definitely was one.

GTR1 has the the problem. Value Line screens are off of there for the same reason (my cease-and-desist, not waiting for Robbie to get his own).

11 Likes
6

Thanks for the info – and also, thanks for all you have done and continue to do for the MI community!

4 Likes
7

Good to see you on the “new” free fool boards, Jamie!

2 Likes
8

Does anyone know why the difference in results on backtest.org vs GTR1? Specifically, I was looking at Nets-nets-grahmafied. From 2015 through 2021 it had a GAGR of 24% on backtest.org, and about 11.5% on GTR1 (hold 10). That’s not a small difference!

9

If I recall correctly, one of the big differences is that backrest.org is 1 datapoint per month whereas GTR1 is an average of all start days. I’m sure there are other differences too.

2 Likes
10

John, backtest.org is limited to being a monthly-start backtester only. Also, it can only test the predefined screens from the StockInvestorPro database. Except for blends and screens of screens, no customization is possible.

GTR1 is an order of magnitude more advanced, with tens of features that have a significant learning curve. But the most important feature is that it’s got daily prices for every stock in every exchange that Robbie loads back to the dawn of history for each exchange.

So when you run a backtest in backtest.org, it’s from one defined Friday (or Monday I forget which) close to one month later’s close. That’s it. GTR1 is testing every possible start date from the date you start, appropriate to the holding period you choose and includes lag. This daily start eliminates the variability - luck - and inaccuracy from a single-day-a-month start.

When you read through the archives about old screens being “torpedoed”, it’s because the backtested CAGR coming from backtest.org’s monthly starts were proved to be statistically false by GTR1’s daily starts.

FC

7 Likes
11

Thank you for the detailed response. Very helpful!

1 Like
12

Thanks for that response F.Circus!

13

@gritton Im huge fan of your backtest.org. But i found a bug. Where should i report it to? I dont see any contact info in your site. The screen builder for some reason cant do divide. For example: when i tried this, it wont work: (cash_y1 / ltdebt_y1) > 0

1 Like
14

This is just as good a way to contact me :-).

While there is indeed a big, it’s little more subtle. The trouble is the way it handles a division by zero. While it would make sense to throw out the stock that caused it, it throws out the entire set of stocks the that date. The reason (and the difficulty in fixing) is that after some minimal parsing, the entire Screen Builder code is passed straight to perl, which like to just fail on a divide-by-zero.

There is a workaround though. You can pre-require that the field in question has a value, for example:

ltdebt_y1 & (cash_y1 / ltdebt_y1) > 0

  • Jamie
4 Likes
15

@gritton I was experimenting creating my own screen on your backtester and I was interested by the field netin_g1q5 whose description is “Net Income-Growth from Q5 to Q1” I am new to mechanical investing and investing in general (i’m in highschool) so I was a bit confused when I saw Q5. I could not find any information on google or the AAII website so i decided to contact you.

Is quarter 1 two years ago and quarter 8 is the most recent quarter on your backtester?

Thanks

16

I’m afraid I don’t have a definitive answer to that one. I just got that short description from the SI Pro data files, with a little explanation as you’ve found. I haven’t actually traded any of the SI-based screens myself, as my personal trading has been with timing futures instead of stock picking, since the time I had remove the Value Line screens from the backtester.

Given that is “from Q5 to Q1,” I think Q5 is before Q1. So my best guess is Q1 is the most recent reported quarter, and Q5 is one year before that. That also goes along with a lot of the fields mentioning Q0, which I think would be the current (not yet reported) quarter.

Hopefully someone on the message board will jump in with a good answer.

  • Jamie
1 Like
17

Data Table Name: NETINC_Q1, Q2, . . .Q8
Data Category: Income Statement - Quarterly
Field Type: Dollars in millions (-999999.9 to 999999.9)
Net income for each of the last eight quarters. Calculated by
taking Income after taxes, adding or subtracting Adjustments
to income, and adding in nonrecurring adjustments.

Q1 is the most recent Quater

1 Like
18

Thank you Jamie and @RAMcN ,

Also, If you don’t mind me asking, what is the field “perv”, “PE relative value”?