Near miss screening

22nd Nov '13 by PhilHofGrowYourDough

9 comments 10737 reads

Hi,

Richard Beddard recently published an article claiming that the strategy of reading daily company reports and using the Checklist feature of Stockopedia was preferable to screening alone as a screen provides no scope for flexibility or interpretation of the screening rules. A stock either meets all of the screen criteria or it doesn't.

Richard's argument is that this strict application of screening criteria means that stocks with great potential can get overlooked.

Well ... I for one don't have time to read announcements for every company and so screening really helps me.

However, I'm interested in the idea that my screens might have excluded some great opportunities, so it got me thinking ...

What if I had greater control over the screening mechanism? For example ...

1) If I have 10 screening criteria, it would be nice to be able to specify a 'criteria tolerance' which specifies how many mismatches I'm prepared to tolerate. So if I set this value to 1, then the result set would include stocks that matched at least 9 of the criteria. The stocks that are partial matches could be highlighted appropriately.

2) Continuing that idea, I could indicate which screening criteria are mandatory, i.e. must match

3) Could it be extended further to include a tolerance on the screening criteria themselves, e.g. mkt cap > £50m, however i'm prepared to consider > £30m if it's the only exception

This increased flexibility would help to spot the near misses.

I'd be interested in your thoughts!

Thanks
Phil

Unlock with Email

Unlock with Google

Already have an account?
Login here

DJLJ23

22nd Nov '13

Hi Phil,
I use advfn to read daily reports, it has a feature where you identify the key words/phrase your interested in and give them a colour. i.e "lower than expected" might be red, "greater than expected" green
this makes reading the report quite easy, sea of red, probably not worth reviewing further, sea of green may be one to get in quick.
I have suggested including in stockopedia
David

cig

You can solve the screening discrete cut-off problem once and for all by replacing binary selection criteria by rank measures and aggregate them. The output of the screening process is then not a fixed list of stocks but an ordering of the entire input universe.

That said you may still miss things because they're too far down in the list, but it does spare you the nasty edge effects of binary criteria.

RobertB73

I've built spreadsheets incorporating different financial criteria, and used colour coding to indicate the stock's ranking in each criteria based on quintile. Not exactly what you're looking for, but it does allow me to scan a list, and see quickly the stocks which were top of the class in more aspects than others, and include stocks for further analysis despite not meeting a particular criteria.

PhilHofGrowYourDough

Hi David, yes that is a nice feature. However, time at the start of the day is at a premium in my world. I tend to screen & research in the evening and buy/sell the next afternoon.

Hi cig, well if I understand your solution correctly it might well open up ranges however it would potentially generate many more stocks to be processed, especially if I flex the ranges of several criteria. Rather than flexing on one of ten criteria and only including stocks that failed one criteria test, you'd be potentially including stocks that would not pass the 9 out of 10 rule.

Hi Robert, yup I do this too to process my screened stocks and to monitor my portfolio.

Personally, where possible I'm looking to reduce the amount of spreadsheet based work. I appreciate all of your comments and I hope the ideas are of some use to the development team.

No stock is excluded in the methodology I'm suggesting, you just get an ordered list, and it's up to decide if you look at say the top 20 or go a bit further down. The criteria are not ranges either (which would reintroduce an edge effect), you just apply a shaping/weighting or ranking function to each metric. The end result should be more robust than binary selection for a given number of stocks: that is the top 20 thus obtained should be a better match for what you want than the 20 stocks obtained by filtering using binary criteria and ranges. Setting it up may be a bit less user friendly though.

I'm not suggesting stocks would be excluded rather additional stocks would be included, e.g

If I had 10 rules, two of which were

1) 'mkt cap > £50m' but I'm prepared to accept 'mkt cap > £35m' if it's the only exception
2) 'net debt < 0' but I'm prepared to accept 'net debt < 3 * operating profit' if it's the exception

If I just set the criteria as 'mkt cap > £35m' & 'net debt < 3 * operating profit' then I'd get stocks that are less than £50m and carrying debt. Sure I could sift them manually afterwards manually but it takes time. As a former software developer i want user friendly features that automate and enhance the sophistication of my workflow. It's also a feature that gives extra power to the key selling point of Stockopedia.

snickers

23rd Nov '13

What you're asking for is a way to customise the aggregate quality / value / momentum rankings on this site - to pick your own mix of criteria, weight them as you might want (peg ranking*3, profit margin rank*2..).
There's surely a 'good enough' level with screens though - I'd have thought you'd want to diversify, so catching that extra near miss company won't make a difference, except by luck. Outliers on screens, when I've graphed them (that is, apparent bargains) always seem to be outliers for a good reason, which you won't find in the numbers. And what would a software developer be doing with his time anyway if Stockopedia had done all the work? Enjoy the fun of the coding!

Hi snickers,

Actually what drove the thread was Richard Beddards article. See my original post. So I guess he feels there is potential in the stocks that just missed the cut.

Well as a former software engineer I'm actually in a sort of frustrated position. I cant programmatically access the data (e.g. via an API) and the features that I'd like to play around with will (understandably) probably not be prioritised. I could scrape the data but that would 1) be an incredibly ugly and brittle solution; 2) probably breach several aspects of the Stockopedia agreement; 3) almost certainly get my account suspended or terminated.

Even if it was accessible through an API I'd have to find the time to develop 'stuff'.

I'll sidestep your perspective on developers ;-)

Have a good weekend
Phil

Another reason why it's a useful feature is to highlight data/screen issues, e.g. consider Fairpoint screened against one of my screens called Value screen

http://www.stockopedia.com/checklists/value-screen-14840/?ticker=LON:FRP

It fails on criteria "PE 2yr rolling <= 15" with a reported value of 81.7.
So being flagged as a near miss might highlight that:

1) There might be an underlying data issue
2) My screen might need to be tweaked or forked (e.g. perhaps 2yr estimates are not appropriate for small caps)

Onwards & upwards

Near miss screening

Unlock the rest of this article with a 14 day trial

About the Author

PhilHofGrowYourDough

9 comments

Unlock the metrics for all these stocks and more...

Unlock the metrics for all these stocks and more...