Quote from Index Maximus:
Why did your indicator logic on ScottD's thread fail to produce sustained positive results ?
Hershey trading logic put to the test:
http://www.elitetrader.com/vb/showthread.php?s=&threadid=147441&perpage=6&pagenumber=1
Fail
By the time May rolled around we had our ideas collected. The logic sheet posted at the beginning of the thread did enable us to lay a foundation that was good.
Each version had some refinements of the basic core first shown.
As those observers who code could figure out, the EOB aspect was a good starting point. We were very fortunate to be able to put in place the first level of filtering as well; no one picked up on that, however. Scott is a champion and he followed the programming standard of not moving on until, at a given level, all problems were solved.
The standard for programmers is to have positive results in order to expend more time.
The chart he tested me with was a real draw for me. I felt he had terrific insight on how to get an initial fair evaluation of the potential for using indicators on markets. I particularly enjoyed how the project started and continued.
In trading, as in coding, there is a crossover point where the profits made far outweigh the costs of modelling, developing and coding an ATS. As has been pointed out, there are many many refined methods of trading successfully using an ATS connected to a platform. We used a good set of values for costs and slippage on the core versions.
If you look at the Sweeps Chart's nine tables, we got part of the Coarse sweeping done.
The Tucson crew did a four aspect documentation effort and we worked on a core and three shells. Our Excel support aspect was supported by logic sheets, chart snagits, and text using dragon 9. The core was about 600 lines in Excel.
Each shell was larger than the prior and the first shell's completed document was over 120 pages. We did not submit it because the core did not produce positive results and that was deemed to be a requirement for moving forward.
All shells were fleshed out, in terms of logic, filtering and gating.
The posted results from Scott are very accurate. He is very skilled and does super testing of data. He also provided terrific data and information for us to work with in Tucson.
The core used scalar data values exclusively.
All shells would have used binary vector values from indicators only. We all took very seriously the specified limitations on using only the information in the chart he used to test me. The core filter(s) he coded were really terrific. There was no way, on that level, we could overcome EOB relative indicator values as compared to individual bar price volatility ranges (which we could not employ). All shells eliminated this problem when used manually with intrabar indicator data.
So as you say the logic was a failure for the later core versions; the initial versions were profitable.
Using the platform we worked on, was terrific; I have it running on the ES daily. All of the panes for the indicators have lites operating as well (31). What I do manually, is use them as desired. Primarily I use volume leading price in one pattern on nested fractals. None of this has to do with indicators.
Over 5500 people looked at Scott's test chart. A lot looked at my post of the trades and many looked at Scott's super-position of the trades on Scott's answer sheet. We all worked hard together and everyone learned something.
Obviously, reading indicators is like tape reading and it lets a person call the profit segments at at least the level of the ATR.
ET, by enlarge, feels that the thread was a failure.