1991issue C021-7

A fuzzy membership checklist for trade signals

Build the trade decision as one graded pre-trade checklist. Score every rule input on a 0-to-1 membership scale, collapse those scores with a single min or max step, and only then decode a crisp enter, exit, or stand-aside signal so the whole procedure can be tested as one rule.

Fuzzy decision scores occupy the interval from 0 to 1, so a market condition can be treated as partly true to that degree instead of only true or false.
AND combinations take the min operator and OR combinations take the max operator, collapsing left-hand-side scores into one rule membership before any action is decoded.
Defuzzification reverses the same membership table used in fuzzification, reading the action score back to a crisp instruction.
Editorial: treat scoring, the min-or-max collapse, and the final decode as one pre-trade checklist so enter, exit, and stand-aside stay testable as a single procedure.

Entries in this reading1 entry

Current trackPre-trade checklist245 readings

One checklist from score to signal

A pre-trade checklist is a single procedure that scores rule inputs against market state, combines those scores, and maps the result to an enter, exit, or stand-aside signal. Fuzzy decision scores occupy the interval from 0 to 1, so a market condition can be treated as partly true to that degree instead of only true or false.

The archive workflow keeps those steps inside one rule path. Inputs are scored first. The scores are then collapsed. Only after that collapse is a crisp instruction read out.

Scoring inputs with membership functions

A membership function is a 0-to-1 mapping that states how fully a crisp observation belongs to a linguistic label. It maps a crisp reading onto that scale by assigning 1 at the designated typical value, 0 outside the accepted band, and interpolated scores in between.

A second-order curve is often used in place of a straight line while keeping the same 0-end, midpoint, and 1-end. Fuzzification is this conversion of a measured market variable into membership scores on linguistic sets.

Production rules in an elementary expert system

An elementary expert system is an inference engine plus a knowledge base of if-then production rules. The knowledge base is the collected production rules that the inference step consults. The inference engine is the module that presents current data to the rule set and decides which rules fire.

A production rule is an if-then statement pairing a condition set with a conclusion the procedure may execute. The left-hand side is the condition portion that must be satisfied before the rule fires. The right-hand side is the conclusion portion that becomes a candidate action when the rule fires. A rule fires when every left-hand-side condition holds, and the right-hand side then becomes a candidate for execution.

Fuzzy rules and the min-or-max collapse

A linguistic label is a graded category such as positive-big or negative-big used in place of a binary true-or-false flag. A fuzzy production rule requires crisp inputs such as price and a price-to-earnings ratio to be converted into linguistic labels before it can fire. One demonstration rule sets the action to negative-big when both inputs belong to the positive-big set.

AND combinations take the min operator, which returns the smallest membership among the conditions. OR combinations take the max operator, which returns the largest membership among the conditions. With a price score of 0.5 and a ratio score of 1.0 on positive-big, the rule membership and the action membership are both 0.5.

Reading the action back to a crisp instruction

Defuzzification is the reverse lookup that turns an action membership score back into a crisp instruction. It reverses the same membership table used on inputs, reading the action score back to an index and then to a crisp instruction.

A spreadsheet with linguistic columns and real-world rows can serve as both the fuzzify and defuzzify interface. Editorial: keep that decode after the min-or-max collapse, so enter, exit, and stand-aside remain outputs of the same checklist rather than separate true-or-false checks.

Membership tables and the generator

Membership tables may be linear or curvilinear and either sigmoidal or bell-shaped. The usual seven labels run from positive-big through zero to negative-big, and neighboring curves should overlap so no value in the variable's range is left unlabeled.

A membership-function generator can build those tables from user-specified 0-edge and 1-edge points, automatically choosing which side of a sigmoid is high, joining opposite sigmoids to form a bell, and writing a seven-column matrix of labels against indexed crisp values.

Stated construction advantages

Stated construction advantages of this checklist form include encoding verbal expert statements as executable rules, simple programming on modest hardware, and execution speed associated with real-time applications. Smaller rule sets and smoother control actions are also asserted.

Editorial: those points are construction claims from the archive workflow. They are not a performance record and they are not a present-day market claim.

Educational research material, not investment advice. Historical source context does not establish present-day performance.

13 of 135 in the Pre-trade checklist track

19911-3 pp.Next on Pre-trade checklistOverconfidence as a skipped pre-trade checklistRising certainty after recent wins is a process input: halt and re-check entry, exit, and stand-aside rules before any new order.

All readings on this track · 135 readings

By fieldDecision process416 readings By year1991166 readings

All 245 readings tagged Pre-trade checklist

Also on Pre-trade checklist5 readings

1984Pairing tradeoffs with pre-trade checklists 1985Event-class hazards and pre-trade survival checks 1986Construct a decision procedure that revises itself 1986A pre-trade checklist for crowded trading methods 1986Two gates for setup and operator readiness