Systems go through the following lifecycle:
- Production Trading
Each step will now be described.
First we must have an idea: A testable hypothesis about how a tradable
instrument will perform under certain circumstances. The accuracy of the predictions made by the hypothesis do not have
to be 100%, or even 75%, but better than 50% is a good starting point.
- Stocks tend to have carry-through on large intra-day moves
- Certain markets over-gap on good or bad news and then reverse in the opposite direction of the
Once we have our hypothesis we can create a complete trading system that implements the idea.
The trading system will include all the elements outlined previously in the 'Our Systems' section of this web site.
Each trading system goes through 3 phases of testing before it can be introduced as a new production
Phase 1 is historical testing of the idea back at least 1000 trades, or 1000 days,
whichever spans the longest time. Although previous performance of a trading system is no guarantee of future success,
systems that have never worked in the past are unlikely to start working in the future.
If the expectancy (average winner/average loser) of the system is not positive in historical testing
it is discarded as a false hypothesis. If the system has a very negative expectation we may reverse the original hypothesis
and continue to Phase 2 testing. Occasionally the hypothesis may be modified and re-tested, although it is better to
view this as a new system test.
Phase 2 is the testing of the system for at least 30 real-time trades on
paper, recording trades as if they were real. If, at the end of this period, the expectancy closely matches the historical
expectancy, then we can proceed to phase 3. If the expectancy does not match the one from Phase 1, we can continue with
Phase 2 testing for another set of 30 trades. If we continue to get a negative expectancy for the paper-trading we will
discard the system at this point.
Phase 3 testing involves trading the system with small lots to ensure that we
can really achieve the entry and exit prices estimated by our historical and paper trading. If, after 30 trades we still
have a positive expectancy, then the new system can be implemented in production.
If a system passes successfully through the 3 testing phases, it can be implemented in the production
trading environment. The system will be allocated a portion of the trading account, and each entry signal that is generated
by the system will be traded. Implementation includes a formal definition of the system, automation of the trade entry
signals, and daily monitoring of positions.
The system is traded in production as long as it has a positive expectancy,
and is not temporarily suspended by any of the rules that govern when the system can be traded.
Each production system is monitored daily, weekly, and monthly, to ensure it is operating within
the expected parameters, and does not need to be temporarily suspended, or permanently retired.
Some systems have rules that determine periods when they should be temporarily suspended.
Also, if the expectancy of a system tends towards zero (break even) then we will suspend it in production and revert to paper
trading until the expectancy goes back to the historical norm, or the system is retired.
If the expectancy of a system goes to, or below, zero, and does not recover after paper-trading
for a statistically significant amount of trades it will be permanently retired. Occasionally poor performance is due
to the omission of a rule that would temporarily suspend the system during certain periods of time. If this is the case
the system may be re-implemented with new rules and go back through the whole testing process again.