To clarify, I assumed that position sizing would be based on each month starting at $100,000.
Absolute, I’d like. For now, relative performance is still good.
"t = sqrt(number of trades) * avg trade / std deviation of avg trade"
Another approach would be to combine the t score with optimal f and some additional metrics to get multiple views into a system’s past behavior. A single metric is not likely to be enough to generate a reliable forecast of the future behavior of the system.
–h
If there are two sets of systems (less than 100 and more than 100 trades), I suggest to compare each set separately. Then you have only one formula for each set.
( For me it is simpler, because I do not believe anything with less than 100 trades 
)
Is there some recommendation for minimum of "number of trades"?
30 trades, and longer than a 2 year backtest or record. 30 is all that is required in a normal system. And, I might add, each day of the system is actually a datapoint. That is, a 2 year system has about 500 bits of data, and even if there’s only 30 trades, as long as there’s 30 or more, you have a normal population if you want to just go by trades. Really, each day is either buy sell hold or do nothing. So time is a big factor.
I wouldn’t trust a 1000 trade system with only 30 days of data, I’d be the opposite. 1000 days of data with 30 trades is enough, assuming a win percentage at 70% and above.
>> And, I might add, each day of the system is actually a datapoint.
An interesting way to look at it, Beau.
I wonder if this concept can be extended further - each day is considered a trade. After all, at the end of day, the system chooses not to exit open positions, thus implying that it expects to perform better on the next day.
A positive change in equity is considered a winning "trade", and vice-versa. If system statistics are computed based on this method, I wonder how they would compare against those that are computed on trade-by-trade basis.
When I develop strategies, I personally aim for 5-10 years of walkforward history, with 100+ trades. Not always possible, though…
Add about 1400 basis points now to about 9,891 basis points of outperformance since 3/20/2007 on the spx.
Oooh, yeah, what was that? Go from 450 back to 963 which is really how I should be ranked? What now?
Perhaps a formula which tracks the psychological makeup of the System Developer going forward, which would give warning signs before they self destruct and cause subscribers’ lose loads as has happened before on C2…That could be something!
Maybe, but everyone was turned away near the peak.
