Page 1 of 1

Help with t formula and optimal-f

Posted: Mon Apr 19, 2010 8:56 am
by d123
Hello all.

I have 12 different simple systems which I follow and 2-3 are being used at the time. Systems are either trend following or based on price breakout and time-frames for mentioned systems are: 5min, 15min, 30min & EOD.

As my switches from system to system are not as robust as they should be and as all is work in progress, I felt free to post here and ask for some assistance or direction.

I’ve plotted for each and every system graph of t and f formula.

t = square root (n) * (average trade / standard deviation of trades)

f$ = (largest losing trade)/((((1+win/loss ratio)*percent wins) – 1)/win/loss ratio)

On each graph I am comparing t formula value (and f formula value) after each trade for the entire sample of the system vs. last 400 trades (rolling). I’ve chosen 400 rolling trades as starting point because of relationship of sample size, level of confidence and margin of error. (For two EOD systems I have 150 and 250 trades, so I am still waiting to record enough trades.)

As it can be seen, both, t and f formula, are sensitive to volatility of the market and market conditions changing from trending to mean reversion. (Due to changes of average trade, standard deviation, largest losing trade, Avg.W/Avg.L ratio, as well of drop/rise of W % due to trending on mean-reverting phases of the market on given-time frame for trend following systems.)

So, the question is:

- Would it be wise to abandon for a while (or cut exposure) system A for which t & f formula graphs (for the entire sample) just have entered Lower High/Lower Low phase (for t formula) and Higher High/Higher Low phase (for f formula) and move to system B for which t formula is in Higher High/Higher Low phase and f formula is Lower Low/Lower High regime?

- Would it be wise to abandon for a while (or cut exposure) system A for which last 400 rolling trades t formula value is below t formula value for entire sample (smaller number) & for which last 400 rolling trades f formula value is above f formula value for entire sample (bigger number, more capital required per contract traded)? (Will it be wise to wait for 400 rolling trades t formula value to drop below 1.6 to abandon a system for a while?)

Any ideas, suggestions and thoughts are very welcome.

Thank you all.

Posted: Mon Apr 19, 2010 11:29 am
by sluggo
Users of Trading Blox software tend to pay less attention to measurements such as your "t" and your "f" because these measures focus upon individual trades. Trading Blox users tend to test (and in real life, to trade) many instruments simultaneously (i.e., a "portfolio"), and so these folks tend to pay more attention to the net outcome of all of these simultaneous trades. The most convenient way to do this is to focus upon the daily equity curve, which shows the aggregate result of all simultaneous trades, on a convenient and uniform timescale.

Blox users tend to pay attention to measurements performed upon the equity curve. Several of these are discussed in the book WAY OF THE TURTLE, such as Robust CAGR%, Robust SharpeRatio, Rsquared, Rcubed, etc. There are many many others, of course; most of them try to measure a Gain-Versus-Pain ratio, using various different ways to quantify Gain and Pain; such as: Sharpe Ratio, Sortino Ratio, Return Retracement Ratio, Kestner K-Ratio, Ulcer Index, Lake Ratio, and so on.

Here are some earlier threads on related topics

viewtopic.php?t=3666

viewtopic.php?t=4179

viewtopic.php?t=6182

Posted: Mon Apr 19, 2010 11:41 am
by d123
Thank you sluggo.
Will read through provided links.
Appreciated.