An excerpt from an email from Toby to the LinnSoft Yahoo Group on 03/25/02

I think your comments, Hayden, offer much food for thought. Just how are TI's used anyway? Who can assimilate ALL that I/RT offers? Don't statistics (and TIs) lie - at least some of the time? Does optimization work? How sensitive is a TI to a variation of an individual parameter? How does a non-expected external news event fit in? (Does a 50% drop in SEPR due to an FDA ruling mean that a system which happened to be long in SEPR suddenly becomes inferior and not worth using for buy/sell timing of other stocks?)

Well, I can't answer any of these questions. They are just some of the 'pitfalls' one encounters while building 'experience' in the markets. You can't win all the time.

I happen to like the concept of "optimization". When I had my Mac, I used Behold! to optimize. It's still probably one of the best packages available for that purpose. However, a system optimized in a bull market may be a loser in a bear market. Further, one would like to picture that an optimization output chart should look like a topographic map of some gently hilly countryside. The high point should be unmistakable, but you ought not to 'die' by falling off a cliff when you change your optimized parameter by a couple of units. One wants a smooth response, not spikiness -- suppose the highest optimization spike is so steep that it fit invisibly within the mesh of the increment grid and was never even tested.

Unfortunately, optimization rarely is smooth and has lots of abrupt cliffs. Suppose one is testing an RSI-type strategy with 30/70 reference lines. Presumably, one sells when RSI rises above 70 and then falls below 70. Let's assume that we have 'confirmed' that 30/70 are the optimum values. What happens when, next time, a stock's RSI peaks at 69.5. This isn't a sale that is shifted up or down a couple of days (as might be expected if one were working with moving averages). It's a sale that is totally missed. One's profit has just fallen off the cliff without a safety line.

I have found that "optimizations" have too many cliffs. And just what is one optimizing anyway? Net profit (without regard to number of trades required)? Percentage of winning trades? Average winning trade? Profit per bar held?

I have abandoned the quest for optimization.

Are Trading Systems in that same category? Maybe. I happen to like trading systems conceptually. My favorite is based on the TDI indicator. The system I use results in profitable trades 65% of the time. That's almost two winners for each loser. Yet maximum drawdown is atrocious for many stocks. The system would have me buying LU at $70 and still holding at $7. Unacceptable. Yet every attempt I've made to build stops into the system yields more trades, lower percentage of winning trades, and ultimately lower total profit. Yet this is the Trading System I like best to work with.

Unless one is a day trader, I see no urgency to use a Trading System literally. On the other hand, I think it's a fantastic tool to use interpretatively. It becomes a first-level screener. My buys are selected by the system, but I place the trades (or not) based on how I like the behavior of the two TDI lines.

This takes us down to the first level - raw TIs. The only 'truth' in a TI is the computation. The use of the TI is in the interpretation of the investor. Statistics "lie". So do TIs. The skill is in knowing how to 'mold' a TI. It's all in the interpretation. We know that 5000 users of technical analysis will end up with 5000 different interpretations. That's probably how it should be. Does that mean that T.A. is '"voodoo science", as is often written on the message boards? Well, if I believed that, I wouldn't be a user of I/RT.

What Hayden suggests about the study of TIs is very constructive. The more one knows about TIs, the better off one is. But I think the ultimate mark of a successful investor is not in the choice of TIs and their parameters. It's in how he builds up his experience. Tech Analysis is a great tool, but the better tool is being able to see the Big Picture. Few of us (myself included) really achieve this.

Toby