Catching up with Richard Olsen

Richard Olsen is chairman and CEO of the Zurich-based Olsen Group, which comprises several financial research, trading, and market-service companies: OANDA (www.oanda.com), an online forex brokerage/market maker; Olsen Financial Technologies, which makes software tools for gathering, cleaning, and analyzing high-frequency data; and Olsen Invest, a money-management firm Olsen founded in November 2002.

Olsen Invest has six U.S. dollar accounts (each initially funded with $10,000) with varying risk profiles. Currency Trader interviewed Richard Olsen in its first issue (October 2004), an article that was published and updated in the December 2005 issue of Active Trader magazine.

We talked to him again in early October to discuss some developments with the Olsen Invest “Live Accounts.”


WE: The live accounts seemed to have suffered larger than normal drawdowns in September. Could you comment on the causes?

RO: In the wake of Hurricane Katrina, the U.S. dollar experienced a rapid down move against all other major currencies, especially the Euro. Our models handled the initial impact well.

At the time we were worried the dollar would continue to drop, so to manage risk, we increased the hedge of the dollar to protect ourselves against a further massive drop. But our expectation proved wrong and the dollar recovered. We have a parameter in the model that monitors the bid-ask spread. We use it for unexpected news events, such as 9/11, to switch the model into “conservative” mode whenever the model detects a widening of the spread, [it] adjusts its behavior and becomes more conservative. Typically, OANDA publishes a spread of 1.5 basis points. Temporarily, it halved the spread from 1.5 to 0.8 basis points, which had the effect of priming the model to become more aggressive, and [it] increased its positions significantly.

We don’t want to interfere with the model; we decided to let it work itself out of its positions. In hindsight, this was the wrong decision. In response to the outcome of the German election, the Euro made another significant drop that culminated in massive sell-offs in the Euro, Swiss franc, and other currencies on Friday of the previous week.

There was an additional 1-percent drop in the Euro in the morning on Monday, Oct. 3. The model had to rebalance its positions to maintain its maximum exposure limits. When, at last on Thursday, Oct. 7, the Euro bounced back, the model could not recoup its losses (as would typically have been the case) because it was off-balance due to the increased positions it had inherited.

On Monday, Oct. 10, we decided to cut our losses. We closed the existing positions and restarted the trading models in normal mode without memory of the incurred losses.

Here’s an analysis of these events: The trading models and infrastructure are comparable to a complex automated factory. The safety measures implemented to prevent unforeseen losses were insufficient, otherwise the event would not have occurred — our infrastructure was thus subject to Murphy’s law.

If the operational mishaps had not occurred, the trading models would have generated 0.3-percent return in September, and — including the Euro’s rebound on Oct. 7 – the models would have generated a 1.2-percent return in September and October, which is a very different result from the massive (-9.88 percent) drawdown.

We have restarted the models, but have lowered temporarily the return target of the standard profile from the standard 15 percent to 8 percent. We have started an immediate program to upgrade the trading model infrastructure and improve the defensive mechanisms. As soon as we have completed the enhancements, we will hike the return target to the standard 15 percent.

It’s important to emphasize that the drawdown was not because of the models themselves. The models did their job. The drawdown originated on the part of the defensive mechanisms. Over the past four weeks, we have learned many painful lessons and we are making every effort to enhance these safety mechanisms.

In the middle of July, we released new trading models that allow us to increase the trade frequency and improve the rebalancing of exposure between exchange rates. The models improve the stability of returns.

We are highly confident the long-term performance of our models will fulfill expectations, and the recent drawdown will, from a longer-term perspective, be a temporary (however painful) blip.