You Don’t Back Test Your Life

In his book, PitBull Marty Schwartz writes about one of the greatest individual trading runs in the history of the markets. As a pioneer trader in the S&P futures pit in the early 1980’s Marty noticed that there was a strong relationship between the price of bonds and the S&P futures. If interest rates rose stock index futures dropped the next day. If interest rates declined, equities were bid up. Since bonds stopped trading at 3 PM NY time and S&P futures traded until 4:15 PM there was a little more than an hour where you could watch the bonds trade in the cash market in after-hours trading. If bonds staged a significant rise of 3/4 of a point or more the probability of S&P 500 opening higher the next day was very high. As Schwartz noted, “Being ahead at the opening was like waking up with a woody.”

Here is how he describes the trading run.”All through October, I smacked the S&Ps when they went up and I smacked them when they went down. On October twenty-second, on rumors that the Fed was not going to lower the discount rate before the election, the physicals plummeted in after-hours trading, the S&Ps opened down 1.85, I was short 150 contracts, covered at the opening, and in one minute made $138,750. By the end of the month, I was up $1.4 million. My legs were sore from jumping up and down, my voice was shot from screaming at Debbie on the phone, and Audrey’s ribs were tender from being hugged. In February, when we’d crawled out on a limb and dumped $400,000 into the beach house, our net worth was $1.2 million. Now, in one month, I’d more than doubled that, I’d made more in a month than I’d made in my entire lifetime. I can’t begin to describe that feeling. Every day, for twenty straight days, we’d get in the Eldorado to drive home from work and we’d be, on average, another $70,000 richer. It would have taken me a whole year to make $70,000 if I were still a securities analyst.”

Here are a couple of things to keep in mind about Marty’s “system”. He didn’t backest it on a thousand samples across an array of markets to prove its robustness. He didn’t optimize parameters or run Monte Carlo simulations, or try to see if this “signal” worked on wheat or pork bellies or some other unrelated nonsense. He also stopped trading it the moment it stopped working.

His system -- like all successful trading systems -- was simply a behavioral edge that he exploited until it stopped working. It’s not that he didn’t do research. Before committing capital to the idea, he did check past occurrences and started making relatively small bets until his thesis was proven correct -- but by standards of “data scientists,” it was a woefully inadequate test and yet it was one the most successful individual exploitations of the market ever recorded.

I bring this up because I think most retail traders are far too obsessed with backtests. Backtests are good at only one thing -- showing you how to make money yesterday. If you really want to learn how to make money today and tomorrow and the day after tomorrow you need to stop testing and start doing. Losing, as I noted in last week’s column is the single best test that you can run. Losing in real market conditions will finally tease out the profitable idea in your thesis -- if there are any. Lose and backtest. Lose and backtest. Lose and backtest until you start to win.

Otherwise, you will simply waste all your time trading yesterday’s data while learning nothing about how today’s price action differs. How many times have you seen a system with a perfect equity curve, passing every statistical test under the sun, fail in real market conditions?

100%.

I have never seen a backtested system that could maintain its profitability for more than a few months without serious editing and adjustment to its initial assumptions. That’s because successful trading is a function of understanding past price behaviors and while being finely attuned to any present-day variations in the market. Yes of course price action is cyclical and basic buying and selling patterns persist over and over. After a self-off comes a rebound. After a rally comes a selloff. But the amplitude of each move is highly variable that’s why the future is just different enough from the past that you won’t be able to exploit it mechanically.

Backtests should be viewed not as justifications for systems, but rather as insights into certain behavioral edges that will not last. That’s the other key to understanding backtests. They will all fail under future market regimes -- and if you understand that going in it will be a lot easier to abandon them or modify them when they stop working.

There are literally hundreds of exploitable edges in the FX market every year, but they are fleeting and usually very instrument dependent. There is no “universal” system of trading that will work across all markets. That’s why the most successful individual traders tend to specialize in one market or even one product.

So stop wasting hours on a perfect tweak of yesterday. You don’t backest your life. You live it. Do the same with trading.

Boris Schlossberg

6 comments

  1. Femi says:

    Keep it up Boris. Your weekly thought provoking writeups are a must read for me. It is a nice way to round up the week’s trading while preparing for the next.

  2. Samir musa says:

    Hi Boris,

    You’re philosophical genius in trading.

    It doesn’t matter what your entry what matters is how you exit or manage yourself out of trouble .

  3. Nicholas says:

    This is real life stuff. Thanks for this kick in the pants. I have been guilty of trying to backtest one edge across all majors but they are different beast even though similar.

  4. dmitry says:

    Backtests are good as falsifiers though. If a backtest fails, the chances are that the real-time strategy will fail too, and taking this into account may save you tons of effort and time.

Leave a Comment

Your email address will not be published. Required fields are marked *