Imagine you’ve tinkered for days or even weeks, perfecting a strategy idea that’s showing a whole lot of promise. You’ve meticulously tweaked a mouth-watering Sharpe Ratio out of your backtests….it even survived costs. YES!
Systems go, let’s trade it.
Imagine this new strategy enters a drawdown.…maybe a lengthy one….maybe from day one!
How would you react to such a letdown?
A common response to a long or sharp drawdown is to defer to our self-preservative instincts and pull the strategy entirely. Maybe you’d compare your poor live performance to your promising backtest, turn red and frisbee your laptop out the window.
If you enjoy making money trading you’ll need to do better.
A relatively experienced systematic trader, who we’ll call Jack the Quant, would look at this drawdown scenario a bit differently.
Yes I know Jack is actually a terribly drawn Napoleon Dynamite
Anyway, Jack would rely on:
- an understanding (and acceptance) of the nature of the markets
- a sound systematic research process
- generally chilling the heck out
We’re going to quickly talk about the latter two. You can learn about all three in depth here (or here for existing RW members).
Jack’s knee-jerk reaction to drawdown
Firstly, he’d go make some tea and relax.
Jack knows it would be trivial to show that a strategy with a true Sharpe ratio of 1.5 stands a reasonable chance of having a two-year drawdown. Jack also knows that the strategy he backtested to a Sharpe of 1.5 isn’t likely to deliver him that performance once live.
This implies that in order to decide whether the performance of Jack’s strategy is “unexpected”, he might have to wait years.
That’s assuming he has an accurate picture of what the “expected” performance might be – which he almost certainly doesn’t have.
Quite the conundrum, isn’t it?
This is why having a sound research process to begin with is so important to your success and sanity as a trader.
A sensible research process lets you gather enough evidence to make a smart bet that your strategy has a good chance of paying off in the long run, even if your strategy isn’t too hot out of the gates.
Satisfied with that evidence, Jack’s approach is to:
- size the strategy small
- chill out and let it do its thing for a couple of years
- in the meantime go hunting for other trades.
This might surprise most beginners who would assume a systematic trader would have some kind of automated process for decision making during drawdown. But, Jack’s approach stems from his understanding that trading is hard and that losing money sometimes is completely normal. Sometimes (or actually, most times), the solution is just to relax and keep moving.
Besides, Jack is smart and has structured his portfolio sensibly so that his risk premia tailwind acts as a buffer for when his alpha trades don’t go his way, which is inevitable. In the end, he still has trades on and is making money.
This is the less frenzied, more liveable way to run a trading operation.
So be like Jack — copy his well-rounded approach and keep your eyes open for new trades.
But surely there are circumstances where we’d turn a strategy off absent a couple of years of live performance data?
You bet!
Jack would down his tea quick smart and pull the pin if live trading demonstrated that his backtest assumptions had been violated so badly that his edge was destroyed. Maybe he was a tad optimistic in his execution assumptions during the research phase, and the live market showed that unexpected frictions are too much of a hurdle to surmount.
He might also pull out if the basis for the trade ceased to exist. For instance, say the trade was based on forced flows from ETF rebalancing. If the ETF Jack was trading changed its mandate in such a way that impacted those forced flows, he’d stop trading the strategy.
But that’s a pretty clear cut example. Often trading isn’t clear cut at all.
Say you were running a convergence trade betting that two related financial assets were linked through shared risk factors. It’s a complex question to decide if the dynamics of those risk factors have changed permanently.
In that case again, as in most cases, the best answer is to size it small, chill out, and find more stuff to trade.
Trading can be uncomfortable, you’ll often lose money before you make it….
….this is part of the game, and that’s OK! We just need to make more than we lose. We show you this inside our Bootcamps.
“Okay I understand….but my backtest says my strategy should rock!”
There’s always a temptation to compare live trading results with the backtest to decide whether or not to pull out. But this violates one of the most important tenets of successful systematic trading:
Don’t use a backtest to define performance objectives!
Think about it. You’ve spent days or weeks or longer researching and developing the strategy. Consider how many decision points you encountered along the way. Which universe of assets to trade. Whether to choose this parameter, that parameter, or both. No matter how careful and considered your development process, some amount of optimistic bias creeps in with each decision. This is a fact of life. Unavoidable. It’s part and parcel of strategy development. We can’t not introduce data mining bias. Which means that we can’t trust our backtest to set future performance objectives for us.
Don’t get me wrong, backtesting is probably the most powerful tool in your arsenal. The ability to acquire empirical evidence that an idea worked or not in the past is crucial. Backtesting gives you this ability.
Problems arise, however, when you expect live trading performance to match the backtest.
Again, sensible ol’ Jack knows this.
Is there really nothing to measure?
Not entirely….
There’s always the temptation to try and systematise decisions (they don’t call it systematic trading for nothing). The decision to pull a strategy based on live performance compared to a backtest is no exception. It’s comfortable to know that there’s an algorithm or a set of rules that’s got our back.
This is generally not a good idea.
Generally speaking, to repeat what we talked about earlier, a good systematic trader is likely to approach drawdown by:
- adopting a sound research process and sizing small – before drawdown ever occurs
- chilling out and letting the strategy go to work, even during said drawdown
- always looking for other opportunities to trade
Find yourself faced with a strategy that’s a bit underwater? Does it fly in the face of your backtest? Are you frustrated?
It’s all part of the journey.
But all that said, maybe there IS some small utility in quantifying this approach — we’ll investigate this in part 2 of this series.
While we’re putting that together you can learn more about where exactly algo trading CAN help you trade more profitably by downloading the free PDF below….
[thrive_leads id=’10392′]
1 thought on “A Quant’s Approach to Drawdown: Part 1”