Category Archives: learning

When is past performance predictive?

The phrase “past performance is not an indicator of future performance” is unavoidable when investing. Funds that have out-performed historically aren’t any more likely to perform in the future than random. A recent market rally doesn’t mean things will keep going up. The phrase is everywhere.

A normal person’s first reaction is likely to be “that doesn’t sound right. Isn’t past performance usually the best predictor of future performance? If the past isn’t predictive, how can we know anything?

If you want to really get frustrated, go one level deeper. The statement itself is self defeating. If historically it has been true, that doesn’t mean it will be true in the future. Perhaps in the future, the past will be predictive. Who knows? It’s enough to make you think you should ignore it.

This is actually a struggle that applies in other areas of your life too. Every day we have little heuristics telling us how likely someone is to be late, make a funny joke, skip the gym, or miss a deadline. And the biggest driver of these predictions is usually what that person has done in the past. But people change, so perhaps we shouldn’t.

So in general, we seem to place a lot of weight on past performance, because in our everyday lives it is predictive of future performance. But that seems to not be true when it comes to investing.

A framework for when you should pay attention to past performance

I’ve been struggling with this logical mobius strip for years, and I’d like to share where I’m at.  This is my framework for discriminating between when the past does predict future, and when it doesn’t. Our time and attention is precious, and we should spend it where it’s most effective. The framework should help us focus on how to make good, systematic forward looking decisions. It will help us decide how much weight to put on past evidence depending on it’s usefulness for future decision making.

These three questions are my filters to try to separate useful history from useless history:

Is the root cause clear?

Is the root cause consistent and persistent?

Is it possible for the belief to be held universally and still be true?

To apply the questions, let’s use these three examples:

  • Asset location.
  • Passive funds
  • Highly active funds (i.e. high active share, long only)


Is the root cause clear?

It’s easy to see what historical performance is… but what caused it?

In the case of asset location, the cause of the outperformance is the relative decrease in taxes paid. It’s pretty clear that this is a mechanical relationship: by having the overall portfolio pay less in taxes, it can grow quicker, and the investor keeps more money.

In the case of passive funds, the performance is caused by two things: the index’s performance, and the funds cost and tracking error. The index’s performance will generally be driven by expectations of future economic conditions, and markets (other people’s expectations) are really hard to predict. But the costs and tracking error are pretty consistent – lower is better.

In the case of very active funds, the performance is caused by three things: the index’s performance, the funds cost and the manager’s deviation from the index. The first two are the same as with passive funds: managers are still exposed to economic conditions, and costs mechanically drive down performance. But high active-share managers/funds have significant leeway to have the manager’s skill (or lack of skill) overcome those factors by deviating from the index. To be clear, it’s not guaranteed or even probable a manager has sufficient skill – just that the ability to deviate allows them to express their views.

If an active manager has positive skill, it’s worth considering what the root of that skill is? Better information (obtained legally)? A systematic statistical or economic predictive process? Better trading technology? A gut that can act as a crystal ball? That leads us to the second question.

Is the root cause consistent and persistent?

Once we’ve identified the root case, we can think about how certain those causes are to persist into the future.

The performance due to asset tax location seems likely to persist. I expect there to be taxes in the future, and that different investment returns will be taxed differently. And that different account types will be taxed differently. So pretty much every year I expect the performance to persist. And I’ll know in advance about when it won’t – I’ll have heard about changes to tax codes well in advance of the strategy not working.

The performance of the passive fund is less certain. While over long periods of time I expect economic conditions to improve, it’s definitely possible that won’t happen over shorter periods. I wouldn’t be surprised that for anywhere from 3 to 10 years a passive fund’s index returns could be flat or negative. And it will likely be reduced by whatever the fund costs are. And I can’t predict when it will be positive, negative, or flat. 

The performance of the highly active fund is even less certain. It will have all the issues mentioned about the passive fund, with the additional layer of what drives the manager’s relative performance. And here is where consistency of causes becomes questionable. Sometimes relative performance is poor just because of bad luck. Fund managers retire and change jobs. Market environments change, which can pull the carpet out from beneath strategies premised on consistency or relationships. And, of course in the long run the market is an arbitrage killing machine. It’s always getting more efficient, commoditizing and driving down the cost of systematic strategies, which generally drives down their relative performance. So it’s very difficult to find a root cause that will be certain to persistent into the future.

Is it possible for the belief to be held universally and still be true?  

Any advocate of active management would probably agree in broad strokes with the above, but note that there are processes and principles one can use to pick managers who you expect to outperform in the future. But of course, we don’t all agree on those processes, and we can’t all invest in one manager. If we did, they’d become the entire market.

Active management outperformance requires diversity to persist. So the question of which active managers will outperform is by necessity always going to be a subjective opinion. If we all agreed who the best future active manager would be, they would become the market, and their advantage would disappear. Conversely, the more unconventional and concentrated the active manager or strategy you pick, the higher its potential for upside (or downside).

So if everyone believes the same active strategy will win, it probably won’t. A belief in a single active strategy is not a view that can be held universally without becoming wrong. 

In contrast, believing that markets go up on average, and that costs reduce returns, and that locating assets intelligently reduces taxes doesn’t require a strongly contrarian or diverse views. All investors can believe it, and it wouldn’t crowd out it’s own truth.

This is what I mean by universally – things we can all concurrently believe, act upon, and the belief would remain true and still perform will in the future. This doesn’t mean everyone does believe in it. Just that it would be possible to be universally believed, and still be true and performant.

Putting it to work

So how can we actually use these questions helping us make history more predictive of the future?

Is the root cause of performance clear?

Is the root cause consistent and persistent?

Is it possible for the belief to be held universally and still be true?

For me, it involves strongly focusing on things that have a strong yes to all three questions. That’s generally pushed me to invest my time and energy in:

  • Investing in myself (through learning, training, networking)
  • Trying to be a good friend, colleague and neighbor
  • Being good at my job, focus on providing a valuable service people pay for, work hard.
  • Keeping my personal and investment costs low, or only paying when the net value is certain to be positive.
  • Keep my taxes low by using tax loss harvesting, tax location, specific lot selling, etc. Always play by the rules, and always pay your share.

To be clear, I don’t think these will help me get rich quickly. I do think they’ll take work and perseverance. But I definitely believe that my past returns from investing in these activities will continue to be predictive of future success.


Learning curves

This is the first (brief) official post of this blog. And perhaps a good intro to talk about learning curves.

In March, I went snowboarding for the first time (I was 27 at the time). It was painful and humbling. Nothing prepares you for having to balance and manoeuvre on a completely new surface. By the end of the third day, I was ready to quit – every fall was like feeling all previous falls again at once. On the fourth day there was heavy snowfall – bad news for my skier friends, but at least the falls would hurt less. So I said (literally) “I’ll give it one last try”.

Low and behold, the heavy snow made all the difference. I could now turn and balance. I went down the mountain faster than my skier friend. At one point I simply was tired and so went off a jump into a snowbank. HEY! THIS IS FUN!

Revelation: I often think of myself as “learning” new things, mostly because I read them or think about them. However, this isn’t learning, at least not in terms of motor skills or concepts – it’s just being informed. Actual learning, let’s call it creating a new pattern of ability in your brain, be it snowboarding, surfing, mathematical or statistical concepts is often uncomfortable and frustrating. It involves doing something poorly (at least at first), and knowing you aren’t doing it well. But you have to remember that learning in any domain is accumulatory  – the first achievements are always the hardest, but subsequent achievements leverage previous ones.

I see three basic types of learning curves– increasing marginal gains, decreasing marginal gains, and mixed. Figuring out what drives each of these allows you to understand which one you are on, and what you can expect in the future. This makes sticking out the curve easier or harder, because you know if you just hold out a little longer, you can reach so much further.

learning curves
learning curves
1. Anders K. Attaining excellence through deliberate practice: Insights from the study of expert performance. The pursuit of excellence through education. 2002 ;21.