Killer Gas CO2
Climate change features prominently in headlines across the western world. Even more so in Germany after the green party to a massive chunk of the vote in recent EU elections. The theory as I understand it in very simplified terms claims that greenhouse gases like CO2 cause our climate to heat up. Therefore, reduction of CO2 emissions will save our climate.
Flash Of Insight
Now as a good citizen, I repeatedly try to understand and evaluate the science behind the climate change topic. Yesterday I watched another video to confirm my bias that man-made climate change is a hoax and I can keep breathing with good conscience. (Don’t bother watching unless you speak German).
The portrayed expert, professor of geology Werner Kirstein, explains how climate scientists came to the conclusion that CO2 is a leading indicator for earth’s climate. Looking at a 30 year stretch of historic data for temperature and CO2 concentration which coincidentally shows both rising in unison, experts deduced a causational relation.
Now I am not here to argue in favor or against climate change or the science behind it. I leave that up to others. But in flash of insight it struck me, that there are strong parallels to trading system development and backtesting.
Trading Systems And Backtesting
Anybody who at any point in his life was serious about trading has tried to develop and backtest a trading system using historical chart data. And anybody who tried also knows it is a difficult endeavor. There are various pitfalls to be navigated around before backtest results are feasible and trustworthy. Let’s take a look at two of them.
Backtesting With Too Little Data
One of the simplest and most obvious pitfalls is to use a set of data that is too small to cover all types of market conditions. Say you are backtesting a daily bar bullish breakout trading system on the Nasdaq 100 for the set of data between 2009 and 2019, you will likely have brilliant results and by projection become a billionaire within a couple of months. Backtest the same strategy on different markets, in other times and changing timeframes and you will likely see that your first billion will take a bit longer to achieve.
Manual Visual Dishonest Backtesting
Another mistake I have made for which I don’t know a proper term is that I looked at historic charts and marked entries and exits as I thought I would have taken them as the market evolved. Some of theses tests showed great results!
The only problem was that in my backtest I actually had a view of the market after the fact. I did not utilize a bar by bar replay to simulate the “hard right edge” of a chart. Guess what, I was never able to come even close to the beautifully profitable test results. It simply was not the same. My explanation is that in the backtest I was simply not able to ignore the way the market behaved after my entry. My brain always introduced a bias towards what the market did “in the future”. So there was no honest backtest. It always was a after-the-fact should-have-done trade. It is damn hard to even produce an honest backtest! (By the way, bar by bar replay is the best I can come up with to alleviate this problem.)
So I described two typical backtesting pitfalls that set me back in my studies of trading. The moment I listened to the video about climate change and that the scientists inferred from a 30 year data stretch that CO2 concentration and temperature have a causal relationship, I was struck by the similarities to my experience with trading system development. Could it be that the scientists made these novice mistakes with their data?
I saw how difficult it is to define a hypothesis about how prices will develop based on historical data and find a statistical tradeable edge. The odds are stacked against you in every respect.
I am certain for anyone to find an adequate all-encompassing theory about climate and causal relationships with whatever internal system parameter is exponentially harder and more complex. The likelihood that scientists make mistakes is so high that I remain extremely skeptical of the validity of their models.
In the best case, they are well meaning people blinded by their hybris. They think themselves fool-proof by credential and don’t even consider blind-spots in their process. Climate change is a high stakes game and it would be fun to watch their theories go up in flames some day, were it not such a politically heavyweight topic with the potential to destroy civilizations and generations until disproved.