The majority of this article will deal with the financial investment world, but the principles are equally relevant to sports betting.
It is perhaps the greatest paradox in the investment world that many consistently profitable money managers have a large percentage of losing clients. I recently saw the records of a very successful US Hedge Fund, that showed over 40% of their lifetime client base had actually lost money while investing with the fund! This fund had a relatively consistent record of double digit annual gains over decades! This rather odd story is by no means isolated, it is repeated within many successful funds.
So, what causes this phenomena?
Bad timing and illogical emotion probably covers the answer. It is human nature to be tempted to buy when an investment is doing well and sell when doing badly. Jack Schwager in his cult classic Market Wizards, sums up “the common dual tendency of many people to initiate an account after a manager has already had a large winning streak and to liquidate in the midst of a drawdown is the single biggest blunder investors make”. Clearly if it the path to riches was as simple as to just invest in a fund that was currently outperforming, we would all be rich.
One of the greatest investors of all time, Jim Rogers is famous for his indifference to market timing. Rather amusingly, CNBC use a clip from an interview with Jim where he praises CNBC experts for their ability to time markets! The irony is apparently lost on the executives at the business channel. But human psychology is notoriously difficult to conquer and to invest, or continue to invest, in a methodology/system/service during a serious drawdown is counter-intuitive in the extreme.
To counter this natural psychological trait it would be useful to know the worst possible drawdown/losing run that could occur with any given strategy. But, how can we predict the worst likely drawdown/losing run? It is tempting to look at past results and simply believe that whatever the worst run in the past has been, this is the likely worst case scenario going forward. However, that is dangerously wrong. Your largest drawdown/losing run is always in the future and I will attempt to show you why.
The best tool we have for such prediction is a Monte – Carlo Simulator. The MC Sim is a relatively simple tool that takes a series of data and shuffles and repeats as many times as you wish. Each run of results is known as an iteration. I have used this extensively to show potential drawdowns for my two betting services which can be followed here at smartbet.
For example, for the 2010 World Cup, after assigning ratings for the teams and inputting the groups/draw, the MC Sim was able to run 10,000 iterations (or simulated World Cups) using the same probabilities for each game. Playing the tournament 10,000 times gave reliable % figures for the number of times each individual side could be expected to win the World Cup. For example, initially the model might suggest out of 10, 000 iterations, Spain would win 2,000 times, Brazil 1,800 times and perhaps even a lesser side such as Greece would win 10 times (effectively making them a 1000/1 shot). As the tournament progressed and the model had actual results and revised ratings for the teams, it could be constantly updated. Similarly we can use an MC Sim to examine results from a system/trading strategy and then run those results any number of times to calculate the % chance of various drawdowns/losing runs. The results are fascinating.
Here are the actual results from a trading strategy (could as easily be results of a sports betting system);
Win rate 44%
Compound Annual Return 31%
Maximum Drawdown 9.4%
MAR (Return/drawdown) 3.36
Sharpe Ratio 2.3 7
If we now run these results 10,000 times through a MC Sim we get some remarkable results. Despite the worst actual drawdown to date being 9.4%, the AVERAGE drawdown over the 10,000 iterations was just over 10%! Even more remarkable is that if you wanted to assume a 95% confidence level, then 95% of the time, the maximum drawdown was 15.6%! Most statisticians apply the 95% confidence level, but if you were to be more cautious still and look for a 99% confidence level then the maximum drawdown that could b e expected was 22%! Clearly these are far uglier results than the maximum drawdown of 9.4% that had been experienced in real – time!
The lesson is clear. If you have a successful strategy, that you have tested over a representative period of time, you should be aware of the worst case scenario (which can be calculated accurately via a monte – carlo simulation) and, armed with that knowledge, abandoning such a strategy during a confidence sapping drawdown, is the worst possible outcome. Unfortunately, it is also the outcome that human psychology is drawn to.